Agents scan this library against open challenges, tasks, and questions and surface who to contact to resolve faster.
Aabhas Sharma — Chief Technology Officer, Hebbia
Signal: 1 (ICP writing about agent costs/reliability). Source: https://www.linkedin.com/posts/hebbia_hebbia-cto-aabhas-sharma-at-humanx-2026-on-activity-7455724160200282112-O-fW and https://www.hebbia.com/blog/why-i-joined-hebbia-as-cto. Company size: ~120-180 (Series C AI-native, enterprise document/retrieval agents for finance). Pain points: token efficiency and citation accuracy across financial-document agent workflows; making retrieval agents reliable/explainable at the scale of customer data. Challenges: building steerable, reliable, explainable agentic systems on large document sets; long-running agent workflows (speaking at Replay 2026 durable-execution conf). Must-haves: token/cost efficiency per agent run, high accuracy/reliability in production. Nice-to-haves: model-level efficiency gains (praised Opus 4.8 token efficiency). ICP confidence: High (technical decision-maker/CTO at AI-native company actively shipping agents in production, in target size band).
Signal: 4. Source: https://www.crunchbase.com/person/abhay-parasnis ; https://theaiagentindex.com/agents/typeface ; Tracxn Typeface profile. Company size: ~186 employees (San Francisco; ~$206M raised, Series B; enterprise content/marketing agents - 'Arc Agents' + Marketing Orchestration Engine, incl. Ideation Agent, Creative Agent, autonomous Brand Agent). Pain points (inferred; no direct quote this run): autonomous agents validating content against brand guidelines reliably; multi-agent orchestration for enterprise marketing; agent cost/governance at enterprise scale. Challenges: making multi-agent content workflows reliable and brand-safe. Must-haves: reliability, brand/governance control, observability. Nice-to-haves: agent cost visibility. ICP confidence: Medium (title is CEO, but a deeply technical founder - former Adobe CTO/EVP - fitting the 'technical co-founder' ICP; ~186-emp Series B company shipping marketing agents). LinkedIn URL not captured this run - do not fabricate.
Signal: 4 (ICP CTO at SaaS company shipping an agentic platform in production). Source: https://www.outreach.ai/ai-agents ; https://www.businesswire.com/news/home/20260427304135/en/ (Outreach launches Omni) ; https://tracxn.com/d/companies/outreach/. Company size: ~1,562 employees (May 2026); later-stage (Series G) revenue-execution SaaS. Actively shipping agents: Outreach is now positioned as the 'Agentic AI platform for revenue teams' — launched Omni (Apr 2026) with AI agents automating forecasting, coaching, deal execution and account expansion across 33M+ weekly interactions. Abhi Abhishek is CTO. Pain points (inferred from product + domain, not verbatim quotes): reliability of revenue agents acting on customer/CRM data; per-run cost and observability at very high interaction volume; controlling autonomous agent actions in a sales workflow. Challenges: scaling reliable, cost-efficient, governable agents across a large enterprise customer base. Must-haves: reliability, control/governance, per-run cost visibility. Nice-to-haves: observability + compounding quality across runs. ICP confidence: Medium (named CTO, 1,562 emp in band, agentic platform in production; caveat: Series G, beyond the A-C stage band).
Abhi Pathak — Chief Product Officer, Suki AI
Signal: 4 (ICP product leader at a qualifying agent company; found via web verification of a Jan 2026 executive appointment). Source: https://www.suki.ai/press-releases/suki-expands-executive-leadership-with-the-addition-of-abhi-pathak-as-chief-product-officer/. Company size: ~413 employees; healthcare AI — ambient documentation + agentic AI Assistant. Appointed CPO Jan 8, 2026; 15+ yrs product leadership (Amazon, Dropbox; most recently CPO at Accolade). Pain points (role-inferred): balancing product-level agent cost vs. reliability; scaling agentic product features across clinicians; per-user/per-interaction economics. Challenges: shipping reliable agentic features fast while keeping inference cost sustainable. Must-haves: visibility into agent cost per interaction; production reliability. Nice-to-haves: cost/perf optimization tooling. ICP confidence: High (product C-suite, AI-focused, at an agent-native 50-2,000-emp company). NOTE: pain points inferred from role/company context, not a verbatim quote.
Signal: 4 (ICP technical co-founder/CTO at agent-native company shipping agents in production). Source: Forbes (https://www.forbes.com/sites/davidprosser/2026/07/29/...), Crunchbase News, company about page; LinkedIn company page confirms size. Company size: 51-200 employees (LinkedIn: 51-200, 134 associated members; founded 2024, SF). Freehand = agentic AI studio for supply chain + finance; autonomous AI "Teams" replace manual procurement, supplier management, invoice checks, and payment ops for F500; reads unstructured data, reasons across contracts/policies, executes in enterprise systems; $75M Series B (Battery Ventures, NewRoad; $100M total); customers Meta, Unilever, J&J, Pfizer, Cardinal Health, Dunkin'. Manohar is ex-Co-Founder & CTO of Pando (TMS/procure-to-pay). Pain points: giving agents the judgment/context to act autonomously and reliably in high-stakes F500 spend workflows; reasoning across messy unstructured data; reducing supply-chain cost. Challenges: reliability + accuracy when agents execute financial decisions. Must-haves: production reliability, decision context, auditability. Nice-to-haves: per-run cost visibility. ICP confidence: High (Co-Founder & CTO at 51-200-person agent-native Series B company).
Signal: 4 (ICP CTO at agent-native healthcare-operations company). Found via web research + verification — LinkedIn content/people search UNAVAILABLE this run (session not authenticated / login wall). Source: https://www.qventus.com/company/newsroom/ (AI Solution Factory launch) ; https://www.linkedin.com/in/abhinavmittal1 ; https://profiles.crustdata.com/company/qventus. Company size: ~259 employees (June 2026); growth-stage healthcare AI (grew ~159 in 2024 to 259 in 2026). Actively shipping agents: Qventus provides AI "teammates"/Operational Assistants that automate hospital operations (patient flow, scheduling, capacity) and launched an "AI Solution Factory" to mass-produce custom operational agents for health systems. Abhinav Mittal is CTO (also titled SVP Engineering since Apr 2023; ex-VP Eng Aura, ex-CTO Nomis Solutions, ex-Intuit; BS CompE Wisconsin-Madison). Pain points (INFERRED): reliability of operational agents acting on hospital systems; per-run cost/observability as the agent fleet scales across health systems; keeping many custom agents governable. Challenges: scaling a fleet of reliable, cost-efficient operational agents. Must-haves: reliability, observability, per-run cost visibility. Nice-to-haves: cost optimization/routing. ICP confidence: High (CTO, technical, 259-emp agent-native healthcare company in band).
Abhinay Vyas — Co-Founder & CDO (Chief Data Officer), RapidClaims
Signal: 2 (non-ICP author's LinkedIn post surfacing an ICP company shipping agents in production — same source as co-founder Jot Sarup Singh added this run). Source: LinkedIn post by Raj P. (AI @ RapidClaims): RapidClaims has 'AI already live in production across US health systems' with 'in-house LLM infra, RL, fine-tuning, agents, evals, retrieval — with a relentless focus on latency, reliability, and cost' (company: https://www.linkedin.com/company/rapidclaims-ai). Abhinay Vyas is Co-Founder & CDO, leading data/ML strategy for the LLM-powered RCM agents (IIT Kharagpur; ex-Chief Data Scientist @ Phable). Company size: ~59-89 employees (Tracxn/PitchBook, 2026); Series A, $11.1M raised (Accel, Together Fund); founded 2023. Pain points: reliability + cost + eval of production medical-coding/denial agents; retrieval/data quality for healthcare RCM; latency at scale. ICP confidence: High (technical co-founder / C-suite data leader at AI-native agent company; headcount in range). Note: 2nd RapidClaims leader added this run (alongside co-founder & CPTO Jot Sarup Singh); pains inferred from the company's documented production-agent focus, not a personal quote.
Signal: 1 & 3 (ICP writing publicly about agent/LLM cost control; TrueFoundry's AI/MCP/agent gateway competes near Portkey & AWS AgentCore). Source: https://www.truefoundry.com/blog/ai-cost-optimization-strategies , https://www.truefoundry.com/blog/best-mcp-gateways , https://www.linkedin.com/in/abhishekch123 . Company size: ~123 (Series A, $19M led by Intel Capital; enterprise AI gateway + MCP/agent gateway; acquired Seldon AI 2026; ex-Meta Sr Staff Eng). Pain points: controlling agent/LLM spend across tool-connected execution; multi-model routing; observability & governance at enterprise scale. Challenges: extending cost control beyond model calls into autonomous agent workflows; rate limiting; secure MCP tool access. Must-haves: per-workflow cost visibility; agent gateway control plane; governance. Nice-to-haves: predictable spend forecasting; multi-cloud routing. ICP confidence: High (verified CTO, ~123 emp, Series A, AI-native building agent infra, actively publishing on agent cost/control).
Adam Guthrie — Co-Founder & Chief Technical Architect, Luminance
Signal: 4 (ICP technical co-founder at agent-native company). Source: https://www.luminance.com/about/ ; https://research.contrary.com/company/luminance ; Tracxn (437 employees as of Mar 2026). Company size: 437. Series C ($75M, early 2025; Point72, March Capital). Actively building AI agents: "Legal-Grade" multi-agent platform automating end-to-end legal workflows (creation → negotiation → risk review → compliance) for 1,000+ enterprises. Pain points (inferred from domain + market signal, NOT a verbatim quote): reliability at scale across many agents/workflows; agent orchestration; cost and observability across a multi-agent fleet. Challenges: keeping outputs "legal-grade"/auditable while scaling; institutional memory across agents. Must-haves: reliability, orchestration, observability. Nice-to-haves: compounding intelligence / institutional memory. ICP confidence: High (technical co-founder & chief technical architect; Series C; 437 employees; multi-agent platform).
Adam Seligman — Chief Technology Officer & GM, AI Incubation, Workato
Signal: 4 (ICP-titled technical leader publicly speaking/shipping on enterprise agents at scale). Source: https://www.businesswire.com/news/home/20250814122383/en/Workato-Launches-AI-Research-Lab-and-Accelerates-as-an-AI-First-Enterprise (AI Research Lab for autonomous enterprise agents; Seligman named CTO & GM AI Incubation); https://www.morningstar.com/news/business-wire/20260701133721/workato-launches-workato-labs-with-a-new-open-source-developer-toolkit-for-enterprise-orchestration (Jul 2026); https://www.agentconference.com/speaker/Adam-Seligman (AI Agent Conference 2026, NYC — speaker); Crunchbase person profile. Company size: ~1,517 employees worldwide as of Mar 2026 (Revelio Labs; grew from 969 in 2023) — inside the 50-2,000 band. Ships enterprise agentic orchestration: Workato Labs open-source toolkit, enterprise MCP layer, and an AI Research Lab operationalizing front-office (sales/marketing/CS) and back-office (finance/HR/IT ops) agents with customers. Pain points: operationalizing autonomous enterprise agents across many business functions; governing MCP/tool sprawl; making agent orchestration reliable across 700+ connected systems. Challenges: agent reliability and governance at enterprise scale; standardizing how agents access enterprise tools; demonstrating economics of agent-run work vs. deterministic integration. Must-haves: orchestration reliability, agent/tool governance, enterprise-grade observability. Nice-to-haves: per-agent cost attribution across customer tenants. ICP confidence: Medium (title, headcount and agent-shipping all fit strongly; downgraded from High because Workato is late-stage — Series E, ~$5.7B valuation — outside the stated Series A-C band, and no direct public quote on agent cost/token waste was found this run).
Adam Sypniewski — Chief Technology Officer, Deepgram
Signal: 3 (competitor/ecosystem-adjacent; voice-agent platform). Source: https://tracxn.com/d/companies/deepgram/ ; https://deepgram.com. Company size: ~326 employees (May 2026); Series C, ~$229M raised (Y Combinator, Tiger Global, Nvidia). Deepgram is a voice-AI platform (speech-to-text/ASR, TTS) that has moved into agentic voice with its Voice Agent API for building real-time production voice agents. CTO Adam Sypniewski (Scott Stephenson CEO; Noah Shutty co-founder). Pain points (inferred): real-time reliability + latency for production voice agents; cost per minute/interaction at high call volume; giving customers observability into deployed voice agents. Challenges: helping customers run reliable, cost-efficient voice agents at scale. Must-haves: reliability, latency, cost efficiency. Nice-to-haves: per-run cost/observability tooling. ICP confidence: Medium (named CTO; ~326 emp Series C in band; agentic voice product. Strong caveat: model/infra vendor — likely builds rather than buys an agent operating layer, and is competitor-adjacent; route to positioning review before outreach).
Adarsh Hiremath — Co-founder & CTO, Mercor
Signal: 4 (ICP technical co-founder/CTO at AI-native company building agent systems). Source: https://en.wikipedia.org/wiki/Mercor ; https://jobsbyculture.com/blog/working-at-mercor-2026 ; Bloomberg 2026-04-29 (mercor $10B valuation, ~200 FTEs). Company size: ~200 full-time (plus large contractor network). Stage: Series C, ~$10B valuation. Actively building agents: AI-driven recruiting/matching + powers frontier RLHF data and AI-agent training at scale; product uses AI agents for candidate matching/interviewing. Pain points (inferred from domain, NOT verbatim quotes): reliability and cost of running many AI agents/eval pipelines at scale; visibility into agent behavior across a large matching/interview workflow. Challenges: scaling agentic matching + data pipelines reliably; controlling model/agent spend as volume grows. Must-haves: reliability, cost visibility per agent run, observability. Nice-to-haves: compounding intelligence across runs. ICP confidence: Medium-High (technical co-founder & CTO; ~200 FTEs, Series C; agent-building core, though heavier on data/RLHF than pure production customer-facing agents).
Adi Azarya — Co-founder & VP R&D (Head of Engineering), Unframe (Unframe AI)
Signal: 4 (VP-level technical leader / Head of Engineering at a company shipping enterprise AI agents; web research this run). Source: https://www.calcalistech.com/ctechnews/article/sjzf1rfyzl ; https://www.unframe.ai/about . Company size: ~130 employees (Series B $50M led by Highland Europe, 2026; $100M+ enterprise TCV). Unframe builds an enterprise AI platform that turns business needs into operational AI solutions/agents and moves them into full-scale production. Adi Azarya co-founded Unframe and leads R&D (previously managed R&D activity at Noname Security). Pain points (inferred; no verbatim quote this run): engineering for production reliability/governance across many enterprise agent deployments; scaling agent infra; cost/observability. Challenges: shipping reliable, governed agents fast across heterogeneous enterprise environments. Must-haves: reliability, control/governance, scalable infra. Nice-to-haves: per-deployment cost/observability. ICP confidence: Medium-High (VP R&D = Head of Engineering, a named ICP role; technical co-founder at a ~130-emp Series B agent company). Profile URL not captured — do not fabricate.
Aditya Bansod — Co-Founder & CTO, Luma Health
Signal: 4 (ICP speaking publicly twice in 2026 on multi-agent orchestration and autonomy thresholds). Source: https://podcast.censinet.com/episode-206-healthcare-doesnt-break-in-the-exam-room-with-co-founder-and-chief-technology-office/ (Risk Never Sleeps Ep. 206, 9 Apr 2026) and https://thebigunlock.com/2026/03/02/turning-ai-hype-into-healthcare-execution/ (2 Mar 2026). Company size: ~197-206 (LeadIQ Jul 2026; ZoomInfo band 51-200). Stage: Series C — $130M led by FTV Capital, Nov 2021, $160M total. FLAG: the Series C is ~5 years old and no later round was found. Pain points (from published episode descriptions; verbatim transcripts unavailable): "AI's challenge in healthcare isn't ambition, but execution... most remain point solutions that fail to integrate into the complex workflows." On orchestration: "AI agents coordinate across workflows to ensure patients complete their care journey." On autonomy: advocates "exception-based intervention, in which people step in only when automation encounters ambiguity," with "flexible guardrails, allowing health systems to calibrate automation based on confidence thresholds and maturity." Episodes explicitly cover "the rise of agentic AI, the tension between human-in-the-loop oversight and autonomy." Challenges: setting and tuning confidence thresholds per health system; knowing when an agent run should escalate vs proceed; coordinating agents across multi-step patient journeys. Must-haves: confidence-scored run outcomes; exception/escalation telemetry; cross-workflow agent trace visibility. Nice-to-haves: per-customer cost and volume reporting; automated threshold tuning from production data. ICP confidence: Medium (leaning High) — exact title and verified headcount, active 2026 public commentary on agent orchestration; downgraded on the aging Series C and pain sourced from episode descriptions rather than transcript quotes. Cost/token pain inferred.
Aditya Sundararam — Chief Product Officer (listed as Head of Product on company leadership page), DataBahn
Signal: 1 (ICP quoted publicly on agents needing grounded enterprise context).
Source: https://www.prnewswire.com/news-releases/databahn-launches-federated-search-and-orchestration-...-302842675.html (Aug 2026); title also verified on https://www.databahn.ai/leadership (where it reads "Head of Product" — the Aug 2026 PR uses "Chief Product Officer").
Company size: ~112 employees as of 30 Jun 2026 per Tracxn (search snippet, medium confidence). Series B $40M Jul 2026 led by Insight Partners. Estimate 100–130 — inside band.
Pain points: (verbatim) "Enterprises have spent the past decade moving security data from one system to another simply to ask questions of it… Security teams and AI agents can now search data across every source, retrieve answers grounded in enterprise context, and turn those answers into completed investigations."
Challenges: Agents can't reach the right context without expensive data movement; investigations require multi-source retrieval that inflates token consumption per run.
Must-haves: Context retrieval that doesn't blow the token budget; grounding without copying data everywhere.
Nice-to-haves: Cost-per-investigation attribution.
ICP confidence: Medium — product-leadership seat at an agent-shipping Series B inside the headcount band, with a real on-topic public quote; downgraded from High because the title is inconsistent across the company's own two sources and the quote is context-focused rather than cost- or reliability-focused.
Watch-out: DataBahn markets token-cost reduction itself — partial competitive overlap.
Adrian Buzgar — Engineering Director, Taktile
Signal: 4 (ICP writing publicly about building/shipping agents). Found via web research — Claude-in-Chrome/LinkedIn NOT connected this run.
Source: https://engineering.taktile.com/blog/it-has-never-been-more-exiting-to-be-a-builder/ (19 Feb 2026, co-authored with Principal Engineer Robin Raymond); leadership listing at https://taktile.com/about
Company size: ~224 employees (Tracxn, 30 Jun 2026). Series C — $110M led by Growth Equity at Goldman Sachs Alternatives, announced 24 June 2026. Berlin + NYC. Product is an "Agentic Decision Platform" with an AI Agent Manager and Agent Library shipping agents for AML, KYC/onboarding, claims and SMB underwriting.
LinkedIn URL: https://www.linkedin.com/in/adrianbuzgar/ (listed directly on Taktile's official About page)
Pain points: "We have power-users who burn through token quotas and there are folks who never run out" — uneven/unmetered token consumption across engineering teams. "agentic AI massively accelerates everything, including how painful your bottlenecks are." Cites a Pragmatic Summit stat that "some orgs see 2x production incidents post AI adoption" and adds: "Given the types of sensitive data we process at scale, for us that's clearly not an acceptable outcome." Flags a 4x widening PR-cycle-time delta between teams as evidence of uneven agent leverage. No single view of which teams/agents are burning budget.
Challenges: Scaling internal agent experiments ("Claudius") AND a customer-facing multi-agent product line simultaneously; proving agent reliability to banks and insurers; measuring velocity/cost per team without inventing bespoke instrumentation.
Must-haves: Per-team and per-agent cost/token visibility; production reliability and incident-rate guardrails; audit trail suitable for financial-services compliance; coverage across both internal dev agents and customer-facing product agents.
Nice-to-haves: Velocity/DORA-style metrics tied to agent usage; routing/model-selection controls; cross-team benchmarking.
ICP confidence: High — Director-level engineering leader at a Series C, ~224-person fintech vertical AI company that ships agents in production, publicly writing about token burn and agent-driven production incidents.
Adrian Hupka — Head of Engineering, Tacto (Tacto Technology GmbH)
Signal: 1 (ICP author). LinkedIn post, 88 reactions / 4 comments.
Source: https://www.linkedin.com/search/results/content/?keywords=%22our%20agent%20fleet%22&datePosted=%22past-month%22&sortBy=%22relevance%22
Company size: ~150–153 employees (Tracxn, 30 Jun 2026; PitchBook ~151, Jul 2026). Munich; industrial procurement intelligence SaaS, founded 2020, $59.7M raised over 3 rounds.
Pain points (his own words): "For five months now we have been building with the attitude that 'Coding is Solved.'" / "we officially retired the role of Software Engineer. One role is left: Builder." / "A year ago, with roughly the same team, we ran four high-impact initiatives in parallel. Right now we run fourteen, and team size is converging toward 1-2 people instead of 5-7." / "That is how we bring our agent fleet into the processes customers actually run."
Challenges: Coordinating an agent fleet deployed into live customer industrial/procurement processes (forward-deployed); deciding which bespoke agent work should be promoted into the platform; maintaining oversight as team size per initiative drops 3–5x.
Must-haves: Visibility and control over a multi-agent fleet running inside customer processes; a way to standardise repeating agent patterns into platform.
Nice-to-haves: Per-initiative cost/outcome attribution; FDE-friendly deployment model.
ICP confidence: High — Head of Engineering (ICP title), Series-B-scale AI-native SaaS at ~150 people, agent fleet demonstrably in customer production, and he is publicly the owner of the operating model. Classic 1→many agents scaling profile.
Signal: 4 (ICP technical leader at agent-native company; found via LinkedIn currentCompany people-search of Dataiku engineering leadership). Source: https://www.linkedin.com/in/lavoillotte/ (via https://www.linkedin.com/search/results/people/?currentCompany=%5B%222770554%22%5D&keywords=VP%20engineering). Company size: ~1,000-1,200 (Dataiku, Series F enterprise AI platform; "Everyday AI" + LLM Mesh + agentic AI framework orchestrating many models/agents in production). Pain points (INFERRED from role+company stage, not an individual quoted post): governing cost/token spend across LLM Mesh where many models & agents run per enterprise workflow; per-agent/per-run cost visibility across a large multi-model estate; reliability of non-deterministic agentic workflows at enterprise scale. Challenges: standardizing observability & guardrails across a broad agent/model portfolio; keeping frontier-model spend justified vs cheaper routing. Must-haves: per-run cost attribution + control across LLM Mesh; production reliability/eval for agent workflows. Nice-to-haves: automatic model routing/cost optimization; unified agent cost dashboards. ICP confidence: High (VP Engineering, technical, at in-range 50-2,000-emp company actively shipping agentic AI).
Advith Chelikani — Co-Founder & CTO, Pylon
Signal: 4 (ICP technical co-founder shipping production agents). Source: https://www.usepylon.com/blog/ai-agents-v2 ; https://www.ycombinator.com/companies/pylon-2 . Company size: Pylon ~50-100 employees; Series B (General Catalyst, Bain Capital Ventures); 1,500+ B2B customers (Retool, Hightouch, Deel). Pain points: shipping agentic B2B support that reliably investigates/resolves/acts across channels; AI Agents v2 quality & cost. Challenges: multi-step agent reliability on real customer actions; scaling from single to many agents in production. Must-haves: production reliability & control over agent actions; cost visibility per resolution. Nice-to-haves: seamless human+agent collaboration workflows. ICP confidence: High (CTO/technical co-founder at Series B company whose core product IS production agents).
Signal: 4 (ICP shipping agents). NOTE: LinkedIn/Chrome NOT connected this run — found via web research. LinkedIn URL NOT captured (not fabricating one); profile_url is The Org leadership page where the title was verified. Source: https://theorg.com/org/cytora/teams/leadership-team (CTO, verified by direct page fetch) ; product launch https://www.cytora.com/risk-flow-center/blog/cytora-autopilot-risk-workflows-that-run-themselves ; https://fintech.global/2026/03/18/cytora-unveils-end-to-end-ai-automation-for-insurers/ ; https://www.reinsurancene.ws/zurich-insurance-scales-cytora-ai-platform-across-global-underwriting-operations/. Company size: ~132-138 employees (PitchBook 132; Tracxn 138 as of Mar 2026). London, UK. ~$43.2M raised. Agentic AI in production confirmed: Cytora Autopilot launched March 2026 — agentic workflow orchestration running end-to-end risk workflows without direct human intervention, persistent context across communications over extended periods; scaled at Zurich Insurance; Markel reported +113% underwriting productivity. Pain points (INFERRED from company-published material, NOT personal quotes): agents that hold persistent context across multi-week, multi-message workflows — a direct context/token-growth cost driver; agents must keep running unattended and recover when new data arrives. Challenges: long-horizon agent runs where context re-reading compounds cost; enterprise-grade reliability for Tier-1 insurer deployments (Zurich, Markel). Must-haves: visibility into cost of long-running/persistent-context agent runs; unattended reliability and failure recovery. Nice-to-haves: context pruning/compaction to stop token growth over long workflows. ICP confidence: High — CTO title, 132-138 employees (in band), agentic product live with named enterprise customers. UNVERIFIED: LinkedIn URL, personal pain statements.
Signal: 4 (ICP CTO at a mid-size insurtech actively shipping AI agents in production). Source: https://covergo.com/news/ai-agents-for-insurance-automation/ ; https://fintech.global/2026/03/06/covergo-launches-ai-agents-for-insurance-operations/ ; https://craft.co/covergo/executives . Company size: ~233 employees (as of Apr 30 2026). Stage: Series A ($15M raised); independent, global (HQ Hong Kong). Actively building/shipping agents: Feb 2026 launched AI agents for insurance with THREE agents already deployed in production with tier-1 insurers/brokers — document processing, customer support, and quotation generation across the policy lifecycle. Role: Ahmad Mosa, CTO (20+ yrs insurtech; prior CTO of a SaaS insurtech). Pain points (INFERRED from public positioning + agent launch, NOT verbatim): moving multiple agents from launch to reliable production across regulated insurance workflows; auditability/accuracy of agent outputs acting on policies/quotes; visibility/control as agent count grows across the value chain. Challenges: reliability & compliance of autonomous agents in a regulated, no-code platform serving many carriers. Must-haves: reliability, guardrails/governance, observability across a fleet of insurance agents. Nice-to-haves: cost efficiency per run, cross-carrier scalability. ICP confidence: High (named CTO, 233 emp, Series A, 3 agents confirmed in production).
Ahmed Achchak — Co-Founder & CEO (technical, ex-Datadog), Qevlar AI
Signal: 4 (web-research proxy — LinkedIn/Chrome unavailable this run). Source: https://www.unite.ai/qevlar-ai-raises-30m-to-transform-security-operations-centers-with-autonomous-ai/ ; company profile https://tracxn.com/d/companies/qevlar-ai . Company size: ~70 employees; Paris; founded 2023; ~$30M raised. Technical founder: AI engineer, former Cloud Security PM at Datadog. Co-founder/CTO: Hamza Sayah (also in Brain).
Pain points (from company positioning): running autonomous SOC agents that reliably investigate every alert at analyst depth; accuracy + trust in production autonomous decisions.
Challenges: scaling autonomous investigations without false pos/neg; throughput and cost per investigation.
Must-haves: reliable, auditable agent decisions at full alert volume.
Nice-to-haves: cost-per-run visibility.
ICP confidence: Medium — technical co-founder/CEO at a confirmed 50-2,000-emp company shipping production agents; added alongside CTO, so slightly lower priority to avoid over-indexing one company.
Aidan Gomez — Co-founder & CEO, Cohere
Signal: 4 (co-founder building & shipping an enterprise agentic platform, North). Source: https://cohere.com/blog/north-eap ; https://www.infoworld.com/article/3757962/cohere-goes-north-with-agentic-ai.html ; https://www.upstartsmedia.com/p/cohere-ceo-exclusive-north-agents . Company size: ~500-800 employees (grew from ~250 in mid-2024 to 800+ by early 2026) — in range. Built North (agentic platform) over 18 months; enterprise pilots with RBC, Dell, Bell, Ensemble Health. Technical co-founder (transformer co-author). Pain points: agent reliability, security, and cost/performance at enterprise scale; competing with Copilot/Vertex on efficiency. Challenges: reliable, secure, cost-efficient agent deployment across regulated enterprises. Must-haves: secure + reliable + cost-efficient agent execution. Nice-to-haves: governance/observability. ICP confidence: Medium — foundation-model lab building an agent platform (North); in-range headcount; technical co-founder, though as a model lab likely builds infra in-house (lower buyer intent).
Signal: 4 (technical leader at company building agents). Source: https://www.linkedin.com/in/aidensjlee/ ; company hiring Director of Engineering, Agentic AI (jobs.techstars.com/companies/twelve-labs). Company size: ~180-192 employees (May 2026); Series B ($100M Series B; $207M total). Building agentic video-understanding platform with agent-facing APIs. Pain points: building production-grade agent-facing infra/APIs; cost of video + LLM inference at scale; reliability of agentic video understanding. Challenges: scaling agentic platform and developer experience to production. Must-haves: production-grade agent infra, auth/authz. Nice-to-haves: token/cost efficiency. ICP confidence: Medium (technical co-founder/CTO at Series B; agents adjacent to core video product).
Signal: 4 (ICP senior technical leader at an agent-native company; discovered via LinkedIn people-search of companies actively shipping AI agents) | Source: https://www.linkedin.com/in/ajaychoudary | Headline: Director of Engineering at Kore.ai (enterprise agentic AI platform — multi-agent, voice, RAG) | Company: Kore.ai — actively building/shipping AI agents in production | Company size: ~1,000–1,200 (estimate) | Pain points (INFERRED from role/company context — not a verbatim quote observed this run): agent cost control and per-interaction spend visibility at scale; reliability of agents in production | Challenges: operating many agents reliably and cost-efficiently in enterprise deployments | Must-haves: cost-per-run observability; reliability/guardrails | Nice-to-haves: token-budget tooling; automated regression detection | ICP confidence: High (senior technical/eng leadership at a 50–2,000-employee company shipping AI agents)
Ajeet Grewal — Head of Voice AI (Leading Voice AI), Sierra
Signal: Bucket 4 (ICP writing/building agents) — found via LinkedIn people-search proxy for senior technical AI leaders at agent-native companies; not a specific post. Source: https://www.linkedin.com/in/ajeetgrewal (SF Bay Area). Company size: ~200-500 (Sierra, independent, Bret Taylor & Clay Bavor's customer-experience AI agent company, Series B/C). Pain points (INFERRED from role leading Voice AI at high-volume CX agent co; not a quoted statement): voice-agent reliability in production, latency + LLM cost per conversation at scale, no per-run visibility into agent cost/behavior. Challenges: scaling reliable multi-turn voice agents with guardrails. Must-haves: production reliability + cost-per-run visibility for voice agents. Nice-to-haves: unified eval/observability across human+AI agents. ICP confidence: High — Head/Lead-of-Voice-AI at a 50-2000 emp agent-native company actively shipping agents. Caveat: found via title search, pains inferred (no fabricated quote).
Signal: 4 (CTO at healthcare AI company expanding into agentic/platform capabilities). Source: https://www.suki.ai/news/suki-makes-executive-appointments-due-to-rapid-company-growth/ ; https://www.linkedin.com/in/amwarrier/ ; Tracxn Suki profile (2026). Company size: ~250-350 employees (well-funded, ~$165M+ raised; Punit Soni CEO). Role: CTO overseeing Product Development & Engineering; PhD Neuroscience (UC Davis); previously CTO at Symphony CrescendoAI. Product: Suki Assistant — voice-enabled ambient clinical documentation & reasoning (notes, orders, instructions); the platform business (agentic/API 'Suki Platform') is growing to >50% of revenue in 2026. Pain points (inferred from product domain, NOT verbatim quotes): reliability/accuracy of clinical reasoning agents; latency and per-interaction cost of voice+LLM at scale; safe autonomous action in a clinical setting. Challenges: making agentic clinical workflows reliable and cost-efficient. Must-haves: reliability, accuracy, cost/latency control. Nice-to-haves: observability of per-run cost. ICP confidence: Medium (named CTO, ICP title, company clearly in size band; CAVEAT — Suki is primarily ambient documentation and only expanding into agentic/platform capabilities, so the '5+ agents in production' fit is partial. Also note: another Suki leader, CPO Abhi Pathak, is already in the People Library). LinkedIn captured; verify current tenure before outreach.
Akarsh Mishra — Head of AI Products & Agent Systems, TrueFan AI
Signal: 4 (LinkedIn people search — ICP title "Head of AI Products & Agent Systems" + agents-in-production). Source: https://www.linkedin.com/search/results/people/?keywords=Head%20of%20AI%20agents%20production (profile: https://www.linkedin.com/in/akarsh-mishra-86722a215/). Company size: ~116 employees (Series A, Gurugram; ~$21M raised, Baring PE India / Z3Partners). B2B generative-AI video platform — leads agentic systems and multimodal pipelines for enterprise campaigns (100+ enterprise clients incl. HDFC, Bajaj Finance; can generate 500k videos/min in 175+ languages). Pain points (inferred from role/company): running agentic + multimodal pipelines at very high volume; cost per generation/run at scale. Challenges: reliability and cost control across multi-step agent pipelines serving enterprise campaigns. Must-haves: per-run cost/observability across agent + media pipeline; reliability at burst scale. Nice-to-haves: token/compute right-sizing. ICP confidence: High (Head of AI/Agent Systems, AI-native Series A, 50–2,000 emp, shipping agents).
Akash Magoon — Co-founder & CEO (ex-CTO, Nayya), Adonis
Signal: 4 (technical co-founder shipping autonomous agents in production; also Signal 1 — speaks publicly, incl. Agent Conference). Source: https://www.prnewswire.com/news-releases/adonis-raises-40m-series-c-to-equip-healthcare-providers-with-aidriven-revenue-cycle-operations-302722199.html ; https://www.agentconference.com/speaker/Akash-Magoon ; https://www.alleywatch.com/2026/03/adonis-ai-healthcare-payments-revenue-cycle-orchestration-platform-akash-magoon/ | Company size: ~80+ employees; Series C $40M (Mar 2026, led by Quadrille; General Catalyst, Bling), $95M+ total; founded 2022. Adonis = AI orchestration platform for healthcare revenue-cycle management; deploys Intelligence + AI Agent products that autonomously monitor, detect issues, recommend actions, and progress claims to resolution. 4x revenue growth 2025, NRR >130%. | Pain points: (inferred) health systems fighting rising denial rates need autonomous agents that reliably progress claims; production reliability/accuracy in a regulated, high-stakes billing workflow; auditability of agent actions. | Challenges: scaling many autonomous RCM agents reliably across payers; cost/observability per claim (per-run) as volume grows. | Must-haves: production reliability + audit trail; per-run cost visibility. | Nice-to-haves: agent evaluation, faster iteration. | ICP confidence: High — technical co-founder (previously Co-founder & CTO at Nayya) now CEO at a ~80-emp company shipping 5+ agents in production; ICP explicitly permits technical co-founders/C-suite.
Akash Singh — Co-Founder & CTO, Observe.AI
Signal: 4 (ICP building/shipping AI agents; found via web research — Claude-in-Chrome/LinkedIn NOT connected this run, so prescribed LinkedIn post/people searches were unavailable; pivoted to web research + primary-source verification, consistent with recent prior runs). Source: https://tracxn.com/d/companies/observeai + https://www.observe.ai/ (co-founder/CTO of Observe.AI; note: co-founders Swapnil Jain (CEO) and Jithendra Vepa already in People Library, but Akash Singh was NOT — deduped against full ~781-person library). Company size: ~348 employees (Jun 2026); ~$214M raised (~Series C); Oakland/Bengaluru; builds autonomous voice AI agents + agent-assist for contact centers. Pain points (inferred from company focus, NOT a verbatim quote): reliability and accuracy of autonomous voice agents at high call volume; cost/latency per interaction across 300+ enterprise deployments; observability into agent behavior in production. Challenges: making voice agents trustworthy + cost-effective at scale; eval + monitoring. Must-haves: production reliability, cost control per agent run, low latency. Nice-to-haves: better trace-level observability/eval tooling. ICP confidence: High (technical co-founder/CTO at in-band ~348-emp company actively shipping voice AI agents in production).
Akhil Bavisi — Director of Engineering, Gupshup
Signal: 4 (ICP engineering leader at an agent-shipping company; company-scoped LinkedIn people search "Gupshup director engineering AI agents").
Source: https://www.linkedin.com/search/results/people/?keywords=Gupshup%20director%20engineering%20AI%20agents (profile https://www.linkedin.com/in/akhil-bavisi-a360651/)
Company size: Gupshup ~1,000-1,500 employees (Series F conversational-messaging + agent platform: Conversation Cloud, ACE LLM, agent builder). Under 2,000. Stage later than A-C (flagged); brain already has 2 other Gupshup people (Kunal Patke, Surendranath C) but not this Director.
Pain points [INFERRED from role + company, NOT a verbatim post]: scaling conversational/agentic messaging across very high message volumes; cost-per-conversation control; production reliability of multi-channel agents.
Challenges: holding reliability + unit economics as agent conversation volume scales globally.
Must-haves: per-conversation/per-run cost visibility; reliability guardrails.
Nice-to-haves: routing/model optimization to reduce per-message inference cost.
ICP confidence: Medium-High — Director of Engineering at an agent-native messaging platform in the size band; stage later than A-C (flagged).
Akilesh Bapu — Head of AI Product Development, Hightouch
Signal: 4 (Head of AI at company shipping agents; web research this run — LinkedIn/Chrome not connected). Source: https://job-boards.greenhouse.io/hightouch/jobs/5970143004 ; company/role context https://rocketreach.co/hightouch-management_b43db6d7c19ca279 ; https://unicornscreener.vc/blog/7-ai-agent-startups-funded-by-top-vcs-in-2026. Background: co-founder & former CEO of DeepScribe (medical AI); now leads Hightouch's AI product strategy for autonomous, context-aware marketing agents. Company size: ~450-600 (Hightouch, Series D). Pain points: building autonomous context-aware marketing agents that act reliably on enterprise customer data; cost of running many agents continuously; proving outcomes. Challenges: turning a CDP/data-activation platform into an agentic decisioning product; reliability, brand-safety, and governance of agent actions. Must-haves: reliability, cost/observability per run, guardrails. Nice-to-haves: faster agent iteration. ICP confidence: High (Head of AI, AI-focused product leadership, at 50-2,000-emp agent-shipping company). LinkedIn URL not captured this run — do not fabricate.
Signal: 4 (technical co-founder/CTO publicly posting about agent infrastructure at a qualifying company). Source: https://techcrunch.com/2026/02/11/ai-inference-startup-modal-labs-in-talks-to-raise-at-2-5b-valuation-sources-say/ ; https://modal.com/company ; his LinkedIn activity incl. https://www.linkedin.com/posts/akshat-bubna-188885103_were-excited-to-introduce-modal-batch-the-activity-7331403762755887105-abXv. Company size: ~100-200 (Modal, Series C AI infrastructure; sandbox product reported to run hundreds of thousands of concurrent environments for coding agents and RL pipelines). Pain points: agent sandboxes and inference are a hard cost/perf tradeoff — Modal's own pitch is automatically selecting the most cost-effective GPU capacity across clouds, i.e. cost-per-run is the product. Challenges: reliability and cost of very high-concurrency, long-running agent workloads; multi-cloud capacity arbitrage. Must-haves: cost-efficient, reliable execution substrate for agent workloads. Nice-to-haves: per-workload cost attribution for customers. ICP confidence: Medium — technical co-founder/CTO at a 50-2,000 employee AI-native company deeply in the agent-infrastructure space, but he is a potential partner/adjacent-vendor rather than a pure buyer, and no direct pain quote about his own agent spend was found. Qualify positioning before outreach.
Signal: 1/4 (technical leader at a company actively building & shipping voice agents; found via web research since LinkedIn/Chrome not connected this run). Source: https://techcrunch.com/2026/07/31/smallest-ai-raises-13m-to-build-ultra-fast-voice-ai-that-sounds-genuinely-human/ ; https://laffaz.com/smallest-ai-series-a-seligman-ventures/ ; https://siliconangle.com/2025/10/09/exclusive-voice-ai-developer-smallest-ai-nabs-8m-investment/ . Company size: ~60 employees (founded late 2024; raised $8M seed led by Sierra Ventures + $13M Series A led by Seligman Ventures, 2026). Full-stack real-time voice AI platform owning the whole conversational loop (waveform→response) so enterprises can build/deploy/scale human-like voice agents. Pain points (inferred from product focus; no verbatim quote captured this run): sub-second latency + cost per call at scale, keeping real-time voice agents reliable in production, controlling inference/token spend on high-volume concurrent calls. Challenges: full-stack ownership of STT/LLM/TTS to hit latency and unit economics; scaling concurrent voice agents reliably. Must-haves: low latency, low cost-per-run, production reliability. Nice-to-haves: model efficiency gains, per-run cost visibility. ICP confidence: High (CTO/technical co-founder at a 50+ emp, Series A voice-agent company shipping agents in production; ex-Bosch engineer, IIT Guwahati). Profile URL not captured this run — do not fabricate.
Akshay Buddiga — Co-Founder & CTO, Traba
Signal: 4 (technical co-founder at agent-building company; company surfaced via a Traba "Staff Software Engineer (AI Agents)" hiring post describing their agentic platform, leader then found + web-verified). Source: https://www.linkedin.com/search/results/people/?keywords=Traba%20co-founder%20CTO (profile https://www.linkedin.com/in/akshaybuddiga/). Company size: ~173 (confirmed PitchBook 2026) — Traba, Series A ~$49M (Founders Fund, Khosla, General Catalyst); building an agentic platform (harnesses, evals, orchestration, model strategy) for industrial workforce operations. Pain points [INFERRED from verified role + company's own hiring post, NOT verbatim]: scaling agents 0->1->many in production; agent reliability ("survive contact with reality"); eval/observability; cost + model strategy. Challenges: building reliable agent harness/orchestration/evals at scale; on-call reliability. Must-haves: agent reliability + eval/observability infra. Nice-to-haves: cost-per-run visibility; model routing. ICP confidence: High (technical co-founder/CTO, agent-native platform, ~173-emp Series A; NET-NEW company).
Signal: 4 (ICP technical leader building/shipping agents) + 1 (agent reliability/cost in production). Source: https://theorg.com/org/skit/org-chart/akshay-deshraj ; https://tracxn.com/d/companies/skitai ; https://www.prnewswire.com/in/news-releases/voice-ai-company-vernacular-ai-rebrands-to-skit-secures-series-b-round-of-usd-23-million-from-westbridge-capital-897074893.html . Company size: ~104-400 employees (Skit self-reports 400+; Tracxn ~104 May-2026); Series B ($23M, WestBridge/Exfinity/Kalaari); augmented voice-AI agents for contact centers & collections, 60+ enterprise customers, IIT-Roorkee founder. Pain points (INFERRED): cost-per-call/token economics of voice agents at scale; production reliability/accuracy of agentic voice; no per-run visibility into what each voice agent costs. Challenges: scaling agentic voice across many enterprise deployments and languages while controlling cost and keeping reliability high. Must-haves: per-agent/per-run cost visibility; production reliability & control. Nice-to-haves: compounding quality over time; multi-model routing. ICP confidence: HIGH (Co-founder/CTO, 100-400-emp Series B voice-AI-agents co actively shipping agents in production; worldwide/India).
Akshay Pushparaja — Director of Engineering (Generative AI Products), C3 AI
Signal: 4 (Director of Engineering for GenAI/agentic products; found via LinkedIn people-search 'Director of Engineering AI agents'). Source: https://www.linkedin.com/in/pakshay/. Company size: ~1,060 associated members (self-reported bracket 1,001-5,000; actual headcount ~1,000, within the 2,000 ICP cap). Building agents: yes - C3 ships C3 Agentic AI enterprise applications. Pain points (inferred from role/company): scaling agentic AI products in enterprise production; reliability & cost of GenAI features at scale. Challenges: enterprise-grade reliability + cost control for agentic products. Must-haves: cost + reliability observability per agent run. Nice-to-haves: standardized agent eval/verification. ICP confidence: Medium (Director of Eng at ~1,000-emp AI-native company shipping agentic products; upper-bound size borderline, pain inferred).
Signal: 4 (ICP building/shipping agents in production — found via 2026 research on production-AI-agent companies). Source: https://rasa.com/calm , https://2026.rasa.com/ (video series "The architecture of production AI agents"), TechCrunch Series C coverage. Company size: 180 employees (Feb 2026); Series C ($30M round, ~$83.8M total). Actively building AI agents: Rasa's CALM architecture is a production conversational-agent platform. Pain points (inferred from public positioning, not verbatim quotes): LLM agents behaving non-deterministically / hallucinating in production; enterprises need predictable results. Challenges: making enterprise agents trustworthy and reliable at scale without giving up LLM fluency. Must-haves: reliability, deterministic control over agent behavior in production. Nice-to-haves: cost control / efficiency at scale. ICP confidence: High (technical co-founder & CTO, 180-person Series C AI-native company whose product is production agents; publicly writes about agent reliability).
Signal: Bucket 4 (ICP senior product leader at an agent-native company shipping agents in production; found via ICP-title people search + web verification).
Source: https://theorg.com/org/decagon/org-chart/alan-yiu ; https://decagon.ai/videos/why-i-joined-decagon-alan-yiu-vp-of-product
Company size: Decagon ~150-250 emp, Series C, AI customer-service agents (agent-native).
Pain points (INFERRED from verified role+company, not verbatim): scaling many concurrent customer-service agents in production; per-conversation / per-resolution cost visibility; enterprise-grade reliability & accuracy; proving outcome-based ROI to enterprise buyers.
Challenges: controlling agent cost per resolution at scale; reliability across concurrent agent runs; defending outcome-based pricing with hard unit economics.
Must-haves: production reliability, cost-per-resolution visibility, guardrails.
Nice-to-haves: honest model comparison, replay/simulation of agent runs.
ICP confidence: High — agent-native, Series C, right size; senior AI-focused VP of Product (prev VP Product @ Glean launching Enterprise Search + AI Assistant; Director GenAI @ Meta).
Note: Alan Yiu is net-new; Decagon already in brain via other leaders (Zhang, Sreenivas, Cui, Ha, Liu) — this person not previously captured.
Signal: 4 (ICP technical co-founder at agent-native security company running many agents). Source: https://simbian.ai ; https://tfir.io/simbian-has-an-army-of-ai-agents-with-top-security-skills-and-tools-ambuj-kumar/ ; https://tracxn.com/d/companies/simbian/ . Company size: ~61 (as of May 2026). Series A ($10M raised). Independent. CEO/co-founder Ambuj Kumar (ex-Fortanix founder); CTO/co-founder Alankrit Chona. Actively building AI agents: Simbian fields an "army" of security AI agents — SOC alert-response agents, threat-hunting agents, GRC agents — functioning as independent virtual employees, i.e. clearly 5+ agents in production. Pain points (inferred from domain + public positioning, not verbatim): orchestrating and controlling a fleet of autonomous security agents reliably; agent trust/accuracy in high-stakes SOC workflows; cost/observability as agent count scales. Challenges: reliability and auditability of autonomous agents acting on real security alerts; scaling many specialized agents. Must-haves: reliability, control/guardrails, observability across the agent fleet. Nice-to-haves: cost efficiency per agent run, compounding institutional memory. ICP confidence: High (technical co-founder & CTO of agent-native company with a multi-agent fleet in production; ~61 employees, just inside target band).
Alberto Rivera Martínez — Head of Agents & AI Developer Experience, Vic.ai
Signal: 4 (ICP Head-of-Agents technical leader at a company confirmed to be actively building AI agents). Source: LinkedIn profile above — found via people search scoped to Vic.ai ("Vic.ai autonomous accounting head of engineering"). Company: Vic.ai, autonomous accounting/finance AI (self-driving invoice & AP agents). Company size: 51-200 employees (confirmed via LinkedIn company page, New York). Pain points (inferred from role + company, NOT a verbatim quote): owning the reliability and dev-experience of accounting agents in production; token/compute cost of autonomous invoice-processing agents at scale; giving developers visibility into agent behaviour and cost per run. Challenges: scaling from a few automations to many production agents; agent evaluation and guardrails in a finance/compliance context. Must-haves: cost-per-run observability, reliability/guardrails, good agent developer tooling. Nice-to-haves: centralized control plane for agents. ICP confidence: High (explicit "Head of Agents" AI leadership role at a 51-200 emp agent-native company).
Alex Jin — Co-Founder & CTO, Greenlite AI (Bretton AI)
Signal: 4 (ICP technical co-founder at agent-native company shipping agents in production). Source: https://www.businesswire.com/news/home/20250521200064/en/Greenlite-AI-Raises-$15M-Series-A ; https://www.ycombinator.com/companies/greenlite ; https://tracxn.com/d/companies/greenlite/ . Company size: ~78 employees (Tracxn, Jan 2026). Series A / total ~$90.5M raised. Building/shipping AI agents: a "trusted AI workforce" of compliance agents (KYC onboarding, AML, sanctions screening, continuous monitoring) running in production for regulated banks and fintechs (customers incl. Ramp, Mercury, Betterment, Gusto). Alex Jin is co-founder & CTO (prior: PM/iOS eng at Dropbox). Pain points (inferred from product domain, NOT a verbatim quote): agents must be reliable and auditable enough to satisfy bank/fintech regulators; explainability of every automated compliance decision; scaling a multi-agent "workforce" without losing control or accuracy. Challenges: trust/reliability in a heavily regulated setting; audit trails; scaling from a few agents to a fleet. Must-haves: reliability, auditability/explainability, control at scale. Nice-to-haves: cost/observability per agent run. ICP confidence: High (co-founder & CTO; ~78 employees in band; production agent fleet in regulated fintech).
Alex Kroman — Chief Product and Technology Officer, AssemblyAI
Signal: 4 (ICP technical exec at an AI-native company shipping agent infrastructure). Source: https://theorg.com/org/assemblyai/teams/leadership-team ; https://www.assemblyai.com/about ; https://www.linkedin.com/in/alexkroman/. Company size: ~150-250 (AssemblyAI, Series C speech/voice AI infrastructure; powers voice agents built by its customers). Pain points: voice-agent workloads are latency- and cost-sensitive at high volume; inference economics sit directly in the product's gross margin. Challenges: running speech and LLM inference reliably at scale for customers building real-time voice agents. Must-haves: predictable per-request inference cost and reliability at scale. Nice-to-haves: finer-grained usage/cost attribution per customer workload. ICP confidence: Medium — role (CPTO, C-suite technical) and company size/stage fit cleanly and the company is agent-adjacent infrastructure, but no dated public quote from him on agent cost, reliability or observability was found this run. Qualify further before outreach.
Alex LeBrun — Co-founder & CEO (technical co-founder; ex-Facebook AI, founder of Wit.ai), Nabla
Signal: 4 (technical co-founder at a company actively shipping AI agents; found via LinkedIn people search + web verification).
Source: https://www.linkedin.com/in/alexandrelebrun/ ; Nabla Series C coverage.
Company: Nabla — agentic AI for clinical workflows.
Company size: ~143-171 employees.
Stage: Series C ($70M Jun 2025).
Note: CEO by title but a genuine technical co-founder (built Wit.ai, acq. by Facebook; led Facebook AI applied research) — ICP explicitly allows technical co-founders.
Pain points (INFERRED from role + company, NOT verbatim): strategic bet on agentic clinical AI economics; agent reliability and cost at scale as a company-level concern.
Challenges (inferred): scaling agent deployments across many health systems while keeping unit economics viable.
Must-haves (inferred): predictable agent cost, reliability, safety/compliance.
Nice-to-haves (inferred): cost-per-run analytics, context efficiency.
ICP confidence: Medium (valid technical co-founder; CEO title slightly outside pure IC/eng-leader ICP).
Alex Lunev — Head of Engineering, LangChain
Signal: 1/4 (ICP identified directly via LinkedIn people search "LangChain head of engineering"). Source: https://www.linkedin.com/in/alex-lunev-19a9291/
Company size: ~100-200 (Series B; LangChain/LangGraph agent framework + LangSmith observability)
Pain points (inferred from role): operating agent infrastructure at scale for customers; token/context efficiency; production agent reliability; cost visibility per agent run.
Challenges: scaling framework + hosted agent platform reliably; helping customers control agent cost/observability.
Must-haves: production reliability, agent observability, token/context efficiency.
Nice-to-haves: cross-tool cost benchmarking.
ICP confidence: High (Head of Engineering, agent-native company, in range). New company for Brain.
Signal: 4 (ICP technical co-founder/CTO at an agent-native company). Source: https://www.linkedin.com/in/alexmcleodio/ ; discovered via AI-agent company/funding map (fastaijobs.com AI Agent Company Map 2026) + LinkedIn people-search verification, 2026-07-27 run. Company size: ~130 employees (Tracxn, Jun 2026); Series B, $127M raised (Sequoia led $75M Series B). Serval = AI-native ITSM platform; autonomous AI agents resolve IT help-desk requests/workflows in production for enterprises. Pain points (INFERRED from role+company context, not a verbatim quote): reliability of autonomous agents taking real actions on production IT systems; cost/visibility per resolved ticket as ticket volume scales; safe write-access/permissioning for agent tool calls. Challenges: last-mile reliability (pilot 80% -> production 99%+) for autonomous IT actions; auditability of agent decisions. Must-haves: guardrails + reliability on agent actions; visibility/control over cost per agent run. Nice-to-haves: model-agnostic routing; replay/simulation of agent runs. ICP confidence: High (technical co-founder/CTO; Series B; 130 emp; agent-native, shipping in production).
Signal: 4 (ICP technical co-founder shipping ecommerce support agents). Source: https://www.eesel.ai/blog/who-owns-gorgias ; https://gorgias.com . Company size: 400+ employees. Series C-stage (investors: Shopify, Sapphire Ventures, Alven, CRV, SaaStr Fund). Independent, founders still leading. Actively building AI agents: Gorgias ships AI customer-service/support agents for ecommerce brands (AI Agent for automated resolution across chat/email). Note: Gorgias already appears in the brain via Victor Duprez — this is a DISTINCT person (co-founder & CTO), not a duplicate. Pain points (inferred from domain, not verbatim): reliable autonomous resolution across thousands of merchant support conversations; controlling per-resolution cost; observability/quality of agent responses. Challenges: reliability and accuracy at ecommerce scale, cost efficiency per resolution. Must-haves: production reliability, cost control, observability. Nice-to-haves: efficiency, compounding improvement across merchants. ICP confidence: High (technical co-founder & CTO at 400+ person Series C-stage SaaS actively shipping support agents in production; in target size band).
Alex Shevchenko — Head of Applied Research (leads Ramp Labs), Ramp
Signal: 1 — ICP speaking publicly about agent token/compute costs and efficiency. | Source: https://www.youtube.com/watch?v=trEM9OKr5Sc ("How Ramp built an AI agent that can think outside of tokens") and https://inferencebysequoia.substack.com/p/how-ramp-solved-the-fatal-flaw-in ; also https://www.akashbajwa.co/p/building-ramp-sheets-ramp-labs-and | Company size: Ramp ~1,000-1,500 employees (2026; $44B valuation) — within 50-2,000 ICP band. | Pain points: controlling per-run agent token/compute cost at scale; deliberately biases agents toward cheaper deterministic actions (e.g., Excel formulas over Python code-gen) to cut token spend; runs a stream of public agent experiments (Ramp Labs / Ramp Sheets) where cost-per-run and reliability tradeoffs matter. | Challenges: keeping agent runs cost-efficient while scaling a fleet of agents across finance/procurement workflows in production. | Must-haves: token/cost efficiency and control over per-run agent cost; reliability of agent outputs. | Nice-to-haves: visibility into cost/performance tradeoffs across many concurrent agent experiments. | ICP confidence: High — Head-level technical AI leader at a <2,000-employee company actively shipping AI agent fleets in production, publicly focused on agent cost/token efficiency (direct match to Alpha's value prop). NOTE: profile URL left blank — not verified/available (LinkedIn browsing was down this run; not fabricating a URL). Found via web/podcast research fallback, not LinkedIn.
Alexander Christie — Co-Founder & Chief Technology Officer, Attio
Signal: 1 (ICP technical leader whose engineering org is publishing about agent cost attribution and reliability). Source: https://attio.com/engineering/blog/you-cant-just-prompt-your-way-to-great-ai-features (Attio engineering blog on their in-house "Thread Agent" framework); https://www.gv.com/news/attio-ai-native-crm (GV, on Christie launching Ask Attio and putting agentic workloads at the heart of the platform); https://attio.com/blog/attio-raises-52m-series-b; title corroborated by Crunchbase and Forbes profile. Company size: ~140-177 employees (Tracxn 177 as of 30 Apr 2026; Built In 140; PitchBook 115 — call it ~150). Series B, $141M raised total ($52M Series B led by GV, Aug 2025). 5,000 customers; agent surfaces in production include Ask Attio, Research Agent, Call Intelligence, custom agents in Workflows, AI attributes, plus sub-agents. Pain points (from own engineering blog, verbatim-adjacent): needed "visibility into usage and cost" — every LLM and tool call tracked and attributed to workspace, feature and request; scale of ~600k LLM completions, ~40k tool runs and >1B tokens sent to model providers in a single week; noisy workspaces/features dominating shared inference capacity; agents triggering sub-agents making failures hard to attribute. Challenges: model/provider lock-in and graceful fallback; cost-vs-quality A/B testing across model variants; regression detection when prompts or models change; balancing "immediate" vs deferrable agent runs. Must-haves: per-workspace/per-feature cost attribution, full agent-run tracing incl. sub-agents, fairness/rate-limiting across tenants, cost+latency reporting in the test harness. Nice-to-haves: automatic retries/fallback models on schema violations, LLM-assisted repair of malformed output. ICP confidence: High (co-founder & CTO at a ~150-person Series B AI-native company with many agents in production; documented, first-party pain on cost attribution and agent observability).
Signal: 4 (senior technical leader at a company shipping AI agents). Source: https://theorg.com/org/elevenlabs/org-chart/alexander-holt | company: https://elevenlabs.io/blog/series-d. Company size: ~1,122 employees (verified — Jul 2026); within 50–2,000 headcount band. Ex-Palantir (Forward Deployed Eng); VP Eng at ElevenLabs since Sep 2023. ElevenLabs ships a conversational AI Agents Platform (ElevenAgents) — 250,000+ agents built by developers; enterprise voice/chat agents with reliability, testing, monitoring. Pain points (inferred from VP Eng at an agents-platform company): production reliability of voice/chat agents, cost of custom inference infra at scale, observability into per-agent cost and quality. Challenges: scaling agent inference economically; reliability/latency for real-time voice agents. Must-haves: cost visibility, production reliability, monitoring/eval. Nice-to-haves: model routing, context optimization. ICP confidence: Medium (title = VP; company ships agents and fits 50–2,000 headcount; CAVEAT — ElevenLabs is now Series D ($500M round Feb 2026, $11B valuation), later-stage than the stated "Series A–C" band, so it stretches the funding-stage criterion; pain inferred from role + company). METHOD NOTE: Chrome/LinkedIn NOT connected this run; verified via web research (org chart + company blog). LinkedIn URL not directly verified.
Alexander Luksidadi — CTO & Co-Founder, Rose Rocket
Signal: 4 (ICP shipping agents). Source: https://www.roserocket.com/solutions/agents ; https://www.businesswire.com/news/home/20250210026704/en/Rose-Rocket-Launches-TMS.ai-Ushering-in-the-AI-Native-Era-of-Transportation-Management ; LinkedIn people search run live this session. Company size: ~69-153 depending on source (2026), Toronto Canada, YC S16, Series B $38M (2023). LIVE LINKEDIN VERIFIED this run — "Alexander Luksidadi — CTO / Co-Founder at Rose Rocket (YC S16). We're hiring!". IMPORTANT CORRECTION: web research initially surfaced Ian Logan as VP Engineering at Rose Rocket; live LinkedIn search does NOT show him at the company and instead surfaces Luksidadi as CTO — Ian Logan was DROPPED as a likely departure. Agent evidence: Rose Rocket launched "TMS.ai", an AI-native TMS with a dedicated AI Agents product line automating quoting, tracking, dispatch and data entry for freight brokers and carriers. Pain points: reliability of agents doing tendering/tracking/invoicing autonomously for freight customers; exception-handling visibility as agents take over routine dispatch. Challenges: multi-tenant agent cost tracking across broker customers; cross-border US/Canada ops. Must-haves: per-run cost visibility across customer-facing agent deployments, production reliability monitoring. Nice-to-haves: orchestration tooling for scaling from a few agent types to many (quoting, tracking, exceptions). ICP confidence: High — co-founder CTO confirmed live, funding stage and agent product line both verified. Secondary target at same company: Christopher M., Director of Engineering at Rose Rocket (seen live on LinkedIn, not added this run).
Alexander Matthey — Chief Technology Officer, Parloa
Signal: 4 (CTO leading eng at a qualifying company deploying agents at scale)
Source: https://www.parloa.com/parloa-in-the-press/parloa-adds-alexander-matthey-as-chief-technology-officer/ | https://www.businesswire.com/news/home/20250506194825/en/Parloa-Raises-$120M-Series-C-to-Reinvent-Customer-Service-with-Agentic-AI
Company size: ~300 employees (Series C $120M, $1B valuation; offices Berlin/Munich/NY)
Pain points (inferred): Parloa's AMP lets enterprises "build, test, and deploy millions of AI agents" for contact centers — production reliability at very high agent volume; Matthey (ex-Adyen CTO, scaled eng 250->1500) brought in to bring complex AI apps into production for the world's largest enterprises.
Challenges: reliability & safety of voice agents at enterprise scale; testing/QA across large agent fleets; cost at millions-of-agents scale.
Must-haves: agent management + observability; reliability at scale; robust agent testing tooling.
Nice-to-haves: cost control across the agent fleet.
ICP confidence: High (CTO, ~300 emp, agentic AI platform in production).
Note: LinkedIn slug inferred from his Parloa post activity (linkedin.com/in/alexander-matthey) — verify before outreach; common name.
Alexander Schwarm — SVP of AI & Agentic Factory, Netomi
Signal: 4 (web-research pivot — Chrome/LinkedIn NOT connected this run). Source: https://www.zoominfo.com/c/netomi-inc/472266025 ; https://getlatka.com/companies/netomi . Company size: ~208-220 employees (67 in engineering; Latka/Growjo 2026); Series C; #1 agentic AI platform for CX. Role: SVP of AI & Agentic Factory (joined Jun 1, 2026) — senior technical AI leader (VP-level, above Director threshold). Actively shipping AI agents: enterprise agentic customer-service agents at scale. Pain points (inferred from 'Agentic Factory' mandate): standardizing/scaling many production CX agents ('factory' = 5+ agents); per-run cost and reliability visibility across an agent fleet; moving agents from demo to reliable production. Challenges: operationalizing an agent factory with consistent reliability and cost control. Must-haves: cost-per-run observability, reliability at fleet scale. Nice-to-haves: unified control plane across agents. ICP confidence: High (VP-level AI leader at a ~210-employee company running an agent factory in production).
Signal: 4 (adapted method — found via LinkedIn company People-page directory of an in-band agent-native company; LinkedIn content/post search was low-yield this run, consistent with prior runs). Source: https://www.linkedin.com/company/perplexity-ai/people/ and https://www.linkedin.com/in/alexandr-yarats/. Distinct from co-founder Denis Yarats (already in brain). Company size: 201-500 employees (LinkedIn header). In-band, agent-native (ships answer engine, Comet browser agent, Deep Research agent in production). Owns the core agentic search product. Pain points (INFERRED from role+company, no quotes): per-query/per-agent-run LLM cost at consumer scale (agentic answer engine + Comet agentic browser + Deep Research agent, millions of runs); token/context efficiency; reliability & citation accuracy of agent outputs; latency; near-zero per-run/per-agent cost visibility. Challenges: keeping unit economics sane as agent usage compounds; orchestrating multi-step search->browse->research agents reliably at scale. Must-haves: per-run/per-agent cost observability; reliability guardrails at massive scale. Nice-to-haves: model routing/cost optimization; agent governance. ICP confidence: High (Head of Search/AI, leads the flagship agent product).
Signal: 4 (ICP data/AI leader at agent-shipping company; found via LinkedIn currentCompany people-search of Zapier). Source: linkedin.com/in/alexeivink (Zapier company id 2418251 people search, past-month). Company size: ~1,000-1,800 (Zapier; Zapier Agents / Central AI agents at scale). Pain points (INFERRED from role+company stage, not an individual quoted post): visibility into and attribution of AI/agent spend across the data platform; measuring cost-per-outcome of agentic automations; data/eval infrastructure to keep agents reliable. Challenges: instrumenting per-agent cost & quality signals across a very high-volume platform; justifying frontier-model spend. Must-haves: per-run/per-agent cost visibility; reliability & eval telemetry. Nice-to-haves: automatic routing/cost optimization; shadow-savings metrics. ICP confidence: High (VP Data, senior technical leader, in-range agent-shipping company).
Allan Leinwand — Chief Technology Officer, Webflow
Signal: 4 (ICP CTO at SaaS company pivoting to and shipping AI agents). Source: https://diginomica.com/ai-reshaping-developer-workflows-webflows-cto-explains-how ; https://tracxn.com/d/companies/webflow/ ; https://layoffhedge.com/company/webflow. Company size: ~1,710 employees (May 2026); ~$336M raised, ~$4B val (Series ~C/D). Actively shipping agents: Webflow's 2026 'agentic web' pivot — App Gen (AI app building), AEO agents (launched Team plan May 2026), and agentic workflows across CMS/SEO. Allan Leinwand is CTO (ex-CTO Slack, ServiceNow, Sumo Logic). Pain points (inferred from product + domain, not verbatim quotes): reliability of agents doing production work across customer sites; cost/observability of agent runs at platform scale; controlling agent behavior on customer-facing web output. Challenges: shipping reliable, cost-efficient agentic features across a large customer base during an org-wide agentic pivot. Must-haves: reliability, per-run cost visibility, control/guardrails. Nice-to-haves: observability + compounding quality. ICP confidence: Medium-High (named CTO, 1,710 emp in band, agents shipping in production across an explicit agentic strategy).
Allen Calderwood — Co-Founder & Chief Technology Officer, Hadrius
Signal: Bucket 4 — ICP building/shipping agents at a qualifying company. Found via web research (funding news); LinkedIn/Chrome unavailable this run.
Source: https://fintech.global/2026/07/16/hadrius-secures-27m-to-scale-agentic-compliance/
Company size: 80 employees (Y Combinator company page: https://www.ycombinator.com/companies/hadrius). $27M across seed + Series A led by CRV, ~2026-07-14.
Agent evidence: agentic compliance infrastructure for financial services; 500+ financial firms run compliance programs on the platform. Reported 95% reduction in false positives and 70% reduction in manual compliance tasks. Expanding agentic surface (marketing approvals, trade surveillance, WORM archiving) before end of 2026.
Pain points: CEO Thomas Stewart, quoted in source — "If AI is generating the communications, the marketing, and the trades, only AI can review them at the same scale." Implies pressure to scale agent oversight without adding headcount and to keep agent actions auditable at speed. Inferred for the CTO: reliability and auditability of agents reviewing regulated financial communications.
Challenges: scaling an agentic system in a near-zero-error-tolerance regulated environment while rapidly widening the agent surface area.
Must-haves: auditable, controllable agent execution with full trails; cost-efficient scaling of high-volume agent workloads.
Nice-to-haves: multi-model routing / cost optimization as agent volume grows (inferred).
ICP confidence: Medium-High — clean CTO/co-founder title, headcount 80, Series A, agents in high-stakes production. Quote is CEO-attributed, not Calderwood's own.
Signal: 3 (VP of AI Engineering at AI-observability vendor Coralogix; competitor-adjacent space — surfaced via agent cost/observability searches).
Source: https://coralogix.com/blog/coralogix-strengthens-ai-leadership-with-appointments-of-liran-hason-and-alon-gubkin/ ; https://www.globenewswire.com/news-release/2025/02/27/3033786/0/en/Coralogix-Strengthens-AI-Leadership-with-Appointments-of-Liran-Hason-to-VP-of-AI-and-Alon-Gubkin-to-VP-of-AI-Engineering.html
Company size: ~597–803 employees. Fits 50–2,000.
Background: Co-founder & former CTO of Aporia (AI observability), acquired by Coralogix; now leads AI engineering — building the systems that detect hallucinations, data leakage, toxicity and monitor AI/agent behavior in production.
Pain points (inferred from role/company focus): engineering reliable, low-latency observability over high-volume agent runs; attributing cost/quality per run.
Challenges: scaling agent monitoring infra across customer workloads.
Must-haves: production agent tracing, cost/token attribution, evals.
Nice-to-haves: automated regression detection across agent versions.
ICP confidence: High (technical AI-engineering decision-maker at ~600-person AI-native company building agent tooling). Same partial-buying-intent caveat as vendor peer.
Alon Tron — Co-Founder & Chief Technology Officer, Noma Security
Signal: 3 (engaging with / building in the agent-control and agent-runtime space adjacent to competitor tooling). Source: https://noma.security/blog/securing-the-agentic-frontier-noma-unveils-the-first-real-time-agent-runtime-security-for-cursor/ ; https://www.timesofisrael.com/tel-aviv-startup-raises-100-million-to-defend-against-ai-agent-vulnerabilities/. Company size: ~130 (Apr 2026 aggregator data cited in funding coverage), Tel Aviv Israel, Series B $100M (2026). LIVE LINKEDIN VERIFIED this run — people search returns "Alon Tron — Co-Founder & CTO at Noma Security", Israel. Profile URL not captured in the search result; left blank rather than guessed. Agent evidence: Noma raised $100M Series B specifically to secure AI agent vulnerabilities in production; shipped real-time agent runtime security for Cursor. DIRECT QUOTE from Tron: "offering runtime security access is no longer optional; it is the new standard for agentic AI solutions". Pain points: enterprises deploying agentic AI lack runtime visibility and control — this is Noma's stated market thesis. Challenges: KNOWN GAP — Noma builds FOR agent security; it is not confirmed to run 5+ agents internally in production. Evidence is vendor-thesis-level, not confirmed internal production-agent usage. Must-haves: unknown for Noma's own internal stack. Nice-to-haves: unknown. ICP confidence: Medium — title, company, funding and headcount all verified (title confirmed live on LinkedIn), but the "actively shipping agents in production" company criterion is met by product category rather than by confirmed internal agent fleet. Approach as a peer/ecosystem conversation rather than a pure buyer.
Aman Magoon — Co-founder & Chief Product Officer, Adonis
Signal: 4 (AI-focused product co-founder at an ICP agent-shipping company). Source: https://www.prnewswire.com/news-releases/adonis-raises-40m-series-c-to-equip-healthcare-providers-with-aidriven-revenue-cycle-operations-302722199.html ; https://www.adonis.io/ | Company size: ~80+ employees; Series C ($40M Mar 2026), $95M+ total; healthcare RCM AI-agent orchestration platform. | Pain points: (inferred) building agentic product that autonomously resolves claims reliably in a regulated domain; proving pilot→production outcomes to health-system buyers; visibility into what each agent run costs and does. | Challenges: reliability/quality of autonomous agents across diverse payer rules; measurable ROI per run. | Must-haves: production reliability, observability, per-run cost/outcome tracking. | Nice-to-haves: agent eval + iteration tooling. | ICP confidence: Medium — matches "VP/Head of Product (AI-focused)" persona (CPO at agentic-AI company); product title, hence Medium.
Amanpreet Singh — Co-Founder & CTO, Contextual AI
Signal: 4 — technical co-founder/CTO at a qualifying company building agents. Amanpreet Singh (ex-FAIR/Hugging Face) is Co-Founder & CTO of Contextual AI, an enterprise AI agent / grounded-RAG platform.
Source: https://en.wikipedia.org/wiki/Contextual_AI ; https://www.cbinsights.com/company/contextual-ai
Company size: ~93 employees (confirmed 2026); Series A (~$80M raised).
Pain points: grounding/reliability & factual accuracy of enterprise agents in production; long-context cost and token waste; RAG-driven context inflation increasing cost per run.
Challenges: making agents reliably grounded on enterprise data at acceptable cost.
Must-haves: reliable, grounded agent outputs; cost-efficient long-context handling; production observability.
Nice-to-haves: provider flexibility.
ICP confidence: High — Co-Founder & CTO, ~93-employee Series A AI-native company building enterprise agents. (Note: CEO/co-founder Douwe Kiela moved to Google DeepMind in 2026; Amanpreet Singh remains as CTO.)
Amar Kulkarni — Senior Director of Engineering, Asana
Signal: 4 (ICP engineering leader at a company running a large fleet of production agents). Source: LinkedIn people search "\"Director of Engineering\" AI agents LLM platform observability cost" (US geo filter) -> profile https://www.linkedin.com/in/amarnathkulkarni/ ("Senior Director of Engineering @ Asana", SF Bay Area; profile shows open roles). Corroborating: https://asana.com/resources/ai-teammates-overview (21 out-of-the-box AI Teammate agents for Marketing, IT and Ops; AI Studio), https://siliconangle.com/2026/06/04/asana-launches-ai-powered-products-help-organizations-manage-human-agent-work/ , https://www.unifygtm.com/insights-headcount/asana. Company size: ~1,819 employees across 13 offices (Jan 2025 filing) — within the 50-2,000 band but at the top of it; AI ARR ~$6M from 200 beta customers before GA. Pain points (INFERRED from role + product surface, no verbatim quote captured this run): operating 21+ named agents in production and moving them from beta to GA; coordinating "human and agent work" in one system of record; per-agent cost and quality as agent count grows across customer workspaces. Challenges: keeping a large multi-agent fleet reliable and economically sane while monetising it (AI ARR still small relative to headcount). Must-haves: per-agent cost and quality visibility across a fleet, production reliability. Nice-to-haves: shared observability and eval tooling across agents. ICP confidence: Medium (Senior Director of Engineering is at the ICP floor and the company is squarely shipping many agents in production; downgraded from High because Asana is a public company rather than Series A-C, sits at the top of the size band, and no direct pain quote was captured).
Signal: 2/4 (surfaced via analyst coverage — Josh Bersin's Mar 2026 piece on Gloat entering the agentic-HR war — then confirmed as ICP).
Source: https://joshbersin.com/2026/03/gloat-enters-the-crowded-war-for-ai-agents-in-hr/ ; https://finance.yahoo.com/sectors/technology/articles/gloat-introduces-agentic-hr-end-130000549.html ; https://gloat.com/ ; https://en.wikipedia.org/wiki/Gloat_(company)
LinkedIn: personal profile URL NOT verified this run — company page https://www.linkedin.com/company/gloat. FIND HIS PERSONAL URL BEFORE ANY OUTREACH.
Company size: DISPUTED — 126-177 across trackers (Apr-Jun 2026); one source says ~300 global incl. 150 in the Israel R&D centre. In band on every estimate.
Agent evidence: Agentic HR Platform launched 31 Mar 2026, built on Loomra (Workforce Context Engine: workforce knowledge graph + semantic embeddings + skills inference/clustering + career trajectory modelling + enterprise-scale matching). Multiple named agents incl. a Workforce Insights / Headcount Planning agent. Also shipped into Microsoft 365 Copilot and Teams.
Pain points (INFERRED, plus one strong public signal): their own launch premise is verbatim "agents are only as intelligent as the context they carry" — i.e. they have already concluded the bottleneck is the context layer, not the model. That is a context-engineering/token-waste problem stated in their own marketing.
Challenges: serving agents through a third-party surface (M365 Copilot/Teams) means far less control over invocation volume and therefore over cost; a knowledge-graph-grounded context engine is expensive per call.
Must-haves: context/token efficiency and visibility into what each agent actually loaded per run.
Nice-to-haves: per-surface (Copilot vs native) cost breakdown.
ICP confidence: MEDIUM. Role and agent-activity are strong and the context-cost thesis maps directly onto our pitch. Downgraded because (a) headcount sources conflict 126-300, (b) CTO tenure not re-verified against a 2026 source — Wikipedia/founder records only, (c) no personal LinkedIn URL confirmed.
Signal: Bucket 4 — ICP building/shipping agents at a qualifying company. Found via web research; LinkedIn/Chrome unavailable this run.
Source: https://techcrunch.com/2026/06/15/ai-agents-are-becoming-employees-newcore-emerges-with-66m-to-give-them-identities/
Company size: "more than 50 employees across the U.S. and Israel" (TechCrunch, 2026-06-15). $66M seed at $300M valuation (Cyberstarts, Index Ventures, Evolution Equity). NOTE: seed-stage, not Series A–C — flagged as a deliberate exception given headcount, funding scale and exact thematic fit.
Agent evidence: identity/authorization platform treating AI agents as first-class identities with their own permissions, lifecycle controls and revocation; ships an "Agentic Skill" product plus a mobile app for granting/revoking agent access. Fewer than 10 paying customers, 10+ design partners as of June 2026.
Pain points: CEO Zohar Alon, quoted in source — "We know for sure that the scale and the complexity that those things [AI agents] are going to add to 15- or 20-year-old identity platforms are going to break them." Also describes incumbent agent support as "on the side — it's not integrated," i.e. fragmented control over agent access.
Challenges: building agent identity/permissioning ground-up rather than bolted-on; early-stage production scaling risk (monetization only starting summer 2026).
Must-haves: fine-grained, revocable control over what agents can access and do; integrated human-oversight layer for agent actions.
Nice-to-haves: per-agent-identity cost and usage visibility as adoption scales (inferred).
ICP confidence: Medium — clean CTO/co-founder title, headcount just over the floor, whole thesis is controlling agents in production (strong overlap with Alpha's control pillar). Downgraded because no LinkedIn URL found, quote is CEO-attributed, and the round is seed rather than A–C.
Profile URL: not found this run (do not fabricate).
Amir Haghighat — Co-founder & CTO, Baseten
Signal: 3 (ICP technical decision-maker at inference-infra vendor in the agent cost/production-reliability space — surfaced via agent-cost/observability searches). Source: https://www.baseten.co/author/amir-haghighat/ ; https://www.baseten.co/blog/building-the-future-of-ai-infrastructure-amir-haghighat/ ; https://research.contrary.com/company/baseten. Company size: ~204-238 employees (Feb-Mar 2026; grew from 49 in 2024). Independent; raised $300M at $5B val Jan 2026; 10x YoY revenue 2025. Fits 50-2,000. Technical co-founder & CTO (ex-Head of Eng Gumroad; PhD CS UC Irvine); primary technical voice. Baseten = inference-first platform for deploying/scaling production AI models & agents; "accelerate MVPs into production applications." Pain points (inferred from role/company focus): cost & latency of production inference for agentic workloads; scaling agents reliably from MVP to production; per-run cost/performance. Challenges: production reliability + cost efficiency of model/agent inference at scale across customers. Must-haves: cost-efficient, reliable production inference; observability of model/agent runs. Nice-to-haves: per-run cost attribution; model routing. ICP confidence: Medium (technical co-founder/CTO at ~220-person independent AI-infra company in Alpha's adjacency; build-vs-buy / partner-adjacent — inference vendor, not a pure buyer). LinkedIn URL not verified this run (left blank rather than fabricated).
Amit Carmi — Co-Founder & CEO, Sett
Signal: 4 (technical co-founder / decision-maker at agent-native company)
Source: https://www.bvp.com/news/meet-the-founders-of-sett | https://www.calcalistech.com/ctechnews/article/rjxbyqisze
Company size: ~50 employees
Pain points (inferred): as technical co-founder (ex-Unit 8200) and CEO of an agent-native UA company, owns unit economics of agent-driven creative and end-to-end agent automation (25x cheaper / 15x faster positioning).
Challenges: scaling the agent org; cost/quality tradeoffs across many game clients; reliability of continuous agent output.
Must-haves: efficient, reliable agent execution at scale.
Nice-to-haves: cost/quality observability per agent workflow.
ICP confidence: Medium (CEO rather than eng lead, but technical co-founder and buying decision-maker for agent infra).
Contact/profile URL not confirmed — left blank to avoid fabrication.
Amit Prakash — Co-Founder & CTO, ThoughtSpot
Signal: 4 (ICP technical leader at a company shipping a fleet of production agents). Source: https://www.thoughtspot.com/product/agents ; https://www.techtarget.com/searchbusinessanalytics/news/366636078/ThoughtSpot-automates-full-platform-with-new-Spotter-agents ; https://www.crunchbase.com/person/a-prakash ; Tracxn May 2026. Company size: ~1,702 employees. Pain points (inferred from public product/positioning): ships a suite of production BI agents (Spotter, SpotterModel, SpotterViz, SpotterCode, Spotter 3) that reason step-by-step, "check their own work," and need human-in-the-loop validation and trust/governance ("Spotter Semantics to bring trust and context"). Reliability/trust and governance of many domain-specific agents is core. Challenges: keeping many governed agents accurate and trustworthy across industries; controlling agent behavior at fleet scale. Must-haves: agent trust/governance, human-in-the-loop validation, reliability. Nice-to-haves: per-agent cost/usage visibility. ICP confidence: Medium-High (co-founder+CTO per Crunchbase/TheOrg, one source flags possible role change so slight staleness risk; 50-2,000 emp; agents in production).
Amit Saraswat — Senior Director of Technology, Kore.ai
Signal: 4 (senior technical leader at a company shipping AI agents at scale). Source: https://theorg.com/org/kore-ai/org-chart/amit-saraswat | company: https://pitchbook.com/profiles/company/170569-54. Company size: ~1,250 employees (verified — 1,252 as of Jun 2026); within 50–2,000 headcount band. Kore.ai runs an enterprise AI Agent Platform (XO Platform, GALE) automating 450M interactions/day for 400+ Global 2000 firms. Pain points (inferred from Sr Director of Technology at an enterprise agent platform): reliability/accuracy of agents in production at massive scale, LLM cost control across huge interaction volume, cost-per-run visibility and token governance. Challenges: scaling agent orchestration economically; multi-tenant reliability and observability. Must-haves: cost governance/visibility, production reliability, eval + monitoring. Nice-to-haves: model routing, context optimization. ICP confidence: Medium (title = Director+, company ships agents and fits 50–2,000 headcount; CAVEAT — Kore.ai is later-stage/more established than the stated "Series A–C" band ($296M raised, founded 2013), so it stretches the funding-stage criterion; pain inferred from role + company). METHOD NOTE: Chrome/LinkedIn NOT connected this run; verified via web research (org chart + PitchBook/company profiles). LinkedIn URL not directly verified.
Amit Sasturkar — Co-Founder & CTO, Terret (formerly BoostUp)
Signal: 4 (technical co-founder/CTO shipping a fleet of AI agents in production; surfaced via LinkedIn people search 'CTO AI agents production' and verified).
Source: https://www.linkedin.com/in/amitsasturkar/ ; Terret "Virtual Revenue Fleet" of AI Revenue Agents.
Company size: 57 employees; ~$43-50M raised (Series B).
Pain points: scaling an interconnected fleet of AI agents across the full revenue lifecycle (pipeline gen, deal execution, renewals, expansion); automating up to 80% of tactical sales activity reliably.
Challenges: multi-agent orchestration and reliability at scale.
Must-haves: reliable multi-agent orchestration, production monitoring/observability.
Nice-to-haves: cost visibility per agent/run.
ICP confidence: High — Co-founder/CTO at 57-emp Series B AI-native company shipping 5+ agents; right title, size and stage.
Amit Srivastava — VP & Head of AI, Judi Health (Capital Rx)
Signal: 4 (ICP AI leader building agentic AI + voice agents; found via LinkedIn People search for ICP titles + healthcare agents). Source: https://www.linkedin.com/search/results/people/?keywords=VP%20AI%20agents%20healthcare%20production ; verification: https://hitconsultant.net/2026/07/15/capital-rx-rebrands-as-judi-health/ and https://www.healthcareittoday.com/2025/10/27/capital-rx-announces-funding-round-of-400m.../. Company size: est. ~700-1,500 (Capital Rx, rebranded Judi Health 2026; raised $400M expansion round; 5M+ contracted PBM lives; Judi Cloud platform licensed to health plans/TPAs). AI-native benefit-administration health-tech. Actively building agents (per role): Amit is VP & Head of AI "building Agentic AI and voice agents for healthcare customer-service workflows" (ex-Microsoft, ex-ServiceNow VP). Pain points (inferred): reliability of healthcare CS voice agents in a regulated environment; agent cost + observability; scaling agentic benefit-admin workflows. Challenges: production-grade reliability and governance for agents on sensitive healthcare data. Must-haves: reliability, compliance/governance, cost visibility per run. Nice-to-haves: observability, optimization. ICP confidence: Medium (VP/Head of AI at well-funded health-tech in the size band building agentic AI + voice agents; company size is an estimate — verify headcount; stage later than A-C). LinkedIn profile URL not captured this run — do not fabricate.
Amit Verma — Head of Engineering (Founding Head of Engineering and AI), Neuron7.ai
Signal: 4 (ICP at a company shipping named agents into complex service operations). NOTE: LinkedIn/Chrome unavailable this run — LinkedIn URL surfaced in search results, not opened/verified.
Source: https://www.neuron7.ai/about-us (leadership: "Amit Verma — Head of Engineering") ; https://profiles.forbes.com/members/tech/profile/Amit-Verma-Head-Engineering-Neuron7-ai/f29f758c-99e6-4ef9-854e-27ceaed33b6d
Company size: ~129 (SignalHire/ZoomInfo company-profile snippets; other sources 100-200). San Jose, CA. Series B — $44M (Oct 2024) led by Smith Point Capital (Keith Block), with Nexus Venture Partners, Battery Ventures, ServiceNow Ventures; $63M+ total.
Agents in production: CONFIRMED. Company timeline is unusually concrete about the agent transition: first enterprise deployments Aug 2021 at "90%+ resolution accuracy"; Nov 2025 launched "Neuro, a next-generation AI agent built for complex service operations, moving beyond search and recommendations toward AI that can reason, guide, and resolve." Ships an "Agent Library" and a "Deploy Autonomous Agents" use case. Salesforce Agentforce partner for service. Vertical: medical devices, industrial equipment, high-tech infrastructure.
Pain points: [NOT evidenced from him directly — no sourced quote found; his title is sourced, his commentary is not] [inferred from product framing] Moving a service AI from retrieval/recommendation to reasoning-and-acting; resolution accuracy and trust in safety-critical domains (medical devices); grounding agents in messy service records — they ship an entire product line around "Improve Service Record Quality / Ensure AI Success," which strongly implies data-quality-driven agent failure is a live, recurring problem for them.
Challenges: Maintaining 90%+ resolution accuracy as agents move from recommend to resolve; agent grounding on low-quality enterprise service data; safety-critical blast radius (a wrong medical-device service instruction is not a UX bug).
Must-haves: Agent reliability/accuracy measurement in production; grounding and data-quality guardrails; visibility into why an agent resolved the way it did.
Nice-to-haves: Per-agent cost tracking across the Agent Library; regression detection as agents are added.
ICP confidence: High on ICP fit — Head of Engineering (Director+), ~129 employees, Series B, agents named and shipped. LOW on personal-voice evidence: no sourced quote from him about agents. Lead with product/architecture, not with a quote.
Signal: Bucket 1 — ICP writing about agent/LLM production reliability and control.
Source: LinkedIn content search "agent cost LLM production" (past month) surfaced his post dated 31 Aug 2026; verified on https://www.linkedin.com/in/ghaziamjad/recent-activity/all/
Company size: 501-1,000 employees (LinkedIn company page, Lentra — digital lending SaaS for banks/NBFCs, founded 2019, Pune; Series B).
Pain points (his own words): "Your LLM gives a great answer. Your software can't use it." — "Without structured output, engineering teams end up building parsers, regexes, repair logic and retries around inherently variable language." He closes with "Who owns the complexity required to make its response deterministic enough for software to act on?"
Challenges: making probabilistic model output safe to act on inside a regulated production lending workflow; enforcing output contracts at the model/application boundary; retry and repair overhead absorbed by engineering teams; evaluating whether a model can honour a schema before it goes into a production workflow.
Must-haves: schema-constrained/deterministic boundary between model and application; reliability guarantees before a model enters a production workflow; clear ownership of the "make it deterministic" work.
Nice-to-haves: model-agnostic way to compare which models actually honour output contracts; less hand-rolled parsing/repair code.
ICP confidence: High — VP Engineering (agentic systems, applied AI, platform engineering) at a 501-1K-employee Series B vertical SaaS actively building agentic architectures; posts consistently from an engineering-leadership seat about agent reliability in production, not vendor marketing.
Signal: 2 (quoted in third-party press about agent cost blowout).
Source (title, fetched): https://theorg.com/org/replit/org-chart/amol-jain — "Head Of Product Engineering at Replit since June 2025."
Source (pain, NOT directly fetched — page returned empty body): https://venturebeat.com/orchestration/ai-coding-agents-are-blowing-through-budgets-replit-kilo-code-and-symbotic-explain-how-theyre-managing-it
Company size: ~404–597 (Revelio Labs 404, Mar 2026; Tracxn 597, Jun 2026).
Pain points: Agent cost blowout after broadening agents beyond engineering — per the VentureBeat article, a user on the support side had "blown through an insane amount of money"; root cause was an automation running on a top-tier reasoning model. His stated priorities: cost visibility that isn't "anti-productive," model routing, and sensible defaults.
Challenges: Governing agent spend across non-engineering internal users who have no intuition for token cost; balancing guardrails against developer productivity.
Must-haves: Per-user / per-automation spend visibility with alerting before the bill lands; sane model defaults so cheap tasks don't run on the most expensive model.
Nice-to-haves: Automatic model routing by task complexity.
ICP confidence: Medium — title, company and size all qualify, and the pain is exactly our wedge, but the quote comes from a search-index snippet rather than a page I rendered. RE-CONFIRM the VentureBeat quote in a browser before using it in outreach.
Possible LinkedIn (seen only as a search-result title, not fetched): linkedin.com/in/jainamol — verify before use.
Amrish Singh — Founder & CEO, Liberate
Signal: 1/4 (technical founder building reasoning/voice agents; expressed regulated-agent reliability pain; found via press primary sources, not LinkedIn — Chrome/LinkedIn unavailable this run). Source: https://techcrunch.com/2025/10/15/liberate-bags-50m-at-300m-valuation-to-bring-ai-deeper-into-insurance-back-offices/ and https://councils.forbes.com/profile/Amrish-Singh-Founder-CEO-Liberate-Innovations-Inc/32bd2df6-ff7b-4944-8fff-f7d8a56de617. Company size: ESTIMATED ~60-150, NOT independently confirmed via a headcount database — inferred from Series B ($50M Oct 2025, $300M valuation, $72M total) and 60+ enterprise customers / top-100 carriers with voice agents in production. Role note: "software engineer by training, previously built large core systems"; founder/CEO — qualifies as technical co-founder. (Liberate's VP Engineering Ryan Eldridge is already in the People Library.) Pain points: reasoning-agent reliability in long, regulated insurance conversations; auditability and human-in-the-loop compliance (HIPAA/PCI/SOC2); scaling voice AI across sales/servicing/claims. Challenges: hallucination/accuracy control in compliance-heavy voice; RL tuned for long conversations. Must-haves: auditable, compliant, reliable agents with HITL safeguards. Nice-to-haves: cost efficiency of voice-agent workloads. ICP confidence: Medium (technical founder/CEO at agent-native insurance company; company size estimated, not confirmed).
Amrit Santhirasenan — Co-Founder & CEO (technical), hyperexponential
Signal: 4 (ICP technical co-founder at a Series B company shipping an agentic product). Source: https://finance.yahoo.com/technology/ai/articles/hyperexponential-introduces-hyperoperator-unified-agent-123000213.html ; https://www.reinsurancene.ws/hyperexponential-launches-hyperoperator-to-automate-commercial-pc-underwriting-workflows/ ; https://tracxn.com/d/companies/hyperexponential/ . Company size: Series B London insurtech, founded 2017, $91M raised (incl. $73M Series B led by a16z + Battery); platform trusted for $75bn+ in annual premium (Series B scale implies ~150-250 emp; exact headcount unconfirmed). Stage: Series B. Independent. Actively shipping agents: July 2026 launched "hyperoperator," described as the first end-to-end agentic AI underwriting workbench for commercial P&C — advances broker submissions from intake to decision-ready/priced file inside a carrier's own pricing, appetite & authority controls (live demo: broker email → triaged & priced cyber risk in <3 min). Role: Amrit Santhirasenan, co-founder & CEO; actuary/engineer who built the pricing platform (technical co-founder). Co-founder Michael Johnson. Pain points (INFERRED from positioning, NOT verbatim): making an underwriting agent reliable and trustworthy within a carrier's own authority/appetite/compliance controls; keeping humans on high-value judgment while the agent does prep work correctly; cost/latency at scale across $75bn+ premium. Challenges: agent reliability + governance inside regulated underwriting authority limits. Must-haves: control/guardrails, reliability, auditability. Nice-to-haves: per-run cost visibility, speed. ICP confidence: Medium (technical co-founder/CEO; Series B; agentic product just shipped; note: role is technical co-founder-CEO rather than titled CTO/VP Eng, and exact headcount unconfirmed).
Ana Maria Jaime Rivera — Head of AI & Data Science, Snoonu
Signal: 3 (ICP engaging with competitor/adjacent observability content — named in a Datadog LLM Observability customer case study). LinkedIn profile verified live 2026-09-02: "Head of AI & DS @ Snoonu | Master of Data Science & Innovation @ UTS | Building World-Class AI Teams & Scalable AI Systems | Best Emerging Tech in AI 2024", based in Qatar.
Source: https://www.datadoghq.com/case-studies/snoonu/
Company size: ~900 employees (per the Datadog case study). Snoonu = Qatar-based delivery super-app running production AI agents including the "Genie" shopping assistant and Smart Catalog agents. Series B (2023) / Series C (Jul 2025) per Dealroom.
Pain points: No real-time visibility into what agents were actually doing in production. Her quote: "Datadog lets us see exactly how our AI products behave in production so we can fix issues faster." A colleague in the same study: "When a problem in production was caused by a user, it was very difficult to know what the agent was doing."
Challenges: Reproducing failing agent sessions by hand; unexplained latency spikes; unifying classical ML metrics with agent/tool-call telemetry in one place; debugging agents that touch live commerce flows where failures are customer-facing.
Must-haves: End-to-end agent tracing with tool-call visibility, fast root-cause on production agent failures, session replay/reconstruction.
Nice-to-haves: Custom metrics unifying ML models and agents; proactive anomaly alerting on agent behaviour rather than post-hoc debugging.
ICP confidence: High — Head-of-AI title, headcount squarely mid-band, agents demonstrably in production, and a dated first-party quote naming exactly the visibility gap thealpha.ai sells against. Note she is already invested in a competing observability vendor (Datadog), so the wedge is cost/per-run economics rather than tracing alone.
Signal: 4 (ICP leader via broad "Head of AI ... agents production" people-search + web verification). Source: https://www.linkedin.com/search/results/people/?keywords=%22Head%20of%20AI%22%20agents%20production (profile https://www.linkedin.com/in/anand-gupta-1202/). Company size: ~170 (confirmed Revelio/Tracxn) — Wysa, AI mental-health platform, ~$30.5M raised. Headline: "Deploying LLMs & multilingual agents in production." Pain points [INFERRED from verified role+company, NOT verbatim]: cost + reliability of multilingual LLM agents in production; token/context waste across languages; safety/reliability in a regulated healthcare context. Challenges: scaling agents across languages reliably and affordably. Must-haves: production reliability + cost control. Nice-to-haves: observability/eval. ICP confidence: High (Head of AI, agent-deploying ~170-emp company).
Signal: 4 (ICP at company building/shipping enterprise agents; Leena AI ships "Agentic AI Colleagues" for IT/HR/Finance/Procurement across ~500 enterprise customers). Source: https://www.linkedin.com/in/anand-prajapati/ , https://leena.ai/about-us , and CTO interview https://logit.io/blog/post/interview-with-co-founder-and-cto-anand-prajapati/ . Company size: ~268 employees (YC/Bessemer/Greycroft; ~$40M raised). Pain points (inferred from public product positioning, not a verbatim quote): building and customizing many enterprise agents fast (AI Colleague Studio) while automating complex cross-system workflows (Workday, ServiceNow, SAP); achieving reliable ROI at scale across 500 customers. Challenges: scaling and maintaining agents across many enterprise systems; reliability and behavior control in production. Must-haves: production reliability, fast agent build/customization, enterprise integrations. Nice-to-haves: per-agent cost/observability visibility. ICP confidence: High (technical co-founder/CTO at an AI-native agent company in the target size band, shipping agents in production).
Signal: 4 (ICP technical leader building/shipping agents) + 1 (cost/reliability in production). Source: https://theorg.com/org/gnani-ai/org-chart/ananth-nagaraj ; https://www.deccanherald.com/india/karnataka/bengaluru/bengaluru-based-voice-first-agentic-ai-startup-gnaniai-raises-10-million-3950735 . Company size: ~199-244 employees (LeadIQ ~244 Jun-2026; PitchBook 199). Series B ($10M, Mar-2026, Aavishkaar/Info Edge). Voice-first agentic AI, 200+ enterprise customers, 30M+ voice interactions/day across 12+ languages. Pain points (INFERRED): cost & reliability of agentic voice at massive interaction volume; observability/control across many enterprise agents; multilingual reliability. Challenges: keeping large agent fleet reliable and cost-efficient while expanding globally. Must-haves: per-run cost visibility & control; production reliability. Nice-to-haves: compounding improvement; routing. ICP confidence: MEDIUM-HIGH (Co-founder/CTO at ~200-emp Series B agentic-voice company shipping agents in production).
Andrea Michi — Co-founder & Chief Technology Officer, depthfirst
Signal: 1 (ICP publicly on the record about agent cost economics). Found via web research — Claude-in-Chrome/LinkedIn NOT connected this run.
Source: https://www.businesswire.com/news/home/20260331071577/en/Applied-AI-Lab-depthfirst-Announces-$80-Million-in-Series-B-Funding
Company size: 50 employees (Tracxn, June 2026) — exactly at the ICP floor. Series B — $80M led by Meritech Capital, announced 31 March 2026, less than 90 days after a $40M Series A; $120M total. Applied AI lab; agents that reason across whole codebases and infrastructure. Named customers in the release: ClickUp, Lovable, Supabase, incident.io, Moveworks, plus Fortune 500s.
LinkedIn URL: https://www.linkedin.com/in/andreamichi/ (headline seen in search result: "CTO depthfirst | ex-DeepMind")
Pain points: Frontier-model inference cost for agentic security workloads was bad enough that they trained their own model. Direct quote: "When you own the training process, you can optimize for what actually matters in your domain… The result is a model that can be cheaper to run, better at the task, more responsive to continued investment than a general-purpose system." The release states their in-house model dfs-mini1 "outperformed frontier models while running at 10x to 30x lower cost."
Challenges: Keeping unit economics workable while running codebase-wide reasoning agents for large customers; expanding into more security domains means more agents; noise/precision tradeoffs at scale.
Must-haves: Cost-per-task visibility; routing/model-selection control; domain-specific evaluation harnesses.
Nice-to-haves: Cross-domain reuse of eval + cost tooling as they add model families.
ICP confidence: Medium — High on pain signal and stage (Series B, AI-native, explicitly cost-motivated to the point of training their own model), but 50 employees is exactly at the lower ICP bound and he is a co-founder rather than a non-founder buyer. Watch this one: if headcount grows past ~75 it becomes a High.
Andreas Hauri — Co-Founder & CTO, Unique (Unique AG)
Signal: 4 (technical co-founder at net-new agent-native company). Source: https://www.linkedin.com/in/andreashauri/ ; verified via TechCrunch/startupticker/unique.ai. LinkedIn headline: "CTO & Founder @ UniqueAI | Building the Future of Finance with Agentic AI." Company: Unique AG (Zurich), agentic AI workforce for financial services (asset/wealth mgmt, banking); clients incl. Pictet, UBP, LGT, SIX. Funding: Series A $30M (Feb 2025), ~$53M total since 2021. Company size: EST ~100 employees (NOT exactly verified — estimate from funding/stage/clients). Pain points (INFERRED, NOT verbatim): reliable agentic AI for regulated finance; governance + auditability; cost/observability of agent workflows. ICP confidence: MEDIUM-HIGH — technical co-founder/CTO, agent-native, Series A. Caveats: headcount is an estimate (flag to verify); pain inferred.
Andreas Kollegger — Director of Applied AI Research, Neo4j
Signal: 1 (ICP speaking about agent explainability/decision traces) + 4 (ICP at a company shipping agent infrastructure).
Source: AI Engineer World's Fair 2026 speaker record — https://ai.engineer/worldsfair/2026/speakers.json ; session "Context Graphs for Explainable, Decision-Aware AI Agents".
Company size: ~1,028 employees (May 2026); other sources 767–1,000. Within the 50–2,000 band. Raised $100M in Apr 2026 explicitly to build out AI infrastructure and agentic systems. Ships an agent platform with hosted runtimes, built-in observability, governance controls and marketplace integrations.
Pain points (from his own session abstract, paraphrased): "AI agents can follow prompts and use tools, but often lack the institutional context needed to explain WHY a decision is made"; the reasoning — policies, precedents, past outcomes — is "scattered across systems and human memory".
Challenges: no durable record of agent decision traces, causality or context over time; agents are unexplainable after the fact.
Must-haves: decision-trace capture for agents; explainability/auditability of agent actions in production.
Nice-to-haves: causal modelling of agent decisions over time; institutional-context reuse across agents.
ICP confidence: Medium — Director-level applied AI leader (meets the Director floor), headcount comfortably inside 50–2,000, company demonstrably shipping agent runtime + observability. Downgraded from High because Neo4j is a late-stage company well past the Series A–C band, and because his stated pain is explainability/context rather than cost — adjacent to, but not identical with, the core Alpha wedge.
Andreas Stolcke — VP of AI / Distinguished AI Scientist, Uniphore
Signal: 1 (ICP AI leader at an agent-shipping company; found via web research — LinkedIn/Chrome unavailable this run). Source: https://www.youtube.com/watch?v=j-L21dkeV7o ; https://theorg.com/org/uniphore/teams/artificial-intelligence-division. Company size: ~700–1,100 employees (within 50–2,000 band); Uniphore ships agentic/conversational AI agents in production for enterprise CX. Andreas Stolcke is VP of AI / Distinguished AI Scientist (veteran speech & language scientist). Pain points (INFERRED from role/company context — NOT verbatim): productionizing and scaling AI agents reliably, model/token cost management, quality vs. cost trade-offs across agent workflows. Challenges: balancing agent accuracy/reliability against inference cost at enterprise scale. Must-haves: cost + reliability visibility per agent. Nice-to-haves: routing/optimization. ICP confidence: Medium (clean "VP of AI" ICP title; caveat: research/science-oriented role, and Uniphore is late-stage rather than Series A–C). NOTE: pain points inferred, not quoted.
Andrei Negrau — CEO & Co-founder, Siena AI
Signal: Targeted ICP people-search — net-new agent-native company (Siena AI: autonomous CX/customer-service agents for commerce) not in brain. Source: https://www.linkedin.com/in/negrau/ (LinkedIn people search). Company size: ~50-90 (Series A). Pain points (ICP-inferred; no direct post captured this run): reliability of customer-facing autonomous agents; cost per resolution as ticket volume scales; escalation/quality control. Challenges: scaling autonomous CX agents across many brands while controlling LLM spend and maintaining accuracy. Must-haves: reliability, cost-per-run visibility, guardrails. Nice-to-haves: model routing, eval tooling. ICP confidence: Medium (C-suite founder of agent-native company; note: CEO/business co-founder — technical counterpart is Siena's eng leadership).
Andrew Bihl — Co-Founder & CTO, Numeric (numeric.io)
Signal: 4 (ICP technical decision-maker building/shipping agents in production). Source: Numeric $51M Series B announcement (https://www.numeric.io/blog/numeric-raises-51m-series-b); YouTube "Numeric CTO Andrew Bihl: Building AI-Powered Accounting Automation"; Numeric Engineering Substack. Company size: 51–100 (confirmed range) — ideal ICP. Numeric builds AI accounting agents for close automation and broader finance workflows, incl. an MCP connector to build custom agents and automate multi-step workflows. Their stated architecture routes work between AI (interpret unstructured data), deterministic code (precise calculations), and human oversight (exceptions) — reflecting acute reliability/accuracy pressure. Ex-Segment/Twilio engineer. Pain points (partly stated + inferred): reliability/accuracy of agents in high-stakes financial workflows; controlling cost of multi-step agent runs; auditability of agent actions; scaling from close-mgmt to many finance agents. Must-haves: precision, auditability, cost control. Nice-to-haves: agents that generalize across accounting workflows. ICP confidence: High (CTO + technical co-founder, Series B SaaS, right size, shipping agents in production).
Andrew Filev — Founder & CEO (technical; ex-Wrike founder), Zencoder (For Good AI Inc)
Signal: 4 (ICP building/shipping agents). Source: https://www.linkedin.com/posts/filev_zen-agents-zencoder-the-ai-coding-agent-activity-7328086855474655232-iN9Y ; https://venturebeat.com/ai/zencoder-launches-zen-agents-ushering-in-a-new-era-of-team-based-ai-for-software-development ; https://zencoder.ai/about
Company size: ~50+ engineers (May 2026), founded 2023, independent/self-funded (Filev's exit from Wrike, sold ~$2.25B). Acquired Machinet (2025).
Company context: Zencoder ships autonomous coding agents plus "Zen Agents" (org-wide custom agents + open-source marketplace) and "Zenflow" orchestration layer — team-based AI agents for software development.
Pain points (inferred from product): orchestrating multiple coding agents reliably across a team; making autonomous coding agents production-grade; managing cost/token consumption of agentic coding (1000x more tokens than standard reasoning).
Challenges: scaling from single agent to org-wide fleets of agents; reliability of multi-step agent workflows.
Must-haves: reliable orchestration and control of many agents; visibility/observability into agent runs.
Nice-to-haves: cost controls per agent run; agents that improve over time.
ICP confidence: Medium — technical founder/CEO and clear decision-maker, actively building agent fleets; headcount (~50) is right at the ICP floor, so sizing is a mild flag.
Andrew Gabbeitt — Director of Implementation Engineering, Abridge
Signal: Signal 4 (LinkedIn people search at named agent company; Director-level. Displayed on LinkedIn as "Andrew G."; full surname from profile slug. Sourced by role/company match)
Source: https://www.linkedin.com/search/results/people/?keywords=%22Abridge%22%20director%20engineering
Profile: https://www.linkedin.com/in/andrewgabbeitt/
Company size: ~300-500 (Abridge; clinical AI documentation / clinical AI agents)
Pain points: (inferred, not directly observed) Deploying clinical AI agents into health systems reliably; cost and reliability at production scale
Challenges: (inferred) Making agent behavior predictable and observable across deployments
Must-haves: (inferred) Reliability + cost visibility per deployment/run
Nice-to-haves: (inferred) Governance across customer environments
ICP confidence: Medium (Director-level but Implementation Engineering skews to deployment vs core agent build; company is in-ICP)
Andrew Hsu — Co-Founder & CTO, Speak
Signal: Bucket 4 (ICP technical co-founder running conversational AI agents at consumer scale).
Source: https://theorg.com/org/usespeak/org-chart/andrew-hsu ; https://www.crunchbase.com/organization/speak-3 ; https://asugsvsummit.com/speakers/andrew-hsu ; https://www.youtube.com/watch?v=zZ3yHiaLBKg ("How AI Tutor Works with Speak's Andrew Hsu")
Company size: ~80 employees (reported as "not even 80" mid-2026) — at the lower end of the ICP band but above the 50 floor. >$150M raised from Accel, OpenAI Startup Fund, Khosla, Founders Fund; valued >$1B after a $78M round (Series C stage). Founded 2016 SF with Connor Zwick. #1 education brand in South Korea; subscribers in 40+ countries.
Agent activity (verified): Speak runs an AI language tutor built on speech recognition + LLMs — a realtime conversational agent handling every learner turn, across millions of consumer sessions in 40+ countries.
Pain points (inferred from workload shape, not a quote): consumer subscription pricing against per-turn inference cost is the definitive agent cost-blowout shape — gross margin is a direct function of tokens/ASR seconds per session. Scaling to a new market multiplies inference load without multiplying price.
Challenges: holding realtime latency for spoken conversation while cutting per-session cost; pedagogical quality/reliability of agent responses at scale; multi-lingual model routing.
Must-haves: cost-per-session visibility and per-market attribution; latency-preserving cost reduction; regression detection on tutor quality across model swaps.
Nice-to-haves: automatic model tiering (cheap model for drills, expensive for open conversation); context/token trimming across long learner histories.
ICP confidence: Medium-High — CTO + technical co-founder at Series C scale with a consumer agent product; downgraded from High because headcount (~80) sits near the ICP floor and the agent stack is a single product surface rather than a fleet of 5+ distinct production agents. Publicly speaks (ASU GSV, podcasts), so approachable via technical/cost content.
Andrew Paugh — Senior Engineering Director, Dialpad
Signal: 4 (company-scoped LinkedIn people search of agent-native company Dialpad). Source: https://www.linkedin.com/search/results/people/?keywords=Dialpad%20AI%20director%20engineering | Company size: ~1,200 (est; <2,000). Pain points (INFERRED from role+company): reliability + cost of Dialpad Ai voice/CX agents in production; per-run cost visibility. Challenges: scaling reliable voice-agent infra. Must-haves: production reliability + governance; cost observability. Nice-to-haves: context/token-waste reduction. ICP confidence: High (Senior Engineering Director at agent-native 50-2,000-emp company; confirmed via explicit '@ Dialpad' headline; net-new).
Andrew Qu — Chief of Software (CTO), Vercel
Signal: 1 (ICP writing publicly about agent architecture, tool sprawl and production cost/control). NOTE: LinkedIn/Claude-in-Chrome was NOT connected this run (two retries) — found via web research + primary-source verification instead of the prescribed LinkedIn post search.
Source: https://www.latent.space/p/vercel-agents-new-software (Latent Space interview at AI Engineer World's Fair 2026, Jul 3 2026); https://www.latent.space/p/aiewf26trends; LinkedIn posts "We removed 80% of our agent's tools - Vercel" (https://www.linkedin.com/posts/andrew-qu_we-removed-80-of-our-agents-tools-vercel-activity-7408986983957446656-lm9n) and "AMA with Vercel's CTO: Shipping agents" (https://www.linkedin.com/posts/andrew-qu_ama-with-vercels-cto-shipping-agents-activity-7381767081291268096-qE-5).
Company size: ~847-915 employees (Revelio Labs / Tracxn, 2026). Vercel ships agents itself (eve agent framework, Vercel Sandbox) and hosts agent workloads for customers.
Pain points: agents are unpredictable relative to web apps ("They are not as predictable as web applications... the interaction, interface and outputs are much more dynamic"); tool sprawl degrading agent behaviour — publicly documented cutting ~80% of their agent's tools; infrastructure requirements only surfaced under production load ("A year ago, we did not know sandboxes would become so important, or how much demand there would be for secure code execution and long-running jobs").
Challenges: building a durable framework/harness layer for agents while the requirements keep shifting; secure code execution and long-running job cost/isolation; learning production failure modes after the fact.
Must-haves: visibility into what agents actually do in production; control over tool surface area and context; safe sandboxed execution for long-running agent jobs.
Nice-to-haves: portable skills/knowledge across agent surfaces; standardised agent framework primitives.
ICP confidence: High — C-suite technical leader (Chief of Software / referred to as Vercel's CTO), company well inside 50-2,000 headcount, actively shipping agents in production and publishing about agent cost/control problems.
Signal: 1/4 (senior AI/ML research leader at an agent-active company; surfaced via LinkedIn people/leadership search this run). Source: https://www.linkedin.com/in/seagravesan/ and https://deepgram.com/company/leadership. Company size: ~250-300 (Deepgram, voice AI platform; Series B). Pain points [INFERRED from role+company, not a verbatim quote]: accuracy/reliability of speech-to-action agents in production; controlling inference cost of research-grade voice models at scale; eval/observability of agent behavior. Challenges: closing the gap between demo-quality and production-reliable voice agents; cost of running large models per interaction. Must-haves: reliability + cost visibility per agent run; strong eval tooling. Nice-to-haves: model comparison on cost-per-completed-task. ICP confidence: Medium (VP of Research ~ VP AI/ML; MIT PhD; agent-active company in-band; voice-AI infra caveat).
Andrew Stockwell — Chief AI Officer, Euna Solutions
Signal: 1 (senior AI leader at company scaling agents across product lines). Source: https://eunasolutions.com/resources/euna-expands-ai-leadership-and-capability-for-governments/ ; https://eunasolutions.com/ai-public-sector/. Company size: ~500-700 (ZoomInfo/PitchBook range). Oakville, Ontario, Canada. TITLE CORRECTED THIS RUN: web research listed him as "VP, Artificial Intelligence"; live LinkedIn people search shows "Andrew Stockwell — Chief AI Officer", Oakville ON. Recorded as Chief AI Officer. Agent evidence: Euna named to the 2026 AI 50 list for public-sector AI impact; built a cross-functional AI leadership group (Chief AI Officer plus a separate VP of Applied AI & Productivity) specifically to scale AI across procurement, payments, grants, budgeting and permitting products. Pain points: standing up production AI/agent capability across five-plus distinct government-facing product lines simultaneously — textbook 1-to-many agent scaling. Challenges: public-sector compliance and procurement cycles slow agent rollout; must prove ROI to government budget officers. Must-haves: governance/audit tooling; cost visibility that government finance teams can defend in a budget review. Nice-to-haves: multi-product agent reuse tooling. ICP confidence: Medium — C-level AI mandate, govtech vertical fit and headcount all verified, but FLAG: Euna is PE-owned (GI Partners buyout of former GTY Technology, 2022) rather than VC Series A-C, a deviation from the stated funding criterion. Secondary targets seen live at Euna: Mykola "Myk" Konrad (Chief Product Officer, AI-driven), Ron Houston (Senior Director of Development).
Signal: 4 (ICP writing about shipping agents in production). Source: https://tech.orbitalwitness.com/posts/2026-07-06-no-agent-left-behind/ and https://tech.orbitalwitness.com/posts/2026-03-07-product-engineering-at-orbital/ — title verified live on LinkedIn 2026-08-30 ("CTO at Orbital", London). Company size: ~60-110 (Craft.co ~110, Getlatka 59); London, UK; AI for real estate law; $60M Series B Jan 2026 (Brighton Park Capital, REV/RELX). Pain points: their engineering blog states they have "processed over 100 billion tokens through our AI systems"; document volume grew ~7x in six months (4k to 30k docs/week) while "agent tasks became longer-running and more autonomous"; deploys killing in-flight agent state; "expensive LLM calls fanned out in parallel" being re-executed and paid for twice on retry; hung LLM calls where "the worst failure mode is not an error but a hang". Challenges: noisy-neighbour multi-tenancy across agent workloads; silent failure detection — they ask "how quickly do we notice a failure that arrives silently?"; orchestrating fleets of agents; grading their own agent harness. Must-haves: durable agent state across deploys, dedupe/checkpointing so expensive parallel LLM fan-out is not paid for twice on retry, hang and silent-failure detection. Nice-to-haves: per-tenant cost attribution, automated agent-harness grading/eval. ICP confidence: High — CTO of a Series B agent-native company that publicly documents exactly the cost, reliability and observability problems thealpha.ai addresses; also spoke at AI Engineer World's Fair 2025.
Signal: 1/4 (adapted — LinkedIn auth unavailable this run; identified via web research + verification). Source: https://tracxn.com/d/companies/freed/ ; https://venturebeat.com/ai/freed-says-20000-clinicians-are-using-its-medical-ai-transcription-scribe ; https://sequoiacap.com/article/partnering-with-freed-an-ai-powered-clinicians-assistant/. Company size: ~146 employees (Apr 2026), Series A (~$34M, Sequoia-backed). Freed ships an AI medical scribe used by 20,000+ clinicians in production. Co-founder & CTO. Pain points (inferred from production scale): cost of high-volume ambient transcription/LLM calls per visit; reliability/accuracy of documentation agents at scale; per-clinician / per-visit cost visibility. Challenges: controlling token/compute spend as clinician base grows fast amid rising competition; keeping output reliable in a clinical context. Must-haves: cost-per-run visibility; reliability. Nice-to-haves: observability + eval tooling for the scribe agent. ICP confidence: High — technical co-founder/CTO, agent-native healthcare company in size range. LinkedIn URL not captured this run.
Andy Atwal — Co-Founder & VP of Engineering, AKASA
Signal: 4 (senior eng leader / co-founder at company shipping agentic AI in production). Source: https://www.linkedin.com/in/andy-atwal-6159302/ ; https://www.datanyze.com/people/Andy-Atwal/1956169407 ; https://startupintros.com/orgs/akasa (2026). Company size: ~200 employees; AKASA builds gen/agentic AI for healthcare revenue-cycle (medical coding, prior auth, claims) deployed across 650+ hospitals and 6,500+ outpatient facilities ($120B+ net patient revenue processed). Role: Co-Founder & VP of Engineering (R&D / Healthcare + AI). Pain points (inferred from role + product domain, NOT verbatim quotes): reliability and accuracy of agents on billing/claims data in a regulated setting; observability/auditability of autonomous agent actions; cost efficiency at high claim volume. Challenges: scaling reliable, compliant agents across many health systems. Must-haves: reliability, auditability/governance, control. Nice-to-haves: per-run cost visibility. ICP confidence: Medium-High (named VP of Engineering, ICP title; ~200-emp company in band shipping agentic workflows in production). Same company as Varun Ganapathi (CTO). Found via web research — LinkedIn/Chrome not connected this run.
Signal: 4 (VP Eng at large agent company; LinkedIn people search "Yellow.ai VP engineering AI"). Source: https://www.linkedin.com/in/anikdas/ . LinkedIn headline (his own words): "VP of Engineering at Yellow.ai | Building enterprise-grade agentic AI platforms | Scaling teams & systems globally | Driving innovation in conversational AI." Company size: ~700-1,100 employees (Yellow.ai; enterprise conversational/agentic AI; ICP company already in brain). Pain points (PARTLY from headline, rest INFERRED): scaling agentic AI platforms + teams globally; enterprise reliability of conversational agents; cost at scale. Must-haves: reliability + cost control as agent platform scales. ICP confidence: HIGH — VP Engineering, agent-native, size fits. Caveat: pain partly from headline, partly inferred.
Signal: 4 (ICP AI leader at AI-native company building agents for credit workflows; found via LinkedIn people search 'Head of AI/Director of AI ... agents production'). Source: https://www.linkedin.com/in/aniketdalal/ ; headcount verified via Tracxn/PitchBook (https://tracxn.com/d/companies/credcore). Company size: ~50-60 employees (Tracxn 59 as of Mar 2026; PitchBook 43) — borderline but at/above the 50 floor on most-recent source. CredCore (Series A, $16M) is an AI-native platform turning unstructured credit-deal docs into intelligence for the $5T debt market; models do the heavy reading with senior specialists validating. Pain points (inferred from product): reliability/accuracy of document-reading agents on high-stakes credit deals; cost of running large-context LLM reads across thousands of pages; auditability of AI output. Challenges: hallucination control on financial docs; per-deal inference cost. Must-haves: reliability + explainability of agent output, cost per document/run visibility. Nice-to-haves: automated evals, model routing to cheaper models for easy passages. ICP confidence: Medium (clear Head of AI/ML target title; AI-native, agent-like document intelligence; company size borderline ~43-59).
Signal: 4 (ICP technical leader found via LinkedIn people search at a named agent company). Source: https://www.linkedin.com/search/results/people/?keywords=Ushur%20engineering%20director%20AI . Profile: https://www.linkedin.com/in/anilkumarannem/ . Company size: ~213-284 (agentic Customer Experience Automation platform; Series C). Pain points (inferred from role+company, no direct quote): reliability + cost of agentic CX automation across regulated verticals (insurance, healthcare, finance). Challenges: scaling agent workflows reliably; visibility into per-workflow agent cost. Must-haves: reliability + cost visibility as agent volume scales. Nice-to-haves: anomaly detection; unified agent observability. ICP confidence: Medium (engineering leader (title level unconfirmed) at a confirmed 50-2,000-emp agentic CX company; pain inferred from role+company). Found by scheduled task icp-prospect-signal-scanner run 2026-08-03.
Animesh Kumar — Co-founder & CTO, The Modern Data Company
Signal: 4 (technical decision-maker at a company actively building AI agents). Source: moderndata101.com "The Power Combo of AI Agents and the Modular Data Stack" (co-author; "Animesh Kumar is the Co-Founder and CTO at The Modern Data Company"); themoderndatacompany.com/about. Company size: ~250 (Series B; DataOS; building agentic data systems). Pain points: (company-level) cost/reliability of agentic data workflows; scaling agents on the modular data stack. Challenges: making production agentic systems reliable and cost-efficient over enterprise data. Must-haves: visibility/control over agent cost & behavior in data pipelines. Nice-to-haves: standardized agent orchestration on the data stack. ICP confidence: Medium — technical co-founder/CTO at a Series B agent-building data-infra company; signal is company-level (Signal 4) rather than an individual cost post. Profile URL not verified.
Anique Drumright — Chief Product Officer, Harvey
Signal: 4 (ICP product leader; SaaStr AI Annual 2026 CPO panel on "how to take an agentic demo and turn it into a product enterprises actually trust with real work").
Source: https://www.saastr.com/the-first-44-speakers-for-saastr-ai-annual-2026-the-founders-and-operators-actually-shipping-ai-at-scale/
Company size: ~1,084-1,441 employees (Harvey, legal AI; scaling agents across law firms + enterprises, $11B valuation).
Pain points: turning agentic demos into trustworthy production products; earning enterprise trust in agent output in regulated legal workflows.
Challenges: reliability, auditability, and trust at scale.
Must-haves: agent reliability + auditability + enterprise-grade trust controls.
Nice-to-haves: cost control as agent usage scales.
ICP confidence: Medium (AI-focused product C-suite at agent-native company; note background is product/ops, not core engineering).
Anish Agarwal — Co-Founder & CEO (technical founder), Traversal
Signal: 1/4 (technical founder at AI-native company shipping SRE agents in production). Source: https://www.traversal.com/blog/launch-announcement ; https://www.traversal.com/about ; Tracxn/PitchBook profiles. Company size: ~100 employees (2026); ~$53M raised, Series A (Mar 2026) led by Sequoia & Kleiner Perkins. Product: AI SRE agent platform that analyzes logs, metrics and traces to autonomously root-cause production incidents ("cutting AI research meets world-class engineering"). Founders: Ahmed Lone, Anish Agarwal (CEO), Raj Agrawal, Raaz Dwivedi — a research-heavy technical team (MIT/Cornell). Pain points (inferred from product domain, NOT verbatim quotes): reliability of agents operating on live production systems; controlling/observing autonomous remediation agents; per-incident cost of long-running multi-signal agent investigations. Challenges: making autonomous SRE agents trustworthy/accurate at enterprise scale. Must-haves: reliability, control/guardrails on agent actions in prod. Nice-to-haves: cost-per-run observability. ICP confidence: Medium (title is CEO but a deeply technical research founder fitting the 'technical co-founder' ICP; ~100-emp Series A company shipping agents in production). LinkedIn URL not captured this run — do not fabricate. Note: could not confirm which co-founder holds CTO; captured the technical CEO.
Ankit Jain — Co-Founder & CEO, Infinitus Systems
Signal: 4 (ICP technical co-founder/decision-maker at agent-native company shipping agents in production). Source: https://datainnovation.org/2025/02/5-qs-for-ankit-jain-founder-of-infinitus-systems/ ; https://healthtechremedy.com/episodes/ai-for-healthcare-administration-ankit-jain-on-voice-ai-with-infinitus-systems ; https://tracxn.com/d/companies/infinitus/ . Company size: ~221 employees (Tracxn, Feb 2026). Series C-stage, ~$51.4M raised (Kleiner Perkins, Coatue, GV). Ankit Jain is a technical founder (ex-Google, worked on/around Google Duplex before founding Infinitus). Actively shipping AI agents: production voice AI agents for healthcare with 100M+ minutes handled across benefit verification, prior auth, and back-office calls. Pain points (inferred from product domain + public interviews, NOT a verbatim quote): making autonomous voice agents trustworthy/accurate in regulated healthcare; unit economics (cost per call/minute) at massive volume; visibility into agent behavior and outcomes. Challenges: scaling reliability and compliance as agent volume grows; oversight/guardrails. Must-haves: reliability, trust, cost efficiency at scale. Nice-to-haves: cost-per-run observability and analytics. ICP confidence: High (technical co-founder & CEO; ~221 employees in band; large production agent fleet). Note: CEO rather than pure eng leader — route economic-buyer messaging accordingly; CTO Shyam Rajagopalan is the technical counterpart.
Signal: 4 (ICP CTO at company scaling agents in production). Source: https://innovaccer.com/leadership ; https://medcitynews.com/2026/04/innovaccer-ai-data/ ; https://callsphere.ai/blog/td30-vrt-innovaccer-cx-ai-payer-provider-2026. Company size: ~1,200–1,500 (Innovaccer; healthcare data/AI). NOTE: later-stage funding (beyond Series C) — headcount and agent focus fit ICP. Founding team member, drives tech strategy; earlier VP of Engineering. Innovaccer investing $250M into its agentic AI platform; CX AI agents handle 4M+ patient calls/month for health plans. Pain points (inferred): reliability and cost of agents running at very high call volume; observability of what agents do/cost per run at 4M+ calls/month; scaling many agents across payer/provider workflows. Challenges: production reliability at massive volume; governance in healthcare; cost per run. Must-haves: reliability at scale, cost visibility per run, observability. Nice-to-haves: agents that compound/improve. ICP confidence: Medium (CTO scaling agents in production at high volume; softened by later funding stage; LinkedIn URL unconfirmed — left empty rather than guessed).
Signal: 1/4 (deeply technical founder building an agentic SOC platform shipping AI agents in production). Source: https://techcrunch.com/2026/05/12/exaforce-raises-125m-series-b-to-build-ai-for-catching-and-stopping-cyberattacks-as-they-happen/ ; https://theoutpost.ai/news-story/exaforce-raises-125-m-series-b-... ; https://www.linkedin.com/in/ankur-singla-4863801/ ; Tracxn Exaforce profile. Company size: ~107-130 employees (107 per Tracxn Jun 2026; press notes headcount grew past 130; in 50-2,000 band). Stage: Series B $125M at $725M valuation (May 2026), $200M total (Series A $75M year prior). Ships agents: Exaforce's platform uses AI agents ("Exabots") with deep data analysis to automate security operations; 3-year-old startup, SOC automation in production. Role: Founder & CEO but a deeply technical co-founder — former CTO at Aruba Networks and founder of Volterra (edge/cloud infra, acquired by F5); co-founded Exaforce with Jakub Pavlik (Pavlik already in People Library). Fits ICP "technical co-founder / C-suite" like existing Abhay Parasnis-style entries. Pain points (inferred from domain + product, NOT verbatim quotes): reliability of autonomous SOC agents acting on security telemetry; false-positive/accuracy control; cost and observability of agents processing large data volumes per investigation. Challenges: making agentic security decisions trustworthy, controllable, and cost-efficient at enterprise scale. Must-haves: reliability, human-oversight/control, per-run cost visibility. Nice-to-haves: cross-run observability/compounding quality. ICP confidence: Medium-High (named technical founder-CEO, ~107-130-emp Series B, agentic SOC product in production; caveat: title is CEO though background is technical/CTO-level). Found via web research (LinkedIn/Chrome unavailable this run).
Signal: 1 (ICP publicly writing about agent reliability and cost at scale). LinkedIn verified live 2026-09-02: search result reads "Current: Founder (CEO + CTO) at CoRover", Bengaluru, India, ~30K followers.
Source: https://techgraph.co/interviews/india-bharatgpt-corover-ankush-sabharwal-on-ai-enterprises/ (Oct 2025)
Company size: ~60-85 employees (Tracxn / Inc42 / PitchBook, 2026); Series A, last round Jun 2025. CoRover.ai = full-stack conversational agentic AI platform (BharatGPT); deploys multilingual/multimodal agents at population scale for Indian public-sector and enterprise clients (IRCTC among them).
Pain points: Proving agents are reliable, compliant and explainable under contractual SLAs at population scale. Quote: "Our Prompt Response Layer implements guardrails, audit logs, authenticity checks on content, and data governance practices that guarantee outputs at scale as being reliable, compliant, and explainable. We have even signed SLAs with 99%-100% accuracy... By deploying secure, multilingual, multimodal AI agents at a population scale, our clients drive tangible outcomes: 70% cost reduction, 10X revenue growth."
Challenges: Contractual accuracy SLAs on non-deterministic systems; avoiding "pilot purgatory"; running many agents across languages/modalities without unit economics blowing out; per-client cost accountability when he is selling a cost-reduction outcome.
Must-haves: Reliability guarantees and audit trails per agent, evidence to defend SLA compliance, governance at agent level.
Nice-to-haves: Per-client and per-agent cost/ROI reporting he can put in front of customers.
ICP confidence: High — technical founder holding the CTO seat, Series A, headcount inside the band, agents demonstrably in production at very large scale, and he sells on cost outcomes, which makes cost-per-run visibility a commercial need rather than a nice engineering feature.
Anoop Mohan — Chief Product and Technology Officer (appointed Mar 2026), Augury
Signal: 4 (ICP announcing/shipping a multi-agent product). Found via web research — LinkedIn/Chrome unavailable this run.
Source: https://www.augury.com/media-center/press/augury-shaping-the-future-of-production-with-the-industrial-ai-workforce/ (18 May 2026)
Company size: ~348–355 employees (LeadIQ ~351 as of Jul 2026 across 5 continents; Tracxn 350 as of 30 Jun 2026; PitchBook 348). Industrial AI for manufacturing — machine health + role-based agents. Series F, but headcount is squarely in band.
Agent evidence: announced Augury's "Industrial AI Workforce" — role-based Reliability, Maintenance and Operations agents built on an "Industrial Context Graph," combining Augury machine-health data, AVEVA CONNECT operational context and Google Gemini reasoning. ICL Group is running the Reliability and Operations agents as design partner. Design goal is to "eliminate 'swivel chair' operations." His LinkedIn vanity URL is literally /anoopmohanagenticai/ — agentic AI is his stated identity.
Verbatim: "By combining Augury's Machine data with AVEVA's operational context and Gemini's reasoning capabilities, we've created a foundation with an Industrial Context Graph powering intelligent industrial agents. This allows us to move beyond detecting problems to understanding them and guiding action in real time."
Pain points: gap between insight and action — detection without guided action; plant workers jumping between disparate systems to find answers; agents that don't map to how reliability/maintenance/ops roles actually work.
Challenges: building a continuously evolving context layer tying machine-health signals to process/operational/environmental data; getting agents to reason about cause-and-effect across a production system rather than one machine; validating agent value with design partners.
Must-haves: enriched context graph as the agent substrate; long-context reasoning models; role-scoped agents embedded in daily workflows; partner data integrations (AVEVA CONNECT, Google Cloud).
Nice-to-haves: faster root-cause and yield analysis; self-optimizing/adaptive production loops; measurable autonomy progression from reactive to autonomous production.
ICP confidence: High — CPTO at a ~350-person AI-native manufacturing vendor with a multi-agent product in customer validation as of May 2026, newly appointed (Mar 2026) so actively shaping the stack.
Signal: 4 (ICP technical leader shipping agents). Found via LinkedIn people search ('CTO agentic AI startup') + web verification. Source: https://www.linkedin.com/in/ansarimi/ | Company: Botminds.ai — governed agentic AI platform for lending/credit/regulated document operations; AI agents read borrower docs, spread financials, apply credit policy, produce decision-ready files. Company size: ~53-79 (multiple sources, ~59). Pain points (inferred from role/product, not verbatim quotes): reliability & accuracy of document agents in regulated workflows; explainability/auditability; cost per document/run across high volumes; scaling agents across enterprise customers. Challenges: human-in-loop governance at scale; controlling LLM spend on high-volume pipelines. Must-haves: governed, explainable, auditable agent outputs; cost control per run. Nice-to-haves: observability dashboards, model routing. ICP confidence: High — CTO/co-founder, ~59 employees confirmed, agent-native.
Anton Matsiuk — Director of Cloud Infrastructure, NiCE Cognigy, NiCE Cognigy
Signal: 4 (ICP senior engineering leader at a company shipping AI agents). Source: https://www.linkedin.com/in/anton-matsiuk (found via NiCE Cognigy People directory, filter 'Director'). Company size: 201-500 employees (verified); Cognigy ships contact-center AI agents. Pain points (inferred from Director of Cloud Infrastructure remit at an agent platform): infrastructure and compute cost of running LLM agents at scale, scaling infra as agent/conversation volume grows, no clear per-run cost attribution. Challenges: keeping infra cost efficient while agent workloads compound. Must-haves: cost visibility per agent/run, autoscaling economics. Nice-to-haves: token/context-waste reduction. ICP confidence: Medium (Director of Engineering/Infra owning agent cost + scaling at a verified agent company).
Signal: 3/4 (technical co-founder/CEO of Lovable, a hyperscale AI app-building agent; as one of the largest agent-token consumers in the market, Lovable faces acute LLM cost + reliability pressure — directly relevant to an agent operating/cost-control layer). Source: https://se.linkedin.com/in/antonosika ; Bloomberg (Lovable $400M ARR) ; https://tracxn.com/d/companies/lovable . Company size: ~1,107 employees; Series B, ~$6.6B valuation (NOTE: headcount within 50-2000 but this is a unicorn-scale, late Series B — larger than the typical A-C target). Ex-first-engineer at Sana Labs; creator of GPT-Engineer. Pain points: massive LLM token spend to run the build agent; per-generation unit economics; reliability/quality of agent-generated apps; controlling agent behavior at scale. Challenges: cost per generation at $400M ARR scale; agent reliability; observability into what each run costs/does. Must-haves: per-run cost visibility & control, reliability, observability. Nice-to-haves: model routing/optimization. ICP confidence: Medium (technical CEO shipping an agent product; headcount in range; scale/stage large).
Signal: 1 (ICP senior technical leader at an agent-shipping company; found via web research — LinkedIn/Chrome extension NOT connected this run, so prescribed LinkedIn searches unavailable; pivoted to web + primary-source verification). Source: https://www.boringbusinessnerd.com/startups/aisera ; https://tracxn.com/d/companies/aisera ; https://getlatka.com/companies/aisera. Company size: ~340 employees (2026; range 238–340 across sources). Aisera ships agentic AI ("AI Copilot"/AgentiveGPT) automating IT, HR, finance, and customer service — many agents in production; ~$257M est. ARR; acquired by Automation Anywhere (2025), operates as distinct agent product line. Antonio Nucci is CTO (formerly CTO of Advanced Technologies & AI at Cisco). Pain points (INFERRED from role/company context — NOT verbatim quotes): running large-scale agentic automation across many enterprise workflows implies agent cost-at-scale, reliability/quality in production, and per-workflow cost visibility. Challenges: keeping multi-domain agents reliable at enterprise scale post-acquisition integration. Must-haves: production observability + cost control across a large agent fleet. Nice-to-haves: model routing / token-budget optimization. ICP confidence: Medium-High (verified C-suite technical leader at a 50–2,000-emp agent-shipping company; company is Series D / acquired rather than Series A–C, noted as caveat). NOTE: pain points are inferred, not quoted — LinkedIn engagement content was not accessible this run.
Anubhav Sharma — Head of Agentic AI, Jeeva AI
Signal: 1/4 (LinkedIn people search for ICP agentic-AI leaders; verified via web). Source: https://www.linkedin.com/search/results/people/?keywords=Head%20of%20AI%20Agents | profile https://www.linkedin.com/in/anubhav25. Company size: ~152 employees, Series B (~$27-32M total; investors incl. Marc Benioff, Jack Altman, Sapphire Ventures). Agent activity: Jeeva ships autonomous AI sales agents / "digital workers" that research, enrich, personalize, qualify and automate across the sales pipeline. Pain points (INFERRED from role/company, not a verbatim quote): running many autonomous sales agents at volume; per-agent/per-run cost as agents chain tasks; reliability of unsupervised outreach; visibility into what each agent costs and does. Challenges: keeping autonomous agents accurate and on-budget while scaling from a few to many digital workers. Must-haves: per-run cost visibility + control, reliability guardrails for autonomous outreach. Nice-to-haves: model routing to cheaper models per task, eval/observability. ICP confidence: High (Head of Agentic AI = senior technical AI leader at a ~152-emp Series B agent-native company; textbook ICP).
Signal: 4 (senior AI leader at a company shipping AI agents; surfaced via LinkedIn People directory). Source: https://www.linkedin.com/in/anubhav25/ (Jeeva AI = "digital workers"/AI SDR agent platform). Company size: ~150 (PitchBook 2025; smaller end of range). Pain points (inferred from role — owns agentic AI at a firm running many customer-facing autonomous SDR agents): agent cost blowout as agent volume scales across customers; reliability of autonomous agents in production; token/context waste in long multi-step outreach. Challenges: keeping per-agent unit economics viable at scale; knowing what each agent run costs. Must-haves: cost visibility per agent run; production reliability/guardrails. Nice-to-haves: cross-agent orchestration + shared context. ICP confidence: Medium (title & agent focus strong; company size ~150 per PitchBook but sources vary). NOTE: pain points inferred from role/company, not verbatim quotes.
Signal: 4 (ICP speaking at AI Engineer World's Fair 2026 on building/shipping healthcare AI agents in production). Source: https://www.ai.engineer/worldsfair/schedule ; funding verified via prnewswire (Anterior $40M round, total $64M, Feb 2026). Company size: ~50 employees (Series B, NEA/Sequoia). Company ships an AI agent 'Florence' automating clinical prior authorization for health plans (cut processing time ~74%). Pain points (inferred from role/domain): reliability & accuracy of clinical agents in a regulated setting, per-authorization cost as volume scales, auditability/traceability of agent decisions. Challenges: scaling one agent to many health-plan deployments while keeping quality high. Must-haves: production reliability + explainability/traceability for clinical agents. Nice-to-haves: per-run cost attribution and continuous improvement of agent quality. ICP confidence: High — Head of AI at a 50-person Series B that actively ships agents in production; squarely in ICP.
Apurv Agrawal — Co-Founder & CEO, SquadStack
Signal: 1 (ICP writing about agents in production, cost/security). Source: https://www.linkedin.com/posts/apurvagrawal_agentic-ai-in-call-center-what-actually-activity-7384483896178057216-ykrM ("Agentic AI in call center — what actually works"). Company size: ~187-573 employees (2026), Series B (₹140 Cr / ~$17.6M, Bertelsmann India, Chiratae, Blume); AI-native contact-center platform, Humanoid AI Agents trained on 600M+ real conversations, agents in production for Kotak/Axis/Bajaj. Note: Apurv is CEO (co-founder); Vikas Gulati is co-founder & CTO. Pain points: production accreditation/audit cost of agents replacing crown-jewel workflows, security review, moving agents from pilot to trusted production. Challenges: reliability + trust for customer-facing voice agents at scale, multilingual. Must-haves: production-grade reliability, audit/observability, cost control per interaction. Nice-to-haves: faster agent iteration. ICP confidence: Medium (technical co-founder/CEO, 50-2000 emp, AI-native agent company; decision-maker but not CTO-titled).
Signal: 4 (ICP technical co-founder building agents in production). Source: https://fortune.com/2026/04/27/avoca-ai-agents-missed-calls-hvac-plumbing-roofing-kleiner-perkins-chen-shrivastava-braswell/ ; https://globalindianalpha.com/meet-apurva-shrivastava-indian-origin-engineer-who-turned-a-missed-call-idea-into-1-billion-ai-startup/ ; https://www.ycombinator.com/companies/avoca. Company size: ~85 employees (NYC); $125M+ raised, $1B valuation, Series B; 800+ customers, ~$1B job bookings managed. Studied CS at MIT; prior AI/engineering roles at Apple, Sunshine, and Retool (engineering) — technical co-founder / CTO-equivalent building the agent product. Avoca ships AI voice/chat/SMS/email agents for service industries — multiple agents in production. Pain points (inferred): reliability + latency of real-time voice agents; cost per interaction at scale; agent quality across 800+ customers. Challenges: production reliability of live voice agents. Must-haves: reliability, low latency, cost control per run. Nice-to-haves: per-agent cost visibility, observability. ICP confidence: High (technical co-founder building the agent platform at an 85-person Series B AI-agent-native company; same account as Tyson Chen).
Aravind Bala — Co-Founder & CTO, SeekOut
Signal: 4 (ICP building/shipping agents) — found via web research; LinkedIn/Chrome extension NOT connected this run, so prescribed LinkedIn post/people searches were unavailable; verified via BusinessWire, company site, The Org, Crunchbase.
Source: https://www.businesswire.com/news/home/20260416289383/en/SeekOut-Names-Sean-Thompson-as-CEO-to-Lead-the-Agentic-AI-Recruiting-Revolution ; https://theorg.com/org/seekout/org-chart/aravind-bala
Company size: ~300 employees (Series C, ~$188M raised, founded 2017, Bellevue WA). "Agentic AI recruiting platform" — SeekOut Spot / Recruit ship AI screening, outreach & evaluation agents over 1B+ candidate profiles.
Role note: New CEO Sean Thompson joined May 2026; Aravind Bala continues as CTO leading product, engineering, and innovation (co-founder Anoop Gupta → Exec Chairman).
Pain points (INFERRED from domain/role, not verbatim quotes): running many candidate-screening/outreach agents reliably at scale; per-interaction LLM cost as agent volume grows across 1B+ profiles; output quality/verification so agents deliver "interview-ready" candidates without human rework.
Challenges: reliability & accuracy of autonomous screening agents at enterprise hiring scale; cost control as agent usage expands.
Must-haves: production reliability, cost predictability per run, verifiable agent outputs.
Nice-to-haves: observability into per-agent/per-run cost and quality.
ICP confidence: High — CTO/technical co-founder at a 50–2,000-emp company actively shipping vertical AI agents.
Signal: 4 (ICP building agent platform; AI Engineer World's Fair 2026 speaker). Source: https://www.ai.engineer/worldsfair/schedule — talk "Preferences > Benchmarks: Model Routing for How Teams Actually Build." Company size: ~1,500 employees (public; fits 50-2,000 headcount but note later stage vs. Series A-C). Leads GPU compute, inference-as-a-service, and the DigitalOcean GradientAI platform, which lets developers build/deploy GenAI agents. Pain points: model routing/selection for real-world agent builds, inference cost efficiency at scale, giving teams practical (not benchmark-driven) ways to build agents. Challenges: scaling reliable inference for many customer-built agents, cost-per-run visibility, routing between models. Must-haves: cost-aware model routing, inference reliability/observability for agent workloads. Nice-to-haves: per-agent cost attribution, budget guardrails. ICP confidence: Medium — VP Engineering over an agent-building platform at a right-sized-by-headcount company; caveat: infra provider and public-stage rather than Series A-C AI-native. Note: web research + verification; LinkedIn not connected this run.
Arjun Nagulapally — President & Chief Technology Officer, AIonOS
Signal: 1 (ICP writing/quoted publicly on agent cost control). Found via web research; title VERIFIED live on LinkedIn people search 2026-08-31 ("CXO @ AIonOS... Current: President & CTO at AIONOS", Hyderabad, India).
Source: https://inc42.com/features/the-enterprise-fight-against-runaway-ai-costs/ ; https://tele.net.in/arjun-nagulapally-president-and-cto-aionos/
Company size: ~1,014 worldwide as of Dec 2025 (Revelio Labs). InterGlobe x Assago JV, agentic AI for travel/telecom/logistics. In 50-2,000 band.
Pain points: Runaway AI cost on high-volume agent workloads. Documented a program removing LLM calls from steps that don't need them (account lookups, eligibility checks, field validation belong in a DB or business rule, not a model). On a telecom project processing >1M interactions/month, moved ~30% of routine steps to ordinary software and cut total cost per resolved interaction by 35%.
Challenges: Deciding per-step whether a model call is economically justified across a large agent fleet; no native way to see cost per resolved interaction without building it.
Must-haves: Cost-per-run / cost-per-resolved-interaction visibility at the step level; ability to route deterministic steps away from the model.
Nice-to-haves: Automated detection of steps where the model is being used unnecessarily.
ICP confidence: High — CTO-level, in headcount band, actively running agents at >1M interactions/month, and has an on-record, person-attributed cost-per-run reduction program. Caveat: JV-funded rather than a classic Series A-C cap table.
Arjun Prakash — Co-Founder & CEO (ex-Palantir), Distyl AI
Signal: 4 — builds frontier-lab-grade enterprise AI agents for Fortune 500 (forward-deployed engineering + proprietary agent platform, "Palantir model for the agent era"). Source: https://www.crunchbase.com/organization/distyl-ai ; https://distylai.substack.com/p/nvidia-enterprise-agents. Company size: 159 employees (Tracxn, May 2026); $1.8B valuation; Series A/B; founded 2024. Other technical co-founders: Tianhao "Tom" Hu (ex-Palantir/Anduril), Sahil Patil, Derek Ho. Pain points: scaling many bespoke enterprise agents into reliable production; turning forward-deployed learnings into reusable, controllable agent tooling. Challenges: reliability/consistency of agents across Fortune 500 workflows; visibility into agent behavior and cost at scale. Must-haves: production reliability + governance for enterprise agents. Nice-to-haves: cost attribution per client/agent. ICP confidence: Medium (technical co-founder/CEO; 159 emp; actively shipping enterprise agents; note services+platform hybrid).
Arkadiusz Kwapiszewski — Head of Agent OS (Product), PolyAI
Signal: 4 (product leader owning PolyAI's "Agent OS"; found via LinkedIn people search of PolyAI). Strategically high-relevance: he owns an internal agent operating layer — directly analogous to thealpha's category.
Source: https://www.linkedin.com/search/results/people/?keywords=PolyAI%20head%20of%20engineering%20director%20AI
Company size: ~250-350 employees; PolyAI, Series D, enterprise voice AI agents. In-band on size; stage past Series C noted.
Pain points (inferred): building/operating an internal agent OS — orchestration, control, cost visibility, and reliability across production agents.
Challenges: giving teams visibility/control over what agents cost and how they behave per run.
Must-haves: agent control, cost-per-run visibility, reliability.
Nice-to-haves: guardrails, governance.
ICP confidence: Medium-High — Head of Agent OS (AI-focused product leadership, ICP), agent-native company, in-band size.
Signal: 4 (ICP engineering leader at validated target company Dataiku; sourced via LinkedIn People search, run 2026-08-01). Source: https://www.linkedin.com/in/apichery/ . Company size: ~1,000-1,200 (Dataiku — enterprise AI/ML platform shipping agentic AI: LLM Mesh, Dataiku Answers, AI Agents; AI-native, in-range). Pain points (role/company-contextual): controlling LLM/agent cost across many enterprise workloads; per-agent/per-model spend visibility; production reliability/guardrails at enterprise scale. Challenges: cost attribution & observability across heterogeneous models/agents; reliable agent infra for enterprise. Must-haves: per-agent cost & token visibility; production reliability. Nice-to-haves: model routing/eval automation. ICP confidence: High (VP Engineering at 50-2,000-emp AI-native company shipping agentic products).
Arnie Englander — Co-founder (Tech & Product lead), Model ML
Signal: 4 (technical co-founder shipping agents in production; Signal 1 — public founder POV on AI in finance). Source: https://finance.yahoo.com/news/exclusive-model-ml-financial-research-175821164.html ; https://www.bloomberg.com/news/articles/2025-11-24 ; https://ycombinator.com/companies/model-ml | Company size: ~80 employees (Y Combinator); ~$87.5M total, $75M round (Nov 2025, FT Partners; QED, YC, LocalGlobe); founded 2023 (San Francisco). Model ML = AI agents automating investment-banking/financial-research "grunt work" (analysis, modeling, documents). Co-founders: Chaz Englander (CEO) and Arnie Englander (leads tech & product). | Pain points: (inferred) reliability/accuracy of finance agents on high-stakes work; heavy token/compute cost of multi-step research/modeling agents; per-run cost visibility for enterprise finance customers. | Challenges: scaling many agents reliably in regulated finance; controlling cost-per-run and context waste. | Must-haves: production reliability + auditability; per-run cost tracking. | Nice-to-haves: agent eval/iteration tooling. | ICP confidence: Medium — technical co-founder leading engineering/product at an ~80-emp agent company; no explicit CTO title (hence Medium, not High).
Artur Kuzmin — Head of AI, Squire
Signal: 4 (ICP AI leader at a vertical SaaS building AI; found via LinkedIn people search for ICP titles + agents). Source: https://www.linkedin.com/in/arturkuzmin/ ; headcount verified via PitchBook/LeadIQ (https://pitchbook.com/profiles/company/156488-14). Company size: ~230 employees (PitchBook 232; other sources 201-500). Squire (YC '16) is a Series D vertical SaaS for barbershops (booking, payments, POS). Pain points (inferred): standing up AI/agent features (support, scheduling, ops automation) on top of a payments/booking platform; controlling LLM cost as AI features roll out to a large SMB base; reliability of AI in customer-facing flows. Challenges: proving ROI of AI features; per-feature token/cost attribution. Must-haves: cost visibility per AI feature, reliable production behavior. Nice-to-haves: eval tooling, model routing. ICP confidence: Medium (clear Head of AI target title; in-range Series D SaaS; agent-shipping is inferred from the Head-of-AI mandate, not explicitly stated).
Signal: 1/4 (ICP at AI-native company shipping agents at massive scale). Source: https://en.wikipedia.org/wiki/Anysphere ; Contrary Research report https://research.contrary.com/company/cursor. Company size: ~150–300 employees (2026), AI-native; Cursor powers 50K+ businesses and 1M+ daily active users. Pain points: running multiple coding agents in parallel and "background agents" in production; inference/token cost economics at extreme scale (billions–trillions of tokens); reliability of autonomous multi-file agentic edits across huge codebases. Challenges: agent reliability across very large codebases; controlling inference/token cost per agent run; orchestrating many parallel agents safely. Must-haves: cost-efficient inference at scale; reliable multi-agent orchestration. Nice-to-haves: granular per-agent cost visibility. ICP confidence: High — co-founder & CTO (verified title; Sualeh Asif is CPO, Arvid Lunnemark is CTO) and decision-maker at a ~150–300 person AI-native company shipping coding agents in production. Profile URL left blank (not fabricating LinkedIn URL).
Arvind Jain — Founder & CEO, Glean
Signal: 1/3 (ICP writing about controlling agents, agent observability, and cost). Source: https://www.linkedin.com/in/jain-arvind/ ; Glean "Go Agents" launch + framework for controlling AI agents (nojitter, glean.com/blog/go-agents-launch). Company size: est. ~900-1,100 employees (~$300M ARR May 2026, +89% YoY) — within 50-2,000 band. Technical founder (ex-Google distinguished engineer; Rubrik co-founder). Glean builds/ships dozens of enterprise agents with open interoperability. Pain points: controlling AI agents at scale; full observability into every agent run (inputs, tool calls, LLM decisions, outputs); optimizing quality vs speed vs cost per task; reliable context + guardrails. Challenges: making agents plan/self-evaluate/act reliably; secure, scalable deployment across many agents. Must-haves: Debug/Trace-level observability, per-task model/cost optimization, guardrails. Nice-to-haves: cross-provider model flexibility, agent interoperability. ICP confidence: High — public focus on agent control/observability/cost; ships many production agents. Note: Glean also builds its own agent-control layer, so evaluate as build-vs-buy.
Arvind Rangarajan — Sr. Director of Engineering, Zapier
Signal: 4 (ICP engineering leader at agent-shipping company; found via LinkedIn currentCompany people-search of Zapier engineering leadership). Source: linkedin.com/in/arvindrangarajan (Zapier company id 2418251 people search, past-month). Company size: ~1,000-1,800 (Zapier; shipping Zapier Agents / Zapier Central AI agents across millions of automation workflows). Pain points (INFERRED from role+company stage, not an individual quoted post): agent cost at massive workflow scale (per-run LLM spend multiplied across huge automation volume); no per-agent/per-run cost visibility; reliability of non-deterministic agent steps inside customer automations. Challenges: keeping agent unit-economics viable at platform scale; controlling token blowout from retries/loops in customer-triggered agents. Must-haves: per-run cost attribution + hard cost ceilings; production reliability guardrails. Nice-to-haves: model routing to cut spend; cost dashboards per workflow/agent. ICP confidence: High (Sr. Director of Engineering, product & platform, in-range agent-shipping company).
Arvind Sundararajan — Co-Founder & CTO, Abacus.AI
Signal: 4 (ICP building & shipping enterprise AI agents). Source: https://abacus.ai/about , https://www.crunchbase.com/organization/realityengines . Company size: ~249 (Series C, $140M+ total raised, backers incl. Coatue/Tiger/Khosla; enterprise generative-AI + AI-agent platform 'DeepAgent' / AI Control Center; 1M+ customers incl. Fortune 500). Pain points: deploying & orchestrating dozens-to-hundreds of autonomous agents per org; agent reliability & control at scale; inference/compute cost. Challenges: unifying agent operations under one interface; governance across enterprise data; R&D/compute spend. Must-haves: multi-agent orchestration; agent observability & control; cost management. Nice-to-haves: no-code agent building; unified AI ops. ICP confidence: High (verified CTO, ~249 emp, Series C, AI-native shipping agents in production).
Signal: 4 (ICP CTO at agent-native company). Source: https://www.prnewswire.com/news-releases/eudia-secures-up-to-105m-in-series-a-funding-led-by-general-catalyst-to-transform-legal-work-through-ai-powered-augmented-intelligence-302375331.html ; Tracxn (191 employees as of May 2026). Company size: 191. Series A ($105M, led by General Catalyst). Actively building AI agents: "Augmented Intelligence" / multi-agent platform automating enterprise legal workflows (contract review, compliance, negotiation). Pain points (inferred from domain + market signal, NOT a verbatim quote): reliability and control of agents on high-stakes legal work; scaling agents across Fortune 500 legal teams; visibility into agent cost and behavior. Challenges: accuracy/auditability in regulated legal context; orchestrating many agents. Must-haves: reliability + control + auditability. Nice-to-haves: cost-per-run visibility. ICP confidence: High (technical co-founder/CTO, ex-Cresta; Series A; 191 employees; agent-native).
Ashish Nagar — Founder & CEO, Level AI
Signal: 4 (ICP building/shipping agents in production). Source: https://www.unite.ai/ashish-nagar-ceo-founder-of-level-ai-interview-series/ and https://thelevel.ai/blog/levelai-series-c-funding. Company size: ~135 employees, Series C ($73.1M raised), AI-native contact-center company shipping multiple agents in production (virtual agents, automated QA/scoring, real-time agent assist). Technical founder — IIT Delhi B.Tech, ex-Amazon Alexa conversational-AI team. Pain points (inferred from role + scale of agent fleet, not a verbatim quote): running multiple LLM agents across 100% of customer interactions → inference cost at volume, per-conversation cost visibility, production reliability/accuracy at enterprise scale. Challenges: scaling agent quality across many enterprise deployments while controlling spend. Must-haves: cost-per-run/per-conversation visibility, reliability guardrails. Nice-to-haves: cross-agent observability, compounding quality over time. ICP confidence: High — clean role, confirmed 50-2000 headcount, AI-native shipping 5+ agents.
Ashwin Kulkarni — Director, AI & Engineering, Demandbase
Signal: 4 (ICP publicly identifies as building agentic AI + multi-agent systems; amplifies company agent launches). Source: https://www.linkedin.com/in/ashwin-a-kulkarni/ — headline "Building Agentic AI, Multi-Agent Systems & Enterprise AI Platforms"; reposted Demandbase's AI Chat launch (agent that digs through data and surfaces next-best actions). Found via LinkedIn people search "VP Engineering agents in production LLM observability" (2026-08-29). Company size: ~750–1,129 employees (Tracxn ~987 Jun 2026; Revelio 1,129 Dec 2025). B2B GTM/ABM SaaS, late-stage private. Pain points: running multi-agent systems inside an enterprise SaaS platform where agents query large customer data sets per request. Challenges: scaling from single AI features to a multi-agent enterprise AI platform; 2x co-founder now owning AI + engineering. Must-haves: multi-agent orchestration that holds up on enterprise data volumes. Nice-to-haves: cost/usage visibility per agent feature to price AI Chat-type features. ICP confidence: High — Director-level AI owner, ~1,000-person B2B SaaS actively shipping agent products.
Signal: 4 (ICP technical leader building/shipping agents in production). Source: https://www.linkedin.com/in/sreenivasashwin , https://research.contrary.com/company/decagon , Tracxn Decagon profile | Company size: 434 employees (as of May 2026). Series B+ AI-native conversational-AI platform deploying autonomous customer-experience agents for enterprises. Sreenivas is technical co-founder & CTO (ex-Palantir deployment strategist; co-founded Helia; Stanford MS). Pain points: operating many enterprise customer agents in production with high reliability; per-conversation/per-resolution cost economics (Decagon bills on usage, so inference cost drives margin); ensuring agents behave correctly across many enterprise deployments. Challenges: reliability and quality at enterprise scale; controlling agent behavior; cost visibility per run. Must-haves: production reliability, observability, cost-per-run visibility. Nice-to-haves: agent behavior guardrails/control at scale. ICP confidence: High (CTO / technical decision-maker at a 50–2,000-employee AI-native company shipping agents in production).
Attila Brozik — Chief Technology Officer, DigitalGenius
Signal: 4 (ICP senior technical leader at a company shipping AI agents). Found via LinkedIn people search targeting mid-size autonomous customer-service agent companies — Signal 1-3 content/comment searches this run surfaced mostly vendors/consultants and people already in the brain.
Source: https://www.linkedin.com/search/results/people/?keywords=DigitalGenius%20engineering%20AI%20agents
Company size: ~100-150 employees (ecommerce customer-service automation, founded 2014, ~$25M raised). Within ICP 50-2,000.
Pain points (inferred from role/company context, not a direct quote): autonomous AI agents resolving high-volume ecommerce support tickets; LLM cost-per-resolution economics as ticket volume scales; reliability/accuracy of fully autonomous resolutions in production.
Challenges: keeping per-conversation inference cost predictable across many brand deployments; maintaining resolution quality without runaway token usage.
Must-haves: visibility into agent cost per run/conversation; reliability guardrails for autonomous execution.
Nice-to-haves: cross-deployment cost benchmarking; automated optimization of model/prompt routing.
ICP confidence: High (CTO at a mid-size company whose core product is autonomous AI agents).
Signal: 4 (ICP shipping agents). Source: https://inc42.com/person/notulatul/ ; https://convin.ai/news-collection/series-a-startup-funding-india. Company size: ~150-194 (GetLatka ~151, Growjo ~194, 2026). Bengaluru, India. Series A $6.5M led by India Quotient, Aug 2024. Agent evidence: Convin runs an "AI Agent Platform" with multiple live agent types per its own product pages — a conversation-intelligence agent (real-time transcription, summarization, coaching), an AI Phone Calls agent (automated inbound/outbound calling), and an Automated QA agent. Pain points: keeping multiple concurrent contact-center agents accurate and low-latency across large conversation volumes. Challenges: scaling from a few agent types toward a broader agentic suite while proving ROI to enterprise CX buyers. Must-haves: production reliability, per-agent and per-call cost and performance visibility. Nice-to-haves: unified observability across the different agent product lines. ICP confidence: High — current CTO/co-founder confirmed (IIT-Delhi alum, founding team since 2020), headcount and Series A stage both in band, multiple named agents in production. NOTE: the People Library is already dense with Indian CX/voice-AI companies (Kore.ai, Yellow.ai, Gupshup, Uniphore, Observe.AI, Haptik, Level AI, Skit.ai, Gnani.ai) — Convin is net-new but competes in that same crowded space, so differentiate outreach accordingly.
Signal: 4 (ICP engineering leader at validated target company Dataiku; sourced via LinkedIn People search, run 2026-08-01). Source: https://www.linkedin.com/in/aur%C3%A9lien-coquard-84b35914/ . Company size: ~1,000-1,200 (Dataiku — enterprise AI/ML platform shipping agentic AI: LLM Mesh, Dataiku Answers, AI Agents; AI-native, in-range on headcount). Pain points (role/company-contextual, not from an individual post this run): controlling LLM/agent cost across many enterprise workloads; per-agent/per-model spend visibility; production reliability and guardrails for agents at enterprise scale. Challenges: cost attribution and observability across heterogeneous models/agents; scaling agent infra reliably for enterprise customers. Must-haves: per-agent cost & token visibility; production reliability. Nice-to-haves: model routing/eval automation, regression testing. ICP confidence: High (VP Engineering at 50-2,000-emp AI-native company actively shipping agentic products).
Aurélien Gervasi — Head of Engineering, AI, Back Market
Signal: 4 (senior AI eng leader; found via people search 'Head of Engineering AI agents SaaS'). Source: https://www.linkedin.com/in/aureliengervasi/ . Company size: ~650-750 employees (Back Market, refurbished-electronics marketplace — in range). Role: Head of Engineering, AI. Pain points (inferred from role/company, no direct post captured this run): operationalizing AI/agent features in a consumer marketplace — production reliability and cost efficiency. Challenges: scaling AI features to production with predictable cost. Must-haves: cost/observability tooling for LLM+agent workloads. Nice-to-haves: eval/quality tooling. ICP confidence: Medium (senior technical AI leader at in-range company; 'actively shipping 5+ agents' not directly confirmed this run — verify agent maturity before outreach).
Signal: 3/4 (ICP building agents; explicit agent cost-control engagement). Source: https://www.unifygtm.com/blog/introducing-nextgen-ai-agents ('better performance at a fraction of the previous cost... agents now run at 0.1 credits, a 10x improvement'); https://www.upstartsmedia.com/p/unify-ai-gtm-pipeline-40-million-raise . Company size: Unify ~50-100 employees; Series B ($58M from Thrive, Emergence, OpenAI Startup Fund, Battery; ~$260M val). Pain points: making agents cost-practical at scale; deploying agents across entire outbound GTM workflows. Challenges: agent performance-vs-cost trade-off; evals & reliability across GTM workflows. Must-haves: cost-efficient agent runs at scale; reliable multi-step outbound agents. Nice-to-haves: signal-detection quality; data/partnership layer. ICP confidence: Medium (CEO/co-founder decision-maker at Series B agent company; explicit cost-control signal; technical co-founder Connor Heggie already tracked).
Signal: 4 (ICP building & shipping autonomous AI agents). Source: https://openai.com/index/rox/ , https://theorg.com/org/rox/org-chart/avanika-narayan , https://www.linkedin.com/in/avanika-narayan-a35a02a3/ . Company size: ~133 (unicorn round at $1.2B val; seed led by Sequoia, Series A led by General Catalyst + GV; autonomous AI sales agents in production). Pain points: making LLM agents reliable & efficient for enterprise workflow automation & data wrangling; token/compute efficiency (Stanford Hazy Research work on efficient/on-device LMs — 'Minions'). Challenges: agent reliability at scale; efficiency per task; grounding in enterprise data. Must-haves: efficient & reliable agent execution; workflow automation. Nice-to-haves: cost-efficient / on-device inference. ICP confidence: High (verified technical co-founder / AI Lead; ~133 emp; AI-native shipping agents).
Aviad Berman — Sr Director, Engineering — Platform, Data & AI, Melio
Signal: 4 (ICP speaking publicly about building/shipping agents). Title verified live on LinkedIn people search 2026-09-02: "Sr Director, Engineering - Platform, Data & AI at Melio", Israel.
Source: https://langtalks.ai/events/ai-sdlc — talk "Building Multiplayer Software Engineering" (Aug 2026, Tel Aviv) on what agents can own across coding, testing, deployment and ops. LinkedIn URL href-verified from search results.
Company size: ~693 (Tracxn, May 2026). Tel Aviv / New York. B2B fintech (SMB payments), Series D-stage but headcount well inside the 50-2,000 band. NOTE: LinkedIn slug /company/melio/ is a different Luxembourg company — do not use for headcount.
Pain points: He owns the platform layer that agents run on across the SDLC. That is exactly the seat where per-run cost, guardrails and reliability gaps surface first — agents spreading across many engineering teams with no central spend or telemetry layer.
Challenges: platformising agents so many teams can use them safely; adding guardrails and spend control without slowing developers down; proving the agent programme pays back.
Must-haves: a central gateway / run-level observability for agent executions; per-team budgets and caps.
Nice-to-haves: eval harness; failure-mode analytics; adoption reporting to leadership.
ICP confidence: Medium-High — Sr Director (Director+), ~693 employees (in band), explicit ownership of an agent platform, active public speaking on it. Slight discount: fintech rather than AI-native, and pain is evidenced by role + talk topic rather than a direct cost quote. Net-new name AND net-new company.
Aviad Sharfshtein — Senior Director, Engineering Group Lead, Gong
Signal: 4 (ICP Senior Director / Engineering Group Lead at an agent-shipping company). Source: https://www.linkedin.com/in/aviad-sharfshtein-0294124/ (found via LinkedIn people search 'Gong director engineering AI agents'). Company size: Gong ~1,100-1,300 employees (independent; revenue-intelligence platform shipping AI agents for revenue teams). Pain points (inferred from role, not a verbatim quote): leading an engineering group delivering reliable production AI/agent features; balancing quality, latency and LLM cost. Challenges: production reliability and cost control of agent features at scale. Must-haves: reliability + observability across agent features; spend visibility. Nice-to-haves: per-agent cost attribution and anomaly alerts. ICP confidence: High (Senior Director / Eng group lead at an in-range, independent agent-shipping company).
Ayush Pallav — Director - AI Voice and Infrastructure, Level AI
Signal: 4 (ICP AI/infra leader at an agent-shipping company; company-scoped LinkedIn people search "Level AI engineering").
Source: https://www.linkedin.com/search/results/people/?keywords=%22Level%20AI%22%20engineering (profile https://www.linkedin.com/in/ayush-pallav/)
Company size: Level AI ~200-400 employees (Series C; contact-center conversational + voice AI agent platform, IIT Guwahati alum).
Pain points [INFERRED from role + company, NOT a verbatim post]: real-time voice-agent latency vs. cost trade-offs at scale; per-run cost visibility across large call volumes; production reliability of streaming voice agents.
Challenges: keeping voice-agent infra reliable and cost-efficient as concurrency and agent count grow.
Must-haves: per-run/per-agent cost observability; reliability guardrails for real-time voice agents.
Nice-to-haves: model/routing optimization to cut inference cost.
ICP confidence: High — Director owning AI Voice + Infrastructure at an independent 50-2,000-emp agent-native company; net-new.
Signal: 4 (technical co-founder/CTO at a Series B company shipping AI agents in production). Source: https://gc.ai/blog/gc-ai-raises-60-million-series-b-to-give-every-company-a-legal-advantage ; https://www.crunchbase.com/person/bardia-pourvakil ; https://www.bloomberg.com/profile/person/24921333 (Bloomberg: "Bardia Pourvakil CTO/Co-Founder, General Counsel AI Inc") ; https://gc.ai/company/about. Company size: ~120 employees (Apr 2026, per Tracxn; in 50-2,000 band). Stage: Series B $60M (Nov 2025); GC AI (San Francisco, founded 2023). Ships agents: GC AI's product delivers AI agents that redline contracts directly in Microsoft Word and assist with legal review/negotiation; serves 1,000+ legal teams (incl. News Corp, Nextdoor, Zscaler, Vercel). Role: Co-Founder & CTO (technical) — clear ICP. Pain points (inferred from domain, NOT verbatim quotes): reliability/accuracy of agents editing legal documents; governance/auditability of autonomous legal actions; per-run cost and latency at scale across many customer workspaces. Challenges: keeping agent outputs accurate and controllable in a regulated, high-stakes domain. Must-haves: reliability/accuracy, control & auditability. Nice-to-haves: cost/observability per run. ICP confidence: High (named technical co-founder/CTO, ~120-emp Series B, agent-native legal product in production). LinkedIn vanity appears as "bardia-pourvakil" (seen on his LinkedIn activity URL) but exact /in/ profile URL not verified — left blank to avoid fabrication. Found via web research (LinkedIn/Chrome unavailable this run).
Barry McCardel — CEO & Co-founder (ex-Palantir, technical), Hex
Signal: 4 (ICP writing about building agents at a qualifying company). Source: LinkedIn posts ("We got rid of our AI product team", Hex) + $70M Series C (https://hex.tech/blog/series-c/). Company size: 263 (May 2026), Series C, $172M raised. Pain points: operationalizing multiple production agents (Notebook Agent, Generative Apps agent, Threads); how to structure AI/agent work org-wide; making agents reliable enough for analysts to trust. Challenges: scaling agent capabilities across data/warehouse workflows without breaking. Must-haves: reliable multi-agent orchestration over governed data. Nice-to-haves: cost/latency visibility per agent run. ICP confidence: High — Series C, 263 employees, 5+ agents in production, co-founder decision-maker actively posting about agent strategy.
Barry Shteiman — Co-Founder & CTO, Radiant Security
Signal: 3 (ICP in the competitor-adjacent agentic-SOC / observability-and-control space) + 4 (shipping agents in production). Source: https://www.linkedin.com/in/barryshteiman/ ; https://radiantsecurity.ai/about-us/ ; https://www.crunchbase.com/person/barry-shteiman ; Tracxn headcount (Apr 2026 ~60). Company size: ~60 (Crunchbase 51-100; Tracxn ~60 Apr 2026; Series A ~$15M, ~$21M total; founded 2021). Actively shipping agents: Radiant runs an agentic AI SOC that autonomously triages, investigates and responds to security alerts at analyst-grade precision. Shteiman = technical co-founder & CTO (security veteran). Pain points (inferred from product/positioning, NOT verbatim quotes): reliability/precision of autonomous investigation agents (false positives/negatives carry security risk); token/inference cost as alert-triage agents run at 24/7 SOC volume; explainability/auditability of agent decisions for SOC governance. Challenges: scaling many concurrent investigation agents while controlling cost per alert. Must-haves: control/guardrails, cost-per-run visibility, audit/explainability. Nice-to-haves: agents compounding on analyst feedback. ICP confidence: High (technical co-founder + CTO; in-range headcount 60; agentic product in production; cost-per-run pressure inherent to always-on SOC agents).
Ben Allen — Co-founder & CTO, Omnea
Signal: 4 (ICP technical co-founder building/shipping agents at scale; found via web research — LinkedIn/Chrome not connected this run). Source: https://www.omnea.co/resource/series-b-announcement ; https://www.prnewswire.com/news-releases/omnea-raises-50m-to-make-procurement-every-cfos-competitive-advantage-302559153.html ; https://theorg.com/org/omnea/org-chart/ben-freeman (leadership: Ben Freeman CEO, Ben Allen co-founder & CTO). Company size: ~196 employees (Jun 2026), London (+ NY); founded 2022; Series B $50M (Sept 2025, Insight Partners & Khosla; total >$75M). Product: 'agentic operating system for procurement' — AI agents for procurement intake and orchestration; customers incl. Spotify, Wise, MongoDB, Monzo, Adecco, Albertsons; 5x revenue and tripled headcount in 12 months. Pain points: reliability of autonomous procurement agents acting across enterprise systems; governance/auditability of agent actions in a CFO-owned, compliance-heavy workflow; scaling multi-step agent orchestration. Challenges: making agents reliable and controllable across many integrations; cost/observability as agent volume scales. Must-haves: reliability, governance/control, audit trails. Nice-to-haves: per-run cost visibility, context efficiency. ICP confidence: High (co-founder & CTO at a ~196-emp Series B company actively shipping enterprise agents; squarely in the 50–2,000 band).
Ben Avitouv — Field CTO, Wonderful
Signal: 1 (ICP author writing directly about operating agents at scale vs. building them).
Source: https://www.wonderful.ai/blog-articles/the-hard-part-of-ai-agents-isn%E2%80%99t-building-them (1 Apr 2026), byline on wonderful.ai.
Company size: 350 employees as of 12 Mar 2026, targeting ~900 by year-end 2026 — verbatim from Insight Partners Series B release (https://www.insightpartners.com/ideas/wonderful-raises-150m-series-b-to-accelerate-enterprise-ai-adoption-in-30-markets/). Series B, $150M / €129.8M at €1.7B. Amsterdam-based enterprise AI agent platform.
Pain points: (verbatim) "Today's tooling is largely optimized for building agents, not for operating them responsibly at scale." Also (verbatim) "The only way to catch those breaks before customers do is through continuous operational visibility: monitoring tool failures, abnormal conversation volumes, unexpected behavior, and anything else that signals something is off."
Challenges: Agents silently break in production; existing tooling stops at build time; failures surface to customers before they surface to the team.
Must-haves: Continuous operational visibility across agent fleet; tool-failure monitoring; anomaly detection on agent behavior and volume.
Nice-to-haves: Pre-customer-impact alerting; consolidated operating view across dozens of deployed agents.
ICP confidence: High — Field CTO title is the exact buyer seat, headcount verified from primary source and inside band, company is shipping agents to enterprises across 30 markets, and the pain statement is a near-verbatim match to Alpha's positioning ("operating layer, not build layer").
Note: title carries regional scope ("Field CTO Israel" per LinkedIn snippet) — confirm scope before outreach.
Ben Dixon — Co-Founder & CTO, Sona
Signal: 4 (ICP technical co-founder at an AI-native company building/shipping agents; found via 2026 agent-startup research + LinkedIn verification).
Source: https://www.linkedin.com/in/talkingquickly/ ; company context https://www.prnewswire.com/news-releases/sona-next-generation-workforce-management-for-frontline-raises-27-5m-series-a-led-by-felicis-302145146.html ; Series B https://www.felicis.com/blog/sona-series-a
Company size: mid-sized (est. ~100–200; London); founded 2021; Series A $27.5M (2024) + Series B $45M (April 2026, led by N47; Gradient/Google, Felicis, Northzone). AI-native frontline workforce platform (agentic scheduling, forecasting, payroll).
Pain points (inferred from role/company, not quoted): agentic scheduling/forecasting across thousands of frontline workers must be reliable and cheap per run; scaling agent workloads across many enterprise customers; visibility into agent cost/reliability.
Challenges: keeping agent decisions accurate and cost-controlled at scale; observability across a growing set of workforce agents.
Must-haves: cost-per-run visibility; reliability monitoring for production agents.
Nice-to-haves: token/context optimization; per-customer cost attribution.
ICP confidence: Medium-High — Co-Founder & CTO leading the AI platform at a Series B AI-native company in the ICP band; exact headcount not publicly confirmed (estimated 100–200).
Ben Holmes — Director of Agentic Ecosystem, Dialpad
Signal: 4 (company-scoped LinkedIn people search of agent-native company Dialpad). Source: https://www.linkedin.com/search/results/people/?keywords=Dialpad%20AI%20director%20engineering | Company size: ~1,200 (est; late-stage, ~$2.2B val, <2,000). Pain points (INFERRED from role+company; title 'Agentic Ecosystem' is strongly on-thesis): orchestrating and controlling a growing ecosystem of voice/CX AI agents; per-interaction cost and reliability at production scale. Challenges: scaling agent ecosystem reliably; cost visibility per agent/run. Must-haves: production reliability, governance, per-run cost attribution. Nice-to-haves: token/context-waste control. ICP confidence: High (Director of Agentic Ecosystem at agent-native 50-2,000-emp company; confirmed via explicit '@ Dialpad' headline; net-new vs only the CTO in brain).
Ben Houghton — Head of Graph Data Science, Quantexa
Signal: 4 (ICP writing publicly about the prototype-to-production gap) — captured from his own LinkedIn post, ~2 weeks old.
Source: https://www.linkedin.com/in/ben-houghton-b2060951/recent-activity/all/ ; found via https://www.linkedin.com/search/results/people/?keywords=Quantexa%20Head%20of%20Applied%20AI ; company page https://www.linkedin.com/company/quantexa/people/
Company size: 890 associated members on LinkedIn (28 Aug 2026); company page band 501-1K — inside the 50-2,000 range. HQ London, UK. Decision Intelligence platform for financial crime, risk and growth; sold to global banks. Company is publicly positioning as "AI-native".
Pain points / expressed signal (VERBATIM from his own post announcing his Big Data London talk, 23 Sept): "Too much graph ML work fades between the notebook and the production pipeline. I'll discuss where these projects break down, and how Quantexa's Knowledge Graph gives pipelines the context and scalability they need to survive the trip." — i.e. he is publicly framing the problem as production survivability and context/scalability, which is adjacent to the agent-reliability wedge.
Challenges: getting probabilistic models from prototype into governed production pipelines inside a company whose customers are regulated global banks — reliability and explainability bars are unusually high.
Must-haves: production-grade pipelines with sufficient context and scale; demonstrable behaviour in production rather than in a notebook.
Nice-to-haves: reusable context/knowledge-graph substrate that downstream models and agents can share.
ICP confidence: Medium — Head-of level at a confirmed in-band, AI-native company, and he expressed a real prototype-to-production pain in his own words. Held at Medium rather than High because his stated domain is graph ML/knowledge graph rather than LLM agents specifically, and no cost or token signal was captured. Worth qualifying whether Quantexa's agent products (Q Assist and similar) sit under or beside his remit; if not, John Keightley (Head of Product, Platform) and Ahsan Mehmood (APAC Head of Data & AI Solutions) are the alternate entry points identified at the same account.
Verification notes: headcount from LinkedIn People tab. Profile URL confirmed by DOM href extraction. NET-NEW company — Quantexa returned zero matches on dedup.
Ben Liebald — VP of Engineering (Core Product Engineering & AI), Harvey
Signal: 4 (ICP eng leader at AI-native company shipping agents in production). Source: https://www.harvey.ai/company and https://www.linkedin.com/in/liebald/ and https://theorg.com/org/harvey-ai/org-chart/ben-liebald. Company size: ~400-600 (Harvey, legal AI, Series D/E, ~$3B+ valuation, AI-native shipping legal research/drafting agents to Fortune 500 law firms + in-house teams). Pain points: making agentic legal workflows reliable and accurate at enterprise scale; scaling agent product engineering as the company grows "rapidly" (his own hiring posts). Challenges: building a scalable AI platform, security standards, and integrated agent features across many enterprise customers; ex-Figma Sr Dir Eng + ex-Branch CTO so cares about production reliability & infra. Must-haves: reliability/accuracy of agent outputs in high-stakes legal context, ability to scale agent workloads. Nice-to-haves: cost/token efficiency per agent run as usage scales. ICP confidence: High (VP Engineering, decision-maker, AI-native company actively shipping agents in production, in target size band).
Signal: 1/2 (ICP leader quoted in R&D World feature on 'token maxxing' vs production value).
Source: https://www.rdworldonline.com/how-can-organizations-move-ai-from-token-maxxing-to-production-value/
Company size: ~808-920 employees (public BI/AI platform; ships agentic apps via App Catalyst + MCP, repositioning as governance/control-tower over Anthropic/OpenAI/Google models).
Pain points: agent/token cost bill shock ("All of a sudden you get the bill and ask, why are we spending all this money?"); paying repeatedly for tokens ("I don't want to pay for tokens every time my CEO looks at it"); pilots stuck in demo phase / 'pilot purgatory'.
Challenges: moving agents from prototype to governed production — "you can't vibe code governance, security, and distribution"; can't distribute vibe-coded apps to thousands of users.
Must-haves: cost visibility/control on token spend, deterministic lock-in of good outputs, governance/control-tower layer over models.
Nice-to-haves: per-query cost attribution, model routing.
ICP confidence: Medium (senior C-suite AI leader actively shipping agents; headcount qualifies, though Domo is public rather than Series A-C).
Benjamin Gleitzman — Co-Founder, CTO & VP of Engineering, Replicant
Signal: 3/4 (ICP technical co-founder at agent-native voice company; competes in same production-agent space as tracked competitors). Source: https://www.crunchbase.com/person/benjamin-gleitzman ; https://www.replicant.com/about ; https://techcrunch.com/2022/04/26/call-center-automation-software-vendor-replicant-raises-78m/ . Company size: ~110+ (team of 110+ engineers/PMs/conversation designers). Series B ($78M, Apr 2022, led by Stripes; Salesforce Ventures, Norwest; ~$550M post-money). Independent. Actively building AI agents: Replicant runs autonomous voice AI agents ("Thinking Machine") that resolve customer-service phone calls at carrier scale in production. Pain points (inferred from product domain, not verbatim): keeping voice agents reliable and low-latency at high call volume; controlling per-call cost; observability into what agents do on live calls. Challenges: reliability + latency + cost at contact-center scale; handling messy real-world voice input. Must-haves: reliability in production, cost-per-resolution control, latency. Nice-to-haves: cross-agent observability, efficiency gains. ICP confidence: High (technical co-founder + CTO/VP Eng at 110+ person Series B voice-agent company shipping agents in production; in target size band).
Benjamin Levick — Head of Operations & Internal AI, Ramp
Signal: 1 (ICP writing about agent costs/scale & tying spend to business value). Source: https://www.linkedin.com/posts/benjamin-levick_ai-operations-automation-activity-7364717316200099840-dx08 (also X post: 'I lead Ops and Internal AI at Ramp... shipping 1,000+ internal apps and agents every month, mostly with non-engineers'). Company size: Ramp ~1,300 employees (fintech; note: later-stage than Series A-C). Pain points: connecting token/agent spend to business value; AI usage up 6,300% YoY; 1,000+ internal agents shipped monthly mostly by non-engineers -> governance & cost attribution. Challenges: auditing agent outputs at scale, controlling runaway internal-agent spend, measuring ROI per agent. Must-haves: visibility into what agents cost and deliver; cost-per-agent/per-run attribution. Nice-to-haves: enabling non-engineers to build reliable agents safely. ICP confidence: Medium (size & agent-building fit strongly; very strong personal cost-vs-value signal; company is later-stage than A-C).
Benjamin Mayr — VP, Head of Architecture (Co-founder & Chief Software Architect of Cognigy), NiCE Cognigy
Signal: 4 (company-scoped LinkedIn people search — ICP technical leader at agent-shipping company). Source: https://www.linkedin.com/in/benjamin-mayr-660294a7 (via https://www.linkedin.com/company/nicecognigy/people/?keywords=director). Company size: Cognigy unit 201-500 employees (now NiCE Cognigy — 'AI Agents for Your Contact Center'; acquired by NiCE ~$955M, parent >2,000). Pain points (INFERRED from role+company): scalable, reliable architecture for enterprise voice/chat AI agents in contact centers; cost and reliability of agents at contact-center scale; integrating agent platform into NiCE post-acquisition. Challenges: keeping a decentralized, highly scalable agent platform reliable and cost-controlled across large enterprise deployments. Must-haves: reliability guardrails + cost visibility at scale. Nice-to-haves: routing/model optimization. ICP confidence: Medium (VP + Chief Software Architect = strong ICP technical leader and agent-platform builder; but parent NiCE exceeds the 2,000-emp ceiling — flagged, consistent with existing Cognigy entry Klaus Krogmann. Net-new).
Beyang Liu — CTO & Co-founder, Sourcegraph (Amp)
Signal: 1 (ICP writing/speaking about agent token waste & inference cost) + adjacent to competitor tooling (MCP). Source: https://a16z.com/podcast/from-code-search-to-ai-agents-inside-sourcegraphs-transformation-with-cto-beyang-liu/ + https://www.youtube.com/watch?v=_JBT9d2jdI8 (Flowing with Agents — Amp and agentic coding). Company size: ~100-150 (est.; mid-size, post-2024 restructuring). Pain points: token waste and inference cost in coding agents; managing agents built on multiple frontier models; wielding coding agents as a new skillset (90% of his code now from agents). Challenges: controlling cost/quality tradeoff across sub-agents and models; reducing token waste via MCP server. Must-haves: token/inference cost visibility, multi-model orchestration control. Nice-to-haves: cross-model routing efficiency, replayable agent traces. ICP confidence: High (CTO/technical co-founder at agent-building company; explicit token-waste & inference-cost language).
Signal: 4 (ICP senior technical leader at an agent-native company; discovered via LinkedIn people-search of companies actively shipping AI agents) | Source: https://www.linkedin.com/in/bharatkumarrekha | Headline: AVP Engineering @ Kore.ai — Building Enterprise AI Search & Knowledge Platforms | Ex-Microsoft | Company: Kore.ai — actively building/shipping AI agents in production | Company size: ~1,000–1,200 (estimate) | Pain points (INFERRED from role/company context — not a verbatim quote observed this run): agent cost blowout across high-volume voice/chat interactions; no per-run/per-agent cost visibility; production reliability & hallucination control on an enterprise multi-agent platform | Challenges: scaling from a handful of agents to a large multi-agent enterprise platform without cost/reliability regressions | Must-haves: per-run/per-agent cost visibility and budget controls; production reliability guardrails; observability across many agents | Nice-to-haves: context/token-waste optimization; cross-agent benchmarking | ICP confidence: High (senior technical/eng leadership at a 50–2,000-employee company shipping AI agents)
Signal: 4 (ICP senior technical/product leader at agent-shipping company). Source: https://www.linkedin.com/in/bharath-shankar-143b1914/ ; https://in.linkedin.com/company/gnani-ai . Company size: ~199-244 employees; Series B (Mar-2026). Voice-first agentic AI, 200+ enterprises, 30M+ interactions/day. Pain points (INFERRED): productizing reliable agents at scale; per-agent cost/observability; balancing quality vs cost across languages/customers. Challenges: shipping and maintaining a large agent catalog reliably. Must-haves: reliability + cost/observability tooling. Nice-to-haves: compounding quality. ICP confidence: MEDIUM (C-level product+engineering leader at Series B agentic-voice company; second contact at Gnani, distinct from CTO).
Bihan Jiang — Director of Product, Decagon
Signal: 4 (ICP writing about shipping/operating agents in production). Source: https://decagon.ai/blog/autopilot (Jun 9, 2026, authored by her). Title and employer confirmed live on LinkedIn 2026-08-31 — "Bihan Jiang · Director of Product @ Decagon · San Francisco · 19K followers" (exact profile URL not exposed in search results; left blank rather than guessed). Company size: existing People Library records put Decagon at ~200–400 and 501–1K depending on source — in band either way. Stage: Series A confirmed at decagon.ai/blog/series-a ($35M, Accel + a16z); later rounds reported but not primary-source verified this run. Pain points: Jiang writes that even with AI analyzing 100% of interactions, DECIDING which agent improvements matter and manually remediating them is the bottleneck. She states that every update to agent logic risks downstream side effects — a fix in one area quietly breaking behavior in another — so teams either invest heavily in regression testing for every change or ship with incomplete confidence and find out later. Challenges: regression-testing every agent logic change at scale; manual remediation not scaling with agent count; earning enough confidence to ship changes fast. Must-haves: golden test sets, automated regression detection across agent versions, a human approval gate before production. Nice-to-haves: a self-validating iteration loop, blast-radius analysis for agent logic changes. ICP confidence: Medium (Director level qualifies, company is a genuine agent-native company in the size band with multiple existing People Library records, and the reliability/regression signal is first-person and self-authored — downgraded because she is product rather than engineering, and the pain is change-safety rather than cost). NOTE: net-new person; Decagon itself is well-represented in the library.
Bikash Agrawal — CTO, Simplifai
Signal: 4 (ICP technical leader at a qualifying company shipping agentic AI). NOTE: LinkedIn/Chrome unavailable this run; found via web research + primary-source verification. Source: Simplifai leadership announcements + product pages (simplifai.ai; fintech.global Feb 2025 "Simplifai launches Agentic AI"). Company: Simplifai (Oslo, Norway; founded 2017) — agentic AI for insurance/banking automation (claims handling, underwriting, customer service); launched Agentic AI Feb 2025 and expanded the platform since. Company size: ~77–130 employees (2026; Tracxn 77, PitchBook 92, ContactOut 130) — within ICP band. Leadership: Artem Gonchakov (CEO), Dr. Bikash Agrawal (CTO), Nils Slottet (Chief Solution Officer). Pain points (inferred from role/domain, not a personal quote): reliability/accuracy of insurance-claims and underwriting agents in production; controlling LLM cost across document-heavy agentic workflows; scaling agents across multiple insurer clients. Challenges: regulated-industry accuracy, data privacy, cost predictability. Must-haves: reliability, guardrails, cost/observability per run. Nice-to-haves: orchestration + eval tooling for multi-agent workflows. ICP confidence: Medium-High (confirmed CTO at an agentic-AI vendor; headcount in band across sources). Pain framing inferred from documented product/domain, not a fabricated quote.
Signal: 1 & 4 (prominent ICP voice publicly writing/speaking about deploying & controlling AI agents at scale; company ships agents). Source: https://abacus.ai/bindu-reddy , https://ceoworld.biz/2026/02/06/bindu-reddy-building-the-ai-super-assistant-your-agi-control-center/ , https://www.linkedin.com/in/bindureddy . Company size: ~249 (Series C). Pain points: orgs deploying 50-500 agents that autonomously act on enterprise systems; agent control, reliability & cost. Challenges: building an 'AI Control Center' to unify agent ops; keeping agents reliable & cost-effective at scale. Must-haves: agent control center; observability; cost visibility. Nice-to-haves: unified interface across AI ops. ICP confidence: Medium-High (technical co-founder & decision-maker; ex-AWS GM of AI Verticals, ex-Google Head of Product; ~249 emp Series C AI-native).
Binny Gill — Co-Founder & CEO (ex-Nutanix CTO; technical), Kognitos
Signal: 4 (ICP technical founder at agent-native company). Source: https://www.kognitos.com/news/kognitos-launches-neurosymbolic-ai-platform-for-automating-business-operations-backed-by-25m-series-b/ ; https://techcrunch.com/2023/11/30/kognitos-raises-35m-to-help-businesses-automate-back-office-processes/ . Company size: ~100 (as of Mar 2026). Series B ($25M, June 2025; $51.8M total). Actively building AI agents: deterministic/"neurosymbolic" agentic automation platform executing business-ops logic in plain English across 130+ enterprise integrations. Pain points (inferred from domain + market signal, NOT a verbatim quote): making agent execution predictable/deterministic for enterprise back-office; controlling agent behavior at scale; reliability of multi-step agent runs. Challenges: enterprise trust in autonomous agents; auditability. Must-haves: reliability + control over agent behavior; auditable execution. Nice-to-haves: cost visibility per automation run. ICP confidence: High (technical founder/CEO and ex-Nutanix CTO; Series B; ~100 employees; agent-native).
Blake VanLandingham — VP of Engineering, Canary Technologies
Signal: 1 (ICP at company shipping agent fleet). Source: https://boutiquehotelnews.com/news/technology/canary-technologies-ai-agent-builder/ ; https://hoteltechnologynews.com/2025/06/canary-technologies-raises-80-million-to-expand-its-global-presence-and-enhance-ai-powered-hotel-solutions/. Company size: ~371-374 (Revelio Labs 2026). Series C, $80M June 2025. LIVE LINKEDIN VERIFIED this run — "Blake VanLandingham — VP of Engineering at Canary Technologies", Overland Park KS. Agent evidence: Canary launched "Canary AI Agent Studio", a hospitality-specific AI agent builder; $80M Series C raised explicitly to accelerate AI guest technology. Pain points: scaling from a single guest-messaging bot to a fleet of purpose-built agents across front desk, upsell and guest-service workflows; needs per-agent cost/run visibility as hotel customers deploy multiple agents. Challenges: reliability of agents handling real guest interactions at scale; integration across dozens of hotel PMS/POS systems. Must-haves: production observability, cost-per-run tracking, agent orchestration reliability. Nice-to-haves: eval tooling for guest-facing agent quality. ICP confidence: High — title, headcount, funding stage and explicit agent-builder product all verified; hospitality vertical is an under-covered segment in the People Library.
Signal: 4 (ICP CTO at agentic-AI company scaling agents past pilot). Source: https://www.netomi.com/ ; https://craft.co/netomi/executives ; https://wtvbam.com/2026/04/30/ai-customer-service-startup-netomi-raises-110-million/ . Company size: ~170 (as of 2026 Series C). Series C ($110M, Apr 2026, led by Accenture Ventures; prior $30M Series B 2021). Independent. Actively building AI agents: Netomi is an enterprise agentic-AI customer-experience platform ("built for what comes after the pilot") automating support for United, Delta, Paramount, DraftKings — multiple production agents. Pain points (inferred from positioning, not verbatim): moving agents from pilot to reliable production at enterprise scale; controlling cost/quality across high-volume customer conversations; visibility into agent behavior. Challenges: enterprise-grade reliability, guardrails, and cost control at scale across many brands. Must-haves: production reliability, cost-per-resolution control, observability/governance. Nice-to-haves: efficiency gains, compounding improvement. ICP confidence: High (CTO / technical decision-maker at 170-person Series C agentic-AI company shipping agents in production; in target size band).
Bogdan Pietroiu — Co-Founder & CTO, DRUID AI
Signal: 4 (ICP at company actively building/shipping agents; posts publicly about DRUID's agentic AI, the DRUID Agent Marketplace, and enterprise automation). Source: https://www.linkedin.com/in/bogdanpietroiu/ and DRUID posts e.g. https://www.linkedin.com/posts/druidai_how-the-druid-marketplace-works-for-you-activity-7384172308069875712-Vkkr ; company funding context https://therecursive.com/druid-ai-46m-funding-valuation-276m/ . Company size: ~200-300 (Series C, ~$47.5M raised, ~$300M valuation; Bucharest-based, global). Pain points (inferred from public product positioning, not a verbatim quote): shipping and governing large numbers of enterprise agents across many customers/verticals; reliability and predictable behavior of production agents integrated with enterprise systems. Challenges: scaling an agent marketplace/platform where customers deploy many agents; controlling behavior and cost at enterprise scale. Must-haves: production reliability, governance/oversight, integration with enterprise back-office systems. Nice-to-haves: visibility into per-agent cost and performance. ICP confidence: High (technical co-founder/CTO at an AI-native company shipping agents in production, in the 50-2,000 band).
Signal: 3 (ICP on record in a competitor/LLM-gateway vendor case study). Source: https://nexos.ai/customer-stories/payhawk/ and https://procurementmag.com/news/payhawks-new-agents-is-agentic-ai-transforming-fintech — title verified live on LinkedIn 2026-08-30 ("Co-founder & CTO at Payhawk"). Company size: 457 employees (Tracxn, Mar 2026); HQ London UK, R&D Sofia BG; spend-management fintech; Series B, $239M raised, $1B valuation, >$100M ARR. Pain points (his own quotes): "Teams were exploring AI independently: great for innovation, but terrible for coordination, compliance, and budget." The case study lists their named challenges as "there were never enough tokens for every team, volume, and types of AI queries" and a lack of central control. Challenges: four AI agents shipped on top of existing financial infrastructure — "we've built these AI agents on Payhawk's existing financial infrastructure" — in a regulated fintech where agent actions touch money movement. Must-haves: "complete transparency: a feature to track both company-wide AI usage and individual, granular insights"; a token/budget allocation mechanism across teams. Nice-to-haves: per-team chargeback, compliance-grade audit trail of agent decisions. ICP confidence: High — co-founder CTO actively shipping agents in production and explicitly on record about token scarcity, budget and usage observability. IMPORTANT: already buying an LLM gateway (nexos.ai), so this is a displacement or layer-above play, not a greenfield sale.
Brendan Fortuner — Head of Engineering, Ambience Healthcare
Signal: 4 (ICP building/shipping agents in production). Source: https://www.cognitiverevolution.ai/cracking-the-medical-code-why-cleveland-clinic-doctors-love-their-ambience-healthcare-ai-scribe/ ; role/company confirmed via https://theorg.com/org/ambience-healthcare/org-chart/brendan-fortuner and LinkedIn. Company size: ~200-400 (Series C healthtech AI, ~$1B+ valuation; actively hiring Staff SWEs for multi-tenant infra handling millions of patient encounters). Pain points: scaling agentic capabilities (agents that call patients post-visit, synthesize responses, write structured notes into Epic) into production across millions of encounters; reliability of voice-modality agents in a high-stakes clinical setting. Challenges: multi-tenant infra at scale, production reliability where errors are costly, expanding from ambient scribing to autonomous agent workflows. Must-haves: production reliability guarantees, visibility into agent behavior at scale. Nice-to-haves: per-run cost/latency visibility. ICP confidence: High — Head of Engineering (decision-maker) at a Series C AI-native company scaling agents into production.
Brian Barbosa — Director of Software Engineering, Maven AGI
Signal: 4 (Director of Engineering at a company actively shipping AI agents; found via LinkedIn people search + web verification).
Source: https://www.linkedin.com/in/brian-barbosa-0263235/ ; Maven AGI Series B coverage (PRNewswire/Lux Capital, Jun 2025).
Company: Maven AGI — enterprise AI agent platform for customer experience/support; autonomous agents resolving up to ~93% of inquiries; clients incl. Tripadvisor, ClickUp, Rho.
Company size: SIZE CAVEAT — Crunchbase bucket shows 11-50 (likely stale); Series B ($50M Jun 2025, $78M total) + 9 departments + named enterprise clients strongly imply ~50-100 employees now. Treated as borderline-but-qualifying.
Stage: Series B.
Pain points (INFERRED from role + company, NOT verbatim): reliability of support agents in production; cost per resolution/run at enterprise volume; observability into agent behavior/spend.
Challenges (inferred): scaling multi-agent CX deployments while controlling LLM cost and maintaining resolution quality.
Must-haves (inferred): production reliability, cost-per-run visibility, guardrails.
Nice-to-haves (inferred): token/context optimization, eval tooling.
ICP confidence: Medium (Director of Eng at agent-shipping Series B company; headcount at/near the 50-employee floor — flagged).
Brian Moseley — Co-founder & CTO, Sixfold
Signal: 4 (ICP technical co-founder building/shipping agents in production). Source: https://www.theinsurer.com/ti/news/exclusive-sixfold-launches-ai-underwriting-agent-with-straight-through-quote-and-2026-06-12/ ; https://natlawreview.com/press-releases/sixfold-raises-30-million-series-b-build-ai-underwriter ; https://tracxn.com/d/companies/sixfold ; https://www.sixfold.ai/about-us . Company size: 70 employees (May 2026), "50+ engineers, product builders, underwriting experts"; Series B ($30M, Jan 2026). AI-native platform; launched (June 2026) an agentic "AI Underwriter" with straight-through quote-and-bind for P&C insurers — an agentic colleague doing analytical work and holding institutional memory. In production with major carriers (Zurich, Generali GC&C, Guardian, AXIS, New York Life, Skyward Specialty; customers ~$270B gross written premium). Moseley is co-founder & CTO. Pain points: reliability/trust of agents making high-stakes underwriting decisions in a regulated domain; consistency + auditability of agent reasoning; institutional-memory/context management. Challenges: production-grade accuracy and control across large carriers; visibility into agent decisions. Must-haves: reliability, auditability/traceability, behavior control. Nice-to-haves: per-decision cost visibility, model routing. ICP confidence: High (technical co-founder/CTO at a 70-person Series B AI-native company shipping an underwriting agent in production at large enterprises; reliability+governance is the core buying pain). Profile URL left blank (personal LinkedIn not verified this run).
Brian Ngo — Head of Agent Engineering, APAC, Sierra
Signal: 4 (ICP technical leader at agent-shipping company). Source: https://www.linkedin.com/company/sierra/people/?keywords=head%20engineering. Company size: 201–500 employees (per Sierra LinkedIn). Company context: Sierra builds and ships conversational AI agents in production for enterprise CX at scale. Pain points (inferred from role/ICP, not a direct quote): running agent engineering across a fast-scaling APAC deployment — production reliability, per-run cost, and consistency of agents across many customer environments. Challenges: scaling agent engineering org and infra from a few to many production agents across regions. Must-haves: production reliability and observability into agent behavior/cost per run. Nice-to-haves: cross-region standardization of agent tooling. ICP confidence: High — Head-level technical AI leader owning agent engineering at a flagship agent company in target size band.
Signal: 1 (CTO publicly discussing moving agentic AI from pilot to production at scale). Source: https://telecomreseller.com/2026/03/04/dialpad-moving-agentic-ai-from-pilot-to-production-podcast/ ; https://www.linkedin.com/in/brianpetersonbayarea. Company size: 1,478 employees (Dialpad, communications/CX SaaS; agentic AI platform w/ no-code agent building, governance, ROI validation). Pain points: operationalizing AI pilots into production; deploying agents safely; measuring agent impact; scaling agents across the org. Verbatim: "The real question now isn't whether AI works—it's how you deploy it safely, measure its impact, and scale it across the organization." Challenges: pilot-to-production gap; reliability, measurement & governance at scale. Must-haves: safe deployment, impact measurement/observability, governance. Nice-to-haves: ROI validation, cost control. ICP confidence: Medium — technical co-founder/CTO; company ~1,478 employees fits 50-2000 but is late-stage/large; actively shipping agentic AI in production.
Signal: 1 (ICP author writing about agent costs). Source: LinkedIn post (Aug 2026, past-month) surfaced via content search 'our AI agents in production cost scaling' — https://www.linkedin.com/in/brianreale/ . Company size: ~150-250 employees (ProcessMaker, workflow/BPA platform now shipping agentic automation). Pain points (from his post): enterprises struggle to understand the TRUE cost of implementing AI agents — 'not just tokens but also the way workflows get adjusted'; hidden costs nobody budgets for. Challenges: quantifying/forecasting agent operating cost as workflows change; scaling agentic BPA to production. Must-haves: visibility into real per-agent/per-workflow cost beyond raw tokens. Nice-to-haves: cost benchmarking across workflows. ICP confidence: Medium-High (C-suite technical founder, company in 50-2000 range, actively shipping agentic automation, explicit agent-cost pain in an authored post).
Signal: 3 (ICP engaging with competitor content — named in a Humanloop customer case study).
Source: https://humanloop.com/case-studies/filevine
Company size: 917-942 employees (Revelio Labs / Tracxn, 2026). Legal case-management SaaS shipping AI features across a growing product line.
Pain points (direct quote): "I'm convinced the vast majority of companies leveraging generative AI today are operating in the dark." Also: "Before Humanloop, our prompt management and evaluation process was extremely manual."
Challenges: scaling prompt ops and evals across an expanding AI product line; producing evidence of AI quality that leadership can actually read; manual process that does not survive going from one AI feature to many.
Must-haves: a system of record for prompts and agent behaviour; evaluation visibility that rolls up to leadership.
Nice-to-haves: automated regression detection as models and prompts change.
ICP confidence: Medium — headcount is comfortably in band, title is VP level, and the "operating in the dark" quote is the single cleanest articulation of Alpha's visibility thesis found this run. Downgraded from High for two honest reasons: (1) Filevine is Series E, outside the stated Series A-C window; (2) VP of Data Science is adjacent to, not identical with, the VP Eng / Head of AI persona.
LinkedIn URL: not found in any source actually read this run — deliberately left blank rather than guessed.
Bridgette Perrier — Head of Software Engineering, Cognition (Devin)
Signal: 4 — Heads the software engineering organization at Cognition, maker of Devin, the autonomous AI software engineer in production.
Source: https://websets.exa.ai/websets/directory/cognition-executives ; Cognition / Devin (LinkedIn /in slug not confirmed — left blank)
Company size: ~200–407 (Series C, ~$10.2B val, Founders Fund Sept 2025)
Pain points: shipping reliable software while scaling an autonomous coding agent (Devin) in production; agent reliability/quality at scale
Challenges: correctness & reliability of autonomous coding agents; eval at scale
Must-haves: agent reliability/eval, observability into agent behavior
Nice-to-haves: cost controls for large agent fleets
ICP confidence: High — Head of Engineering, Series C, autonomous agent in production
Brij Mohan Singh — Head of AI (Data Science), The Modern Data Company
Signal: 1 (ICP detailing where agent tokens go / hidden AI costs). Source: LinkedIn org posts "AI Costs: The Hidden Data Cleanup Overhead" (The Modern Data Company); moderndata101.com blogs "Building Self-Evolving Knowledge Graphs Using Agentic AI" and "The Power Combo of AI Agents and the Modular Data Stack" (author Brij Mohan Singh). Company size: ~250 (Series B; DataOS data operating system; building agentic data systems; Indore/US). Pain points: hidden token/compute costs inside agentic loops; data-cleanup overhead driving AI cost; lack of clarity on where tokens actually go. Challenges: cost visibility across agentic data pipelines; making agents cost-efficient at data scale. Must-haves: token/cost attribution for agentic workflows. Nice-to-haves: guardrails on agent loops. ICP confidence: Medium — Head of AI/Data Science at a Series B data-infra SaaS building agentic systems, publicly engaging the cost pain. Profile URL not verified to a clean /in/ slug.
Brock Imel — Head of Evaluation & Alignment / QA (Applied AI), Writer
Signal: 4 (leads Evaluation & Alignment and QA / Applied AI at ICP agent company; via LinkedIn People directory). Source: https://www.linkedin.com/in/baimel (PhD Linguist). Company: Writer, 201-500, Series C, enterprise agentic AI. Pain points (inferred, core to eval/alignment function): agent reliability in production; measuring/ensuring output quality; catching regressions/drift. Challenges: building evals that guarantee reliability at enterprise scale; balancing quality vs cost/latency. Must-haves: robust eval/observability tooling; reliability guardrails. Nice-to-haves: automated eval + cost-aware model selection. ICP confidence: Medium (heads a core AI function; exact Director-level title not explicitly stated in headline). NOTE: pain points inferred from role/company, not verbatim quotes.
Signal: Bucket 4 (ICP technical founder at an agent-native company shipping agents in production).
Source: https://interface.ai/blog/interfaceai-unveils-industry-first-agentic-bankgpt-platform/ ; https://theorg.com/org/interface-ai
Company size: ~204 emp (as of Jun 2026), ~$25M ARR; agentic 'BankGPT' platform for banks & credit unions (voice/chat/mobile agents). Well-funded ('most valuable agentic AI company in banking'); ~Series B/C stage.
Pain points (INFERRED from verified role+company, not verbatim): running voice/chat banking agents in production across many financial institutions; regulated reliability & auditability; cost control per interaction; scaling a multi-agent BankGPT platform.
Challenges: reliability + compliance in regulated banking; agent cost per run at scale; model/vendor neutrality.
Must-haves: production reliability, auditability, cost visibility & control per agent run.
Nice-to-haves: model routing, cost circuit-breaker.
ICP confidence: High — Co-Founder & CTO, ~204 emp, agent-native banking, right size/stage.
Bruno Segalla Peres — Director of Software Engineering, Gupy
Signal: 3 (also 1) — LinkedIn post "Autonomous Agents Risk Confidently Wrong Decisions" engaging with enterprise-AI observability/governance tooling (Atlan, Monte Carlo); also posts about making data semantically queryable before wiring it to agent frameworks and about agentic-era infra challenges.
Source: https://www.linkedin.com/posts/brunotsperes_enterprise-ai-confidence-with-atlan-monte-activity-7453054273812086787-zN4E
Company size: ~700–1,200 employees (Brazil's largest HR-tech SaaS, Series C, ~$103M raised; deploying internal AI agents incl. "Guta").
Pain points: autonomous agents making confidently wrong decisions in production; data/semantic readiness before connecting to agent frameworks; agentic-era infrastructure complexity.
Challenges: making agent output trustworthy and safe to ship at enterprise scale.
Must-haves: reliability + governance + visibility into agent behavior.
Nice-to-haves: semantic grounding to reduce hallucination.
ICP confidence: High — Director of Engineering at a 700–1,200-person SaaS actively deploying agents; role, size, and agent-building all match.
Signal: 1 (ICP speaking about agent cost/model routing in production). Source: Chroma interview "Bryan on building AI at Zapier" (https://www.trychroma.com/interviews/bryan-on-zapier) + Zapier agent/model resources. Company size: ~700-900 (mature SaaS, 8,000+ integrations). Pain points: managing agent cost at scale, routes different work to different models and measures the difference rather than betting on one model; reliability/scale of Zapier Agents across 30,000+ actions/MCP. Challenges: cost-scaling of autonomous multi-step agents in production; model selection tradeoffs. Must-haves: cost visibility + model routing to control agent spend. Nice-to-haves: per-agent / per-run cost attribution. ICP confidence: Medium — large mature SaaS (not Series A–C) but shipping autonomous agents in production with an explicit cost-control signal; headcount well within 50–2,000.
Cai GoGwilt — Co-founder & CTO / Chief Architect, Ironclad
Signal: 1 (ICP speaking about building reliable agents in production). Source: "Building Reliable AI Agents with Cai GoGwilt, CTO of Ironclad" (Humanloop, https://humanloop.com/blog/building-agents-with-ironclad) + Ironclad agentic launch (https://ironcladapp.com/resources/articles/ai-agentic-launch). Company size: ~350–400 (~$200M ARR), late-stage. Pain points: building reliable, trustworthy contract agents (Intake, Redline, Conversational Search, drafting/review) in production; monitoring and fine-tuning agents over time so legal teams trust them. Challenges: agent trust/reliability for high-stakes legal workflows. Must-haves: reliable multi-agent orchestration + ongoing monitoring/eval. Nice-to-haves: cost/observability across agents. ICP confidence: Medium-High — co-founder/CTO shipping 5+ specialized agents in production; headcount within range; later stage (Series E) is the only caveat.
Signal: 4 (ICP technical AI leader at an agent-shipping company; surfaced via LinkedIn people/company search this run, no specific post authored). Source: LinkedIn people search 'CodaMetrix director machine learning' + profile in/cansusen. Company size: ~150-200 (CodaMetrix, Series B, Boston; autonomous medical-coding AI agents). Role headline: Director of Applied ML | Healthcare AI | Foundation Models & Data Strategy | PhD in ML. Pain points (role/domain-inferred, not directly expressed this run): cost-effective productionization of foundation models; evaluation/reliability of applied ML in regulated coding; data strategy to control token/context waste. Challenges: scaling applied ML + foundation models reliably and affordably. Must-haves: reliability + cost per run under regulatory audit. Nice-to-haves: eval + observability tooling. ICP confidence: Medium (Director-level applied ML at qualifying agent-native company; pain inferred this run).
Carina Negreanu — CTO (formerly VP of AI; ex-Microsoft Principal Research Manager), Robin AI
Signal: 4 (ICP building agents) + Signal 3 (surfaced alongside legal-AI competitor comparison, incl. Luminance). Source: https://uk.linkedin.com/in/carina-suzana-negreanu ; https://www.artificiallawyer.com/2025/03/18/robin-ais-new-cto-there-will-be-two-types-of-lawyer/ ; https://www.robinai.com/luminance
Company size: ~54 employees (May 2026, down from 100+ after a 2025 redundancy round amid slower growth); ~$71.7M raised across 7 rounds ($26M round Jan 2024; ~$50m in 2023–24).
Company context: Robin AI builds AI agents for contract review/analysis and legal workflow automation (Copilot) for Fortune 500 / enterprise legal teams. Carina moved from VP of AI to CTO in late 2025; ex-Microsoft Principal Research Manager, PhD (Cambridge).
Pain points (inferred from role/domain): making legal AI agents accurate and reliable enough for enterprise contract review; aligning R&D/ML with product; doing so cost-efficiently while the company runs leaner post-layoffs.
Challenges: reliability/accuracy of agentic contract review; efficiency pressure after workforce reduction; controlling agent behavior for high-stakes legal work.
Must-haves: reliability and control of legal agents; cost efficiency given leaner runway.
Nice-to-haves: observability into agent decisions; agents that improve with feedback.
ICP confidence: Medium — clear technical decision-maker (CTO) at a 50+ company actively building AI agents; company recently shrank and growth disappointed (sizing/traction flag), hence Medium not High.
Signal: 4 (ICP technical leader at a qualifying company actively shipping AI agents). NOTE: This run's LinkedIn/Chrome browsing was unavailable (extension not connected), so prospecting pivoted to web research + primary-source verification, consistent with prior runs. Source: Glia leadership pages + company profiles (glia.com/news; cbinsights/salemove; pitchbook). Company: Glia (formerly SaleMove), NYC, founded 2012 — AI-powered contact center / "Banking AI Workforce" for financial institutions; shipping Glia Cortex agentic AI. Company size: ~445–469 employees (mid-2026, multiple sources) — solidly in ICP band. $152M raised across 7 rounds. Pain points (inferred from role/domain, not a personal quote): production reliability & compliance for customer-facing banking agents; agent cost/observability across high-volume contact-center interactions; scaling from single assistant to a fleet of "AI workforce" agents. Challenges: regulated-industry accuracy/hallucination control; visibility into per-interaction agent cost. Must-haves: reliability guarantees, auditability, cost control per run for financial-services agents. Nice-to-haves: cross-agent orchestration/observability tooling. ICP confidence: High (confirmed technical C-suite at a mid-size, agent-shipping fintech; headcount confirmed in band). Pain framing is inferred from documented product/domain, not a fabricated quote.
Signal: 1 (technical leader speaking publicly about AI cost/performance & production latency). Source: https://www.aiengineeringpodcast.com/ (Modulate CTO episode on low-latency, high-accuracy Voice AI) ; https://theorg.com/org/modulate/org-chart/carter-huffman ; https://www.crunchbase.com/person/carter-huffman . Company size: 56 employees, Series A (founded 2017; voice AI for fraud detection & online safety / ToxMod). Technical co-founder/CTO leads research on optimized AI (cites ~1000x cost/performance improvements over leading models). Pain points: cost & performance optimization of AI inference in production; real-time low-latency voice at scale; reliability. Challenges: delivering large cost/perf gains vs frontier models; running voice AI reliably in real time. Must-haves: efficient inference + cost control per interaction. Nice-to-haves: observability into model/agent performance. ICP confidence: Medium — voice-AI/ML company (not classic multi-agent), but technical co-founder at in-range AI-native company shipping AI in production, vocal on cost/perf.
Cassandra (Cassie) Shum — VP of Ecosystem, Product Engineering, RelationalAI
Signal: 1 (ICP speaking about building production-ready agentic AI — QCon AI Boston 2026 speaker, talk: "Retrieval to Reasoning: building production-ready agentic AI systems with knowledge graphs"; leads teams deploying production AI systems).
Source: https://boston.qcon.ai/speakers/cassieshum
Company size: 182 employees (RelationalAI; Series B, ~$144M raised incl. $22.5M Dec 2025 earmarked for agentic capabilities; Berkeley). Within 50–2,000.
Pain points: making agentic AI systems production-ready, robust and explainable; moving beyond RAG to reliable multi-step reasoning; agent reliability grounded in enterprise data.
Challenges: explainability of agent decisions; reliability of multi-step reasoning agents over enterprise knowledge.
Must-haves: production-grade reliability, explainability, grounding/guardrails for agents.
Nice-to-haves: cost/observability visibility across agent reasoning steps.
ICP confidence: Medium-High — VP-level product-engineering leader at a 182-person Series B AI-native company actively building agentic decision intelligence.
NOTE: Discovered via web/conference fallback — Claude-in-Chrome/LinkedIn browsing was not connected during this scheduled run.
Chai Asawa — Head of Engineering, Clinical Decision Support, Abridge
Signal: 1 (ICP discussing agent cost/reliability/routing publicly on a podcast). Source: https://www.latent.space/p/abridge (Latent Space: "AI-Native Healthcare"). Company size: ~450–635 employees (mid-2026), fast-growing generative-AI clinical documentation platform (1M+ clinical encounters/week across 150+ health systems, agentic workflows e.g. prior auth) — in range. Pain points: at 100M+ conversations, focus on "reliability, cost, post-training, model routing, and infrastructure optimization"; runs a layered eval stack (LLM judges, in-house clinicians, third-party + specialty-specific evals, progressive rollout). Challenges: cost and reliability of agentic clinical workflows at massive scale. Must-haves: model routing to control cost, rigorous eval/reliability. Nice-to-haves: unified observability across model routing decisions. ICP confidence: Medium-High (clear technical eng leader at in-range agent-shipping company; exact LinkedIn/title scope slightly lowers it).
Chaitanya Asawa — Head of Engineering, Clinical Decision Support, Abridge
Signal: 4 (ADAPTED — LinkedIn was unavailable this run; sourced from public conference speaker listing, not a LinkedIn post).
Source: https://www.ai.engineer/worldsfair/schedule — AI Engineer World's Fair 2026, "AI in Healthcare" track, Day 4 (Thu 7/2).
Talk (verbatim): "From Ambient Documentation to Clinical Intelligence".
Company size: ~635 employees (Crustdata, June 2026; Tracxn 605 as of 31 May 2026; Revelio 539 as of Dec 2025). Late-stage private (Series E) — headcount fits the 50–2,000 band but stage is later than the Series A–C ICP window.
Pain points: INFERRED FROM TALK TITLE AND TRACK, NOT A VERBATIM QUOTE — the move from a single ambient-scribe agent to a broader "clinical intelligence" agent surface is exactly the 1 -> 5+ agents scaling wall.
Challenges: clinical-grade reliability and accuracy in a regulated domain; expanding agent surface area without expanding failure surface.
Must-haves: reliability guarantees and auditability per agent run.
Nice-to-haves: per-agent and per-run cost visibility as the agent count grows.
ICP confidence: High — Director/Head-level engineering leader owning a named agent product line at an AI-native company of qualifying headcount that demonstrably ships agents in production.
NO LinkedIn URL captured — do not fabricate one; needs enrichment before outreach.
Chaitanya Gharpure — Co-founder & CTO, Sully.ai
Signal: 1/4 (adapted — LinkedIn auth was unavailable this run; identified via web research + verification of company scale and agent-building). Source: https://www.sully.ai/ ; https://tracxn.com/d/companies/sully/ ; https://markets.financialcontent.com/pawtuckettimes/article/marketersmedia-2025-10-20-sully-ai-launches-receptionist-and-voice-agents. Company size: ~60–70 employees (multiple 2026 trackers). Sully.ai ships a multi-agent "AI medical employees" system (scribe, receptionist, voice agents) into health systems, integrated with EHRs — clear 5+ agents in production, Series A (~$35M raised, incl. Jan 2025 round). Pain points (inferred from production profile, not a verbatim quote): running interconnected voice + scribe agents across the patient journey at clinical-grade reliability; cost of long voice/transcription LLM calls at volume. Challenges: keeping multi-agent handoffs reliable and auditable in a regulated (HIPAA) setting; visibility into per-encounter agent cost. Must-haves: reliability/guardrails for clinical safety; per-run cost visibility as call volume scales. Nice-to-haves: cross-agent orchestration observability. ICP confidence: High — CTO title, agent-native company, size in range. LinkedIn URL not captured this run (verify before any outreach).
Chaithanya Yambari — Co-founder & CTO, Zluri
Signal: 4 (ICP writing about agents — first-person authored post on non-human identity / AI agent sprawl). Source: https://www.zluri.com/blog/nhi-governance
Company size: ~125-271 employees (Tracxn ~271; PitchBook ~145) — in range. Series B ($20M, 2023). SaaS management / identity governance, AI-native.
Pain points: AI agent sprawl with no owner or inventory — first-person quote: "AI agent deployment creates credentials to access production APIs... None of these appear in Workday or BambooHR... AI agents deployed by product teams have broad API access with no documented owner." Also cites Copilot Studio users creating 1M+ agents, each minting credentials that authenticate without human involvement.
Challenges: Discovering and classifying every agent/non-human identity across an org; agents proliferating faster than governance can track; no single system of record for what agents exist or what they can touch.
Must-haves: Real-time inventory of every agent running in the environment; ownership attribution per agent; access/permission visibility per agent.
Nice-to-haves: Policy enforcement pre-execution; audit trail per agent action; cost attribution per agent owner.
ICP confidence: High — verified co-founder/CTO title, headcount in range, agent-governance product shipping, and the pain signal is a first-person quote from the target himself.
Chang Liu — Senior Director of Engineering, Head of NLU & MLOps, Moveworks
Signal: Bucket 4 (ICP building agents) — LinkedIn people-search proxy. Source: https://www.linkedin.com/in/changliucmu (Mountain View, CA). Company size: ~500 (Moveworks, enterprise AI assistant/agents). NOTE: Moveworks acquired by ServiceNow in 2025; now operates as an agent unit inside a >2000-emp parent — but the brain already lists Moveworks as an ICP company. Pain points (INFERRED from role): MLOps/agent reliability at enterprise scale, LLM inference cost + latency across an agent fleet, per-run observability. Challenges: reliability + cost control while scaling agents. Must-haves: agent observability, cost-per-run visibility. Nice-to-haves: eval tooling. ICP confidence: Medium — Director+ AI/eng leader; parent (ServiceNow) exceeds 2000 emp, so size is borderline. Caveat: found via title search; pains inferred (no fabricated quote).
Signal: 4 (ICP at a company exposing an agent/MCP surface in production). Source: https://www.shipwell.com/about/leadership, https://www.shipwell.com/mcp-server, https://www.shipwell.com/solutions/document-ai — title verified live on LinkedIn 2026-08-30 ("VP of Engineering at Shipwell", Austin TX). Company size: ~91 employees (PitchBook/LeadIQ 2026) — near the 50-employee floor. Austin TX; multimodal TMS / logistics execution; Series B $35M (2019) plus later rounds; named a Visionary in the 2026 Gartner Magic Quadrant for TMS. Pain points: Shipwell ships a customer-facing MCP Server — "securely connect AI to your transportation systems, partners, data, and workflows" — plus Document AI for freight document processing and AI pricing intelligence. Exposing an MCP surface to customer-side agents means unbounded, third-party-driven tool-call volume that Shipwell pays for but does not control: a direct and unusual cost-control and reliability exposure. Challenges: a small engineering org (he leads roughly 22 per org-chart data) carrying a broad AI surface area; freight document AI is high-volume and token-heavy. Must-haves: rate limiting and cost attribution on an MCP surface consumed by external agents; per-customer spend visibility. Nice-to-haves: reliability telemetry on document AI throughput. ICP confidence: Medium — VP Eng is in-ICP and the vertical and MCP exposure are a genuinely interesting wedge; caveats are that headcount sits close to the 50 floor and the agent evidence is platform/MCP rather than a named autonomous agent, so qualify agent count on the call.
Charles Giardina — Vice President of Engineering, Levelpath
Signal: 4 (newly appointed VP Eng at a company shipping multiple production procurement agents). Source: https://www.businesswire.com/news/home/20260212834265/en/Levelpaths-Breakout-Year-Sets-Stage-for-Charles-Giardina-Appointment-as-VP-of-Engineering ; product: https://www.levelpath.com/blog/delightful-procurement-at-scale-levelpath-has-agents-for-that
Company size: ~100 employees (PitchBook); other aggregators 26-156 depending on collection date. Series B, $55M (2025). AI-native procurement.
Profile URL: not found — do not fabricate. Prior Airbyte-era handle (linkedin.com/in/cgardens) surfaced but could not be confirmed as current; left blank deliberately.
Pain points: Context and data foundation for agents. On-record: "AI is only as powerful as the data behind it. Levelpath's unified platform brings all of procurement into one system with a complete, clean data foundation, giving AI the context it needs to drive real outcomes." Weaker/cost-adjacent rather than an explicit cost complaint — flagged.
Challenges: Newly in seat (Feb 2026) scaling an engineering org that runs at least three distinct agent types (sourcing, contract discovery, risk) on the Hyperbridge reasoning engine; ex-Airbyte VP Eng so likely carries strong opinions on infra cost and observability.
Must-haves: Clean unified context for agents to reason over; ability to scale agent count without proportional engineering headcount.
Nice-to-haves: Per-agent cost and latency visibility; standardised eval/regression tooling across agent types.
ICP confidence: Medium — VP Engineering title and Series B stage are clean ICP fits and agents are demonstrably in production, but the pain evidence is thin and the LinkedIn profile is unverified. Worth a light-touch enrichment pass before outreach.
Charles Hearn — Co-founder & CTO, Alloy (alloy.com)
Signal: 1 (ICP writing directly about agent context, reliability, and control). Source (bylined by him, Jan 2026): https://www.alloy.com/blog/agentic-ai-alloy-native-ai-agents ; also interviewed in CIO on navigating AI agents: https://www.cio.com/article/4123497/what-cios-in-finance-do-to-navigate-ai-agents.html
Company size: ~400-412 employees (2026 aggregators) — in range. Series D ($52M, 2025) — later stage than the A-C preference, flagged. Identity/risk decisioning infrastructure for fintechs and banks.
Pain points: Context starvation and hallucination in production agents. First-person: "AI agents live and die by context. Without it, they guess, and guessing is not acceptable in risk management." And: "I tested a leading KYB agent with a slightly incorrect address... the agent joyfully approved it... LLMs are still too agreeable... they will confidently hallucinate." Per CIO: "the core of decision-making for high-value operations needs to be predictable and reproducible."
Challenges: Making agent decisions predictable and reproducible in a regulated risk/fraud environment; feeding agents the right context without ballooning token spend; preventing agreeable-but-wrong agent approvals.
Must-haves: Deterministic/reproducible agent decisioning; context correctness and provenance; a single source of truth agents act against.
Nice-to-haves: Cost-per-decision instrumentation; automated detection of agent over-approval drift.
ICP confidence: Medium-High — strongest first-person pain language of this run and verified CTO/co-founder at in-range headcount; downgraded from High only because the funding stage (Series D) sits outside the stated A-C band.
Charles Yeh — Co-Founder & Chief Technology Officer, Persona (withpersona)
Signal: 4 (technical co-founder at a company shipping agents AND publicly reasoning about agent behaviour). Source: https://www.linkedin.com/in/charlesyeh ; https://theorg.com/org/withpersona/org-chart/charles-yeh ; https://withpersona.com/releases/summer26/ (Case Review Agents) ; https://withpersona.com/blog/series-d/ ; https://www.forbes.com/sites/rashishrivastava/2025/04/30/ai-is-making-the-internets-bot-problem-worse-this-2-billion-startup-is-on-the-front-lines/. Company size: ~620 employees. Founded 2018, Series D ($200M, $2B valuation), FedRAMP Moderate authorised. Agents in production: Case Review Agents trained on a team's past decisions that review incoming cases and deliver recommendations, plus agent/bot-identity detection work. Pain points (inferred from public product + co-founder commentary, not verbatim): CEO Rick Song's public thesis is that within 2-3 years agentic behaviour will be indistinguishable from human behaviour online and that agents will need to be registered and identified - i.e. this leadership team already thinks in terms of agent identity, attribution and accountability, which is adjacent to Alpha's control/observability pitch. Challenges: agents making recommendations on regulated identity/fraud decisions inside a FedRAMP-authorised environment = auditability and determinism are non-negotiable; agents trained on a customer's past decisions must not drift. Must-haves: auditable, attributable agent actions; reliability/consistency on decisions; scoped access. Nice-to-haves: cost visibility per case reviewed. ICP confidence: Medium-High - co-founder/CTO, ~620 employees (in band), agents shipped in production; the cost signal is weaker than the control/auditability signal, so lead with control. NOTE: not yet contacted; research only.
Cheng Li — Director of Fundamental Research, Machine Learning, CodaMetrix
Signal: 4 (ICP technical AI leader at an agent-shipping company; surfaced via LinkedIn people/company search this run — no specific post authored). Source: https://www.linkedin.com/search/results/people/?keywords=CodaMetrix%20director%20machine%20learning ; profile https://www.linkedin.com/in/cheng-li-1a444191/. Company size: ~150-200 (CodaMetrix, Series B, Boston; autonomous medical-coding AI agents translating clinical notes into billing codes at hospital-system scale). Pain points (role/domain-inferred, not directly expressed this run): research-to-production model cost/compute efficiency; reliability & hallucination control in a regulated coding workflow; per-run/per-encounter cost visibility across millions of encounters. Challenges: moving fundamental ML research into cost-efficient, reliable production agents. Must-haves: reliable production behavior + cost control at claims volume. Nice-to-haves: unified observability of cost + quality per run. ICP confidence: Medium (Director-level ML at qualifying agent-native company; pain not directly expressed this run).
Chia Chen (Emma) Nien — Senior Director, Data Management & Data Science (Clinical Decisioning), Cohere Health
Signal: 4 (ICP technical AI leader at an agent-shipping company; surfaced via LinkedIn people/company search this run, no specific post authored). Source: LinkedIn people search 'Cohere Health machine learning director' + profile in/chia-chen-emma-nien-35a7a665. Company size: ~500-800 (Cohere Health, Series C, Boston). Profile note: Sr Director of Data Science, Insights & Clinical Decisioning since Apr 2021, building 20+ predictive models + data governance. Pain points (role/domain-inferred, not directly expressed this run): reliability/governance of 20+ production models feeding clinical decisioning; cost per run at payer scale; scaling from predictive models to agentic decisioning. Challenges: governing many models reliably and affordably. Must-haves: reliability + governance + cost visibility. Nice-to-haves: unified cost/quality observability. ICP confidence: Medium (Senior Director data science/clinical decisioning at qualifying agent-native company; pain inferred this run).
Chinmay Barve — VP of Engineering, Nooks
Signal: 4 (ICP building/shipping agents in production). Source: https://www.geekwire.com/2026/the-rise-of-vertical-ai-agents-and-the-startups-racing-to-build-them/ and https://www.geekwire.com/2026/san-francisco-ai-startup-nooks-makes-engineering-push-in-seattle/ | Company size: ~400 employees (2026), Series B AI-native sales platform. Prior: Mixpanel, Twitter, Yahoo Search, EA. Publicly spoke (Feb 2026 Nooks Seattle event) on domain-specific AI agents deployed in production. Pain points: operating many vertical agents in production; reliability and quality of agent outputs at scale; standing up an engineering org/hub to support agent development. Challenges: moving agents from demo to reliable production; team scaling. Must-haves: production reliability and observability for agents. Nice-to-haves: cost/performance visibility per agent run. ICP confidence: High (VP Engineering / decision-maker at a 50–2,000-employee AI-native company shipping agents in production).
Chris Aberger — VP, Applied AI / Agentic Workflows, Alation
Signal: 3/4 (ICP VP-level technical AI leader at a mid-size company shipping AI agents; ex-founder of an agent startup). Source: https://www.geekwire.com/2025/alation-acquires-numbers-station-an-ai-data-analysis-startup-backed-by-madrona/ ; https://www.alation.com/news-and-press/alation-announces-agentic-platform-reinventing-the-data-catalog-for-ai-era/ ; https://www.techtarget.com/searchdatamanagement/news/366619841/Alation-unveils-AI-agents-plus-SDK-for-agentic-development ; https://www.alation.com/podcast/episodes/ai-builders-metadata-chris-aberger/ . Company size: Alation ~800-900 employees (well-known enterprise data-intelligence vendor). Stage: later-stage/private (past Series C — NOTE: outside strict Series A-C guideline, but 50-2,000 emp and actively shipping agents). Independent. Actively shipping agents: Alation launched an Agentic Data Intelligence Platform, shipped AI agents + an SDK for agentic development, and acquired Numbers Station (May 2025) specifically to power agentic analytics workflows with enterprise-grade governance. Role: Chris Aberger — VP at Alation leading agentic workflows / AI-native analytics; former CEO & co-founder of Numbers Station AI (Stanford Hazy Research; pioneered LLM agents for structured data) — deeply technical AI leader. Pain points (from his public talks/positioning, paraphrased, NOT verbatim): teams stuck moving agentic workflows from experimentation to production; agents need trusted metadata/context + governance to be reliable; accuracy/reliability of data agents at enterprise scale. Challenges: production-grade reliability & governance for data agents; grounding agents in trustworthy metadata. Must-haves: governance, reliability, context/metadata grounding. Nice-to-haves: faster build-to-deploy, cost/observability. ICP confidence: Medium-High (named VP-level technical AI leader, 50-2,000 emp, company actively shipping agents; caveat: company is later-stage than the A-C guideline).
Chris Gervais — Chief Technology Officer, CodaMetrix
Signal: 1 + 4 (ICP writing/leading on agent reliability & cost; shipping autonomous agents in production). Source: https://www.businesswire.com/news/home/20240815656258/en/CodaMetrixs-AI-Platform-Now-Available-in-Epic-Toolbox ; https://fcventures.com/codametrix-announces-40m-series-b/. Company size: ~139-200 employees (Tracxn/PitchBook/Forbes, 2026); Series B $40M (Transformation Capital); ~$109M total raised; Boston; 500+ hospitals across 27 states; available in Epic Toolbox. Pain points (INFERRED): autonomous medical-coding agents must be accurate and auditable at 500+ hospital scale; RCM revenue impact of agent errors; explainability/defensibility to external auditors and payers. Challenges: reliability + auditability of autonomous coding agents across many health systems; scaling multi-agent RCM inside Epic; per-claim cost/behavior control. Must-haves: reliable, externally-defensible agent output with audit trail; per-run cost + behavior visibility as coding volume scales. Nice-to-haves: LLM cost optimization per claim. ICP confidence: High (CTO owning software eng, data/analytics, and AI/ML; Series B; ~140-200 emp in-band; autonomous agents in production).
Chris Lu — Co-founder & CTO, Copy.ai
Signal: 4 (ICP technical co-founder at a company shipping GTM AI agents; found via web research — LinkedIn/Chrome not connected this run). Source: https://theorg.com/org/copy-ai/org-chart/chris-lu ; https://sacra.com/research/chris-lu-copy-ai-generative-ai-enterprise/ ; https://www.fullcast.com/content/fullcast-announces-the-acquisition-of-copy-ai/ . Company size: ~199 employees (Jun 2026); Copy.ai is a generative-AI GTM platform shipping AI agents/workflows for go-to-market teams. Pain points: reliability of multi-step GTM agent workflows; cost/token efficiency at content+workflow scale; making agent output trustworthy/controllable for enterprise GTM. Challenges: reliability and cost control across many agent runs; governance for enterprise deployment. Must-haves: reliability, cost efficiency, control. Nice-to-haves: per-run cost visibility, context efficiency. ICP confidence: Medium (co-founder/CTO at a ~199-emp company shipping GTM agents, in the size band — CAVEAT: Copy.ai was acquired by Fullcast in Oct 2025, so it is no longer an independent Series A–C company; verify Chris Lu's current status/role before outreach; do not fabricate).
Signal: 4 (ICP technical co-founder building/shipping agentic AI in production). Source: https://www.fieldguide.io/blog/series-c-announcement ; https://www.fieldguide.io/blog/fieldguide-first-aiuc-1-certification ; https://tracxn.com/d/companies/fieldguide ; YC profile. Company size: 202 employees (May 2026); Series C — $75M round led by Goldman Sachs Growth (Feb 2026), $125M total, $700M valuation. AI-native vertical platform for audit & advisory firms; agentic AI automates the end-to-end engagement/assurance workflow. First AI platform for audit/advisory to achieve AIUC-1 certification (a trust/security/reliability standard — direct reliability & governance signal). Szymansky is co-founder & CTO (former tech leader; company built by ex-Big Four practitioners + tech leaders). Pain points: agent reliability/trustworthiness for regulated, high-stakes assurance work; auditable, verifiable agent outputs; governance/security (drove AIUC-1). Challenges: production-grade accuracy & control across many firm engagements; visibility into what agents do. Must-haves: reliability, auditability/traceability, governance/control of agents. Nice-to-haves: per-engagement cost visibility, model routing. ICP confidence: High (technical co-founder/CTO at a 202-person Series C AI-native company shipping agentic workflows in a regulated domain; reliability+governance is core buying pain). Profile URL left blank (personal LinkedIn not verified this run).
Christian Burgas — Head of Technology Platform, Parloa
Signal: 4 (ICP technical leader at validated target company Parloa; sourced via LinkedIn People search, run 2026-08-01). Source: https://www.linkedin.com/in/christian-burgas-03891161/ . Company size: ~300-500 (Parloa — Series C voice-AI-agent platform for contact centers; validated in-range). Pain points (role/company-contextual, not from an individual post this run): platform reliability & observability for voice agents in production; latency and cost-per-conversation at scale; scaling agent infrastructure from pilots to enterprise volume. Challenges: deterministic reliability across non-deterministic agent loops; multi-tenant agent-platform operations. Must-haves: per-agent cost & latency visibility; production observability. Nice-to-haves: automated eval/regression, model routing. ICP confidence: High (Head-of-platform-level technical leader at 50-2,000-emp agent-native company).
Signal: 3 (engaging with / authoring agent control, identity, and AI-gateway content adjacent to competitor space). Source: https://www.linkedin.com/in/ceposta/ ; blog.christianposta.com ("Do AI Agents Need Their Own Identity?"); O'Reilly book "AI Gateways in the Enterprise" (2025). Company size: ~286 employees (Series C, ~$175M raised, $1B val). Deeply technical field-CTO voice on agent IAM (OAuth/SPIFFE, non-human credentials), delegated authorization, context-rich policy enforcement, and putting LLMs/agents safely into production at enterprise scale. Pain points: agent identity/access-management, control and governance of agents at scale, safely running agents in production. Challenges: authN/authZ for autonomous agents across distributed systems; policy enforcement + guardrails; cost/control via AI gateways. Must-haves: agent IAM + governance, gateway-level control. Nice-to-haves: observability across agent runs. ICP confidence: Medium — 286-emp company in band and a strong technical voice on agent control/observability, BUT role is Field CTO (GTM/evangelist) and Solo.io is an agent-infra vendor, so treat as build-vs-buy / partner-adjacent rather than pure buyer.
Christophe Pierret — VP Engineering, SoundHound AI
Signal: 1 (ICP author writing about agent architecture, harness and production AI). Source: https://www.linkedin.com/in/cpierret/ — posts on building an AI harness that captures intent at the speed AI lets you move; SoundHound named Overall Agentic AI Company of the Year (2026 AI Breakthrough Awards) and Leader in 2026 Gartner MQ for Conversational AI. Company size: ~1,000-2,000 (SoundHound AI, public, conversational/voice agents platform Amelia). Pain points: managing production AI agents plus distributed systems at scale; capturing why/decisions as agents build faster than docs can keep up; institutional-knowledge loss as AI accelerates. Challenges: agent reliability and traceability of decisions; scaling agentic voice AI across enterprise. Must-haves: production-grade agent reliability, decision/intent traceability. Nice-to-haves: automated intent-capture harness, cost accounting per harness component. ICP confidence: Medium (VP Eng plus agent-shipping strong; company is public/large-cap, headcount near 2,000 upper edge).
Christopher Martin — Co-Founder & CTO, Rilla
Signal: 4 (ICP technical leader at an AI-native company). Method: web/company research. Source: https://www.rilla.com , Tracxn/Komo profiles, https://www.linkedin.com/in/christopher-martin-5a9221113. Company size: 51-200 employees (2026; ~$78.9M raised, growth stage). Agent activity: speech AI / revenue intelligence for in-person field sales; 'Rick' AI assistant + virtual ridealongs; expanding toward agentic. CAVEAT: primarily speech-analytics/coaching, not confirmed 5+ autonomous agents in production - agent-shipping fit is weaker/INFERRED. Pain points (INFERRED, not verbatim): LLM cost of analyzing large volumes of sales-conversation transcripts; reliability of AI coaching/assistant outputs at scale. Challenges: cost-per-conversation processed. Must-haves: cost visibility as volume scales. Nice-to-haves: model routing. ICP confidence: Medium (right size 51-200, AI-native, CTO; agent-shipping criterion weaker - flagged).
Signal: 1 (senior technical leader at agent-shipping company; found via web research — LinkedIn/Chrome unavailable this run). Source: https://www.boringbusinessnerd.com/startups/aisera ; https://startupintros.com/orgs/aisera. Company size: ~340 employees (2026). Aisera ships agentic AI automating IT/HR/finance/customer service (5+ agents in production); acquired by Automation Anywhere (2025). Christos Tryfonas is Chief Architect (former Head of Engineering, Splunk Behavioral Analytics). Pain points (INFERRED from role/company context — NOT verbatim): scaling agent architecture reliably, controlling per-run/token cost, production observability across many agent workflows. Challenges: keeping a large multi-domain agent platform performant and cost-efficient. Must-haves: cost + reliability visibility per agent workflow. Nice-to-haves: routing/optimization tooling. ICP confidence: Medium (Chief Architect ≈ VP-level technical leader at a 50–2,000-emp agent company; title sits at the senior-technical edge of the Director–VP band; company late-stage/acquired, noted as caveat). NOTE: pain points inferred, not quoted.
Cijo George — VP & Head of AI, Practo
Signal: 4 (surfaced in LinkedIn people search; explicit '@ Practo' headline: 'VP & Head of AI @ Practo | Building Human-centric AI for Healthcare'). Source: https://www.linkedin.com/search/results/people/?keywords=Observe.AI%20VP%20engineering%20AI | Company size: ~1,000-2,000 (est). Pain points (INFERRED from role+company): building/operating healthcare AI (assistant/agent) features reliably and safely; cost + reliability as AI usage scales. Challenges: production reliability + safety/governance for healthcare AI. Must-haves: reliability + governance. Nice-to-haves: cost-per-run visibility. ICP confidence: Medium (senior AI leader — VP & Head of AI — at a 50-2,000-emp company; CAVEAT: Practo is a healthcare platform and its status as an active shipper of 5+ production AI agents is NOT confirmed, so agent-native fit is weaker than the Aisera/Dialpad/Uniphore adds; company net-absent from brain).
Clara Matos — Director of Applied AI, Sword Health
Signal: Bucket 4-adjacent (ICP AI leader at agent-native healthtech). Sourced via LinkedIn people search "Sword Health AI engineering director". Source: https://www.linkedin.com/in/claramatos/ . Company size: ~1,000-1,520 employees (web-verified; Sword Health, ~$3-4B valuation, Series D+ unicorn — note: stage past A-C band, but size in 50-2000 band). Agent-native: builds Phoenix, an AI care-specialist agent giving real-time clinical feedback in production to patients. Pain points (inferred from role/company; no public post captured this run): reliability/safety of clinical AI agents in production, cost per patient interaction at scale, compounding/observability of agent behavior. Challenges: governing autonomous clinical agents; controlling inference cost across a large patient base. Must-haves: reliability + auditability, cost-per-run control. Nice-to-haves: portability, compounding memory. ICP confidence: Medium (Director of Applied AI ✓, size ✓, agent-native ✓; stage Series D+; no direct engagement signal observed).
Clarence Chio — Co-Founder & CTO, Unit21
Signal: Bucket 4 (ICP technical leader shipping agents in production) — found via web research; LinkedIn/Chrome channel unavailable this run.
Source: https://www.unit21.ai/products/ai-agent ; https://www.crunchbase.com/organization/unit21 ; https://www.rsaconference.com/experts/clarence-chio
Company size: ~116–120 employees (Crunchbase/Tracxn, May–Jun 2026). Series C ($45M, Jun 2023; ~$92M total). SF-based, founded 2018 with Trisha Kothari. 200+ financial institutions incl. Intuit, Chime, Sallie Mae.
Agent activity (verified): Unit21 ships "configurable AI agents [that] run the full financial crime lifecycle, detecting risk in real time and carrying out investigations end-to-end." Product page claims 93% false-positive reduction and 80% faster handle times. Multiple distinct agents across fraud + AML + SAR filing = multi-agent production fleet.
Pain points (inferred from product positioning, not a direct quote): end-to-end autonomous investigation agents running at real-time volume across 200+ institutions — per-case inference cost and reliability are direct COGS and compliance-risk lines. False-positive rate is their headline metric, i.e. agent output quality is the product.
Challenges: regulated-industry auditability of agent decisions (SAR filings); keeping per-investigation cost bounded as agents replace human analysts at scale; multi-tenant cost attribution across 200+ institutions.
Must-haves: per-run / per-case cost visibility and attribution by customer; deterministic audit trail of agent actions; reliability guardrails in a regulated workflow.
Nice-to-haves: model routing / cost optimization across investigation steps; eval harness for agent decision quality.
ICP confidence: High — CTO + technical co-founder, 116–120 emp (in range), Series C, agentic AI is the core product not a feature. Author of O'Reilly "Machine Learning & Security"; publicly technical (QCon, RSAC speaker), so reachable via technical content.
Clay Bavor — Co-founder (technical, ex-Google Labs), Sierra
Signal: 4 (ICP technical co-founder at AI-native company shipping agents in production). Source: https://www.linkedin.com/in/claybavor/ , https://www.forbes.com/sites/richardnieva/2025/11/05/sierra-bret-taylor-clay-bavor/ , https://research.contrary.com/company/sierra , https://www.cnbc.com/2026/05/04/bret-taylor-sierra-fundraise-openai.html | Company size: ESTIMATE ~300–700 employees (not independently confirmed this run; Sierra founded 2023, $150M ARR in 8 quarters, $15.8B valuation May 2026 — clearly a mid-size startup within 50–2,000, but exact headcount not verified — do not treat as confirmed). AI-native platform deploying autonomous customer-experience agents that handle real transactions (claims, returns, mortgage refinance) for 40%+ of the Fortune 50. Bavor is technical co-founder (18 yrs at Google leading Labs, AR/VR, Workspace). Pain points: reliability/trust of agents taking real financial actions in production; controlling agent behavior across many enterprise deployments; inference cost at enterprise scale. Challenges: high-stakes reliability; behavior control/guardrails; per-run cost visibility. Must-haves: reliability, control, observability. Nice-to-haves: cost-per-run analytics. ICP confidence: Medium (technical co-founder at an AI-native agent company shipping in production; company size fits by stage but exact headcount unverified this run — flagged for confirmation before outreach).
Signal: 4 + 1 (ICP technical leader shipping agents; also publicly writing about agent operations/governance). Source: https://siliconangle.com/2026/03/09/dataiku-evolving-orchestration-layer-enterprise-grade-ai-agents/ ; https://www.dataiku.com/stories/blog/looking-ahead-with-clement-stenac ; Cobuild GA June 2026; Tracxn May 2026. Company size: ~1,600 employees (Tracxn 1,621; $350M ARR, Series). Pain points (inferred from public product/positioning): customers "building agents instead of just models" at enterprise scale; needs testing, human-in-the-loop hooks, MCP support, and tight control over the tools agents can access. Just shipped Agent Management giving "one place to track and govern agents across the enterprise with KPIs for every agent in operation, regardless of where it runs" — i.e. cross-fleet agent observability + per-agent KPIs/cost. Challenges: governing/observing a sprawling fleet of enterprise agents across heterogeneous runtimes. Must-haves: agent governance, per-agent KPIs, tool-access control, testing. Nice-to-haves: unified cost visibility per agent run. ICP confidence: High (co-founder+CTO; 50-2,000 emp; agents in production; explicit agent-observability pain).
Cody Nutt — Senior Director of Business Systems, Daxko
Signal: 1 (ICP-adjacent author — AI budget owner rather than engineering leader).
Source: https://www.linkedin.com/search/results/content/?keywords=%22our%20token%20costs%22%20OR%20%22our%20inference%20costs%22%20agents%20production&datePosted=%22past-month%22&sortBy=%22relevance%22
Company size: ~887–893 (Tracxn 31 Jul 2026 / LeadIQ May 2026); Revelio Labs 1,171 (Mar 2026). In band on every source. Health/wellness SaaS, Birmingham AL.
Pain points (his own words): "our Claude token costs grew more than 2.5x in two months and I couldn't explain why. Breaking token usage out by model, user, and team is what answered it — one weekly briefing showed over 20% of our Claude spend running through Opus for work Sonnet handles fine. Changed the default, kept the access."
Challenges: Unexplained 2.5x AI spend growth in two months; no native breakdown by model/user/team; frontier-model overspend on routine work; wants AI spend managed alongside cards/AP/procurement rather than in separate tooling.
Must-haves: Model/user/team-level token attribution; a recurring reporting cadence rather than one-off audits; ability to change defaults without cutting off access.
Nice-to-haves: Consolidation into an existing spend-management workflow.
ICP confidence: Medium — Director+ seniority and a real AI budget owner at an ~890-person SaaS, with the exact spend-attribution pain stated verbatim. TWO CAVEATS: (1) his title is Business Systems, not Engineering/AI, so he is not the core technical-AI-leader persona — treat as a budget-side entry point and find the eng counterpart at Daxko; (2) he solved this with Ramp AI Token Spend Management, so this is a competitive-displacement or expansion conversation, not greenfield.
Signal: 4 (infra leader at agent/inference-infra company). Discovery: company-scoped LinkedIn People search. Source: https://www.linkedin.com/in/colin-mcgrath/ | Company: Baseten — AI inference platform for production & agentic workloads. Company size: ~312 employees (PitchBook 2026) — in band. Pain points (INFERRED from role): infra cost at inference scale, reliability/uptime of production inference, capacity/cost per run. Challenges: cost governance + reliability of high-volume inference. Must-haves: cost-per-run visibility, reliability. Nice-to-haves: optimization automation. ICP confidence: Medium (Head of Infrastructure; infra-adjacent ICP; in-band).
Signal: 4 (ICP technical leader at company shipping AI agents in production). Source: https://www.unifygtm.com/blog/series-b , https://www.businesswire.com/news/home/20250714813159/en/ , https://tracxn.com/d/companies/unify . Company size: ~145 employees (Mar 2026), Series B ($40M led by Battery Ventures; OpenAI Startup Fund participated; $260M valuation). Ex-Scale AI ML research engineer. Customers include Cursor and Perplexity. Pain points: running outbound sales AI agents that 'take any action a seller can' — multi-agent orchestration for prospecting/personalization/engagement at scale; keeping agent output quality high while scaling volume 8x. Challenges: reliability and quality control of autonomous outbound agents; scaling many agents without cost/quality blowup; visibility into what each agent is doing and costing. Must-haves: reliability + control of agents at scale, agent observability. Nice-to-haves: per-run cost attribution, model routing to manage LLM spend on high-volume outreach. ICP confidence: High — technical co-founder/CTO at ~145-person Series B AI-native company whose core product is production sales agents. (Note: profileUrl is company site; personal LinkedIn not verified this run.)
Cooper Oelrichs — Head of Engineering for AI, Nelly Solutions
Signal: 1/4 (ICP identified directly via LinkedIn people search, adjacent to Parloa results). Source: https://www.linkedin.com/in/cooper-oelrichs-0977039a/
Company size: ~150-250 (Nelly Solutions, Berlin health-tech; Series B; building AI/agent automation for medical practices)
Pain points (inferred from title + Nelly AI product direction): reliability & compliance of healthcare AI agents; cost of automating patient onboarding/admin/billing workflows.
Challenges: shipping reliable, auditable agents in a regulated healthcare setting.
Must-haves: reliable/observable agents, compliance-safe automation.
Nice-to-haves: cost visibility per automated workflow.
ICP confidence: Medium (clear Head-of-Engineering-for-AI title and in-range headcount; agent-building inferred from Nelly AI direction, not confirmed from a post this run). New company for Brain.
Signal: 4 (senior technical leader at company actively shipping AI agents; found via web research — LinkedIn/Chrome not connected this run). Source: https://rocketreach.co/hightouch-management_b43db6d7c19ca279 ; https://www.linkedin.com/in/corey-stein-63822744/ ; company context https://unicornscreener.vc/blog/7-ai-agent-startups-funded-by-top-vcs-in-2026. Company size: ~450-600 (Hightouch; $150M Series D Apr 2026 at $2.75B; 100%+ YoY growth two years running; enterprise "AI Decisioning" / always-on marketing agents used by Domino's, PetSmart, DraftKings, Ramp, Whoop). Pain points: running always-on autonomous marketing agents reliably at enterprise scale across ads/email/SMS/web; agent cost/token spend as agents proliferate; measurable outcome delivery in production. Challenges: scaling from a data-activation platform to multi-agent orchestration; reliability + governance of agents acting on customer data. Must-haves: production reliability, cost visibility per agent/run, observability. Nice-to-haves: cross-channel orchestration efficiency. ICP confidence: High (SVP Engineering — Head-of-Engineering-level — at a 50-2,000-emp company shipping agentic products in production).
Cyprien de Masson d Autume — Co-Founder & CTO, Reka AI
Signal: 4. Source: https://www.crunchbase.com/person/cyprien-de-masson-d-autume ; https://tracxn.com/d/companies/reka/. Company size: ~50-60 employees (Sunnyvale; ~$170M raised; Series B; agentic multimodal platform; ex-DeepMind founders). Pain points (inferred; no direct quote this run): building reliable agentic platforms with a small team; multimodal agent reasoning cost and efficiency. Challenges: competing on agentic capability with limited headcount; efficiency of agent inference. Must-haves: efficient, reliable agent inference. Nice-to-haves: better agent evaluation. ICP confidence: Medium (CTO and technical co-founder at ~55-emp Series B; size near the 50 floor; more model/agentic-platform than 5+ shipped production agents). Note: name is Cyprien de Masson d'Autume.
Signal: 4 (Head of AI at ICP agent company; via LinkedIn People directory). Source: https://www.linkedin.com/in/danbikel (Writer = enterprise agentic AI). Company size: 201-500, Series C. Pain points (inferred): model quality vs cost tradeoffs; reliability/evals for enterprise agents; context efficiency. Challenges: balancing quality, cost, latency for enterprise agentic AI. Must-haves: eval/observability + cost control. Nice-to-haves: cheaper inference paths. ICP confidence: High. NOTE: pain points inferred from role/company, not verbatim quotes.
Signal: 4 — ICP at a qualifying company shipping agents. Dan Bikel appointed Head of AI at Writer to lead enterprise agentic AI. Writer ships enterprise AI agents in production (AI HQ platform, Writer Agent, Skills) for Fortune 500s.
Source: https://writer.com/newsroom/ ; https://venturebeat.com/ai/writer-unveils-ai-hq-platform-betting-on-agents-to-transform-enterprise-work
Company size: ~563 employees (confirmed, multiple 2026 sources); Series C, AI-native enterprise platform.
Pain points: getting enterprise agents from pilot/demo to reliable production (Writer's 2026 survey: 79% of enterprises face AI adoption challenges despite high investment); agent reliability & governance at enterprise scale.
Challenges: orgs struggle to reach production reliably; change management; controlling agent behavior across many enterprise use cases.
Must-haves: agent reliability, governance/control, per-agent cost visibility.
Nice-to-haves: reusable agent building blocks encoding team standards (Writer "Skills").
ICP confidence: High — Head of AI, ~563-employee Series C AI-native company actively shipping agents in production.
Dan Bălăceanu — Chief Product Officer and Co-Founder, DRUID AI
Signal: 1 (ICP speaking publicly about evaluating/controlling enterprise agents) + 4 (ICP shipping agents).
Source: AI Engineer World's Fair 2026 speaker record — https://ai.engineer/worldsfair/2026/speakers.json ; session "Would your AI agent get the job? A performance review framework for enterprise agents".
Company size: ~211 employees (Tracxn, Apr 2026). Series C — $31M, Sep 2025; $78.5M raised total. Conversational/agentic AI platform; Gartner Challenger, Conversational AI Platforms 2025. FULL ICP match on stage AND headcount.
Pain points (from his own session abstract, paraphrased): no way to compare agent implementations "on the same task, with the same data, against the same KPIs" across agentic frameworks, direct LLM APIs, conversational AI platforms and vertical SaaS; vendor claims are unfalsifiable.
Challenges: enterprises cannot tell whether a deployed agent is actually performing; evaluation is vendor-specific and non-portable.
Must-haves: vendor-agnostic, KPI-based agent performance measurement; per-agent accountability ("treat agents the way enterprises treat new hires — set the role, then review performance").
Nice-to-haves: benchmarking across competing agent build approaches; standardised agent "performance review" reporting for enterprise buyers.
ICP confidence: High — technical co-founder + CPO, 211 emp, Series C, AI-native, agents are the core product. Only caveat: CPO is a product (not engineering) title, though co-founder status makes him a technical decision-maker.
LinkedIn URL: not published in source dataset — do not fabricate; needs manual lookup before outreach.
Dan Eisenberg — Head of Engineering, Hex
Signal: 2 — ICP engaging with non-ICP practitioner content about agent capability/reliability limits.
Source: https://www.linkedin.com/in/dan-eisenberg/recent-activity/all/ (post 2w ago, commenting on Izzy Miller's "DataBench" frontier benchmark launch). Found via https://www.linkedin.com/company/hex-technologies/people/?keywords=head%20of%20engineering
Company size: 51-200 employees (LinkedIn company page, verified this run). Hex = AI-native analytics workspace shipping Notebook/Magic agents in production, Series C.
Observed this run (verbatim): "Agents may be solving Erdős problems but they can still struggle with complex analytics tasks. Incredible work from Izzy Miller measuring frontier model performance on hard data problems, with super interesting results!" He also ships internal generative data apps ("I built a pizza-themed build tracker in Hex as a generative data app... I've never been more motivated to track down flaky tests!").
Pain points: agent reliability on complex, multi-step analytical work — headline benchmark performance not translating into dependable production behaviour; needs real measurement of frontier-model agent performance rather than vibes.
Challenges: evaluating agents on hard, long-horizon data tasks; flaky-test/build reliability at the same time.
Must-haves: trustworthy evaluation of agent trajectories on complex tasks before shipping to users.
Nice-to-haves: cost/latency attached to those evals so model choice is an economic decision, not just a quality one.
ICP confidence: High — Head of Engineering (core ICP title) at a 51-200 employee Series C AI-native company with agents in production, and an explicit public reliability signal.
Dan Feng — Senior Director of Engineering, Maven Clinic
Signal: 4 (ICP speaking at AI Engineer World's Fair 2026 on building AI agents at a health company). Source: https://www.ai.engineer/worldsfair/schedule ; size verified via Revelio. Company size: ~911 employees. Maven Clinic is designing, building and deploying end-to-end AI agents (agent frameworks; 'Maven Intelligence' agentic AI) in production for care navigation. Pain points (inferred): reliability of agents in a healthcare setting, knowledge/context infrastructure to make agents effective, cost and visibility as agents scale. Challenges: deploying trustworthy agents in production across clinical workflows. Must-haves: reliability, observability, context management. Nice-to-haves: per-run cost attribution. ICP confidence: Medium — Director+ engineering leader at a ~911-person company actively shipping agents; in size band but late-stage (Series F) vs. the A-C core target.
Dan Fu — VP of Kernels, Together AI
Signal: 1/4 (ICP speaking/writing about building, managing, and scaling agents + inference cost). Source: https://www.ai.engineer/worldsfair/schedule — talk "Agents at Scale: Inside MiniMax's Model and the Infrastructure Behind It"; also Slush 2025 talk on "building, using, and managing AI agents." Company size: ~410-414 employees, Series B/C (~$3.3B valuation), AI inference/infra platform used to run agents at scale. Pain points: cost and efficiency of inference under agentic workloads (agents consume far more tokens/compute), building efficient GPU kernels to cut cost, scaling agent infra reliably. Challenges: keeping agent-at-scale inference fast and cheap; managing agents in production. Must-haves: low-cost, high-throughput inference; efficient kernels/routing for agent workloads. Nice-to-haves: cost-per-run/agent attribution and observability. ICP confidence: Medium — VP-level technical leader (ex-Stanford, FlashAttention co-author) at a right-sized company whose work centers on agent cost/scale; caveat: Together is an inference-infra provider rather than a vertical agent-app builder. Note: web research + verification; LinkedIn not connected this run.
Dan Neil — Chief Technology Officer, Formation Bio
Signal: 4 (ICP author shipping agents — LinkedIn post: "building a new kind of pharma company: technology, agents, and LLM-integrated systems shipping into production for therapeutic diligence, R&D, and trial execution"; hiring a VP of Engineering). Source: post surfaced via content search "shipping agents in production VP engineering", 3w ago on his profile activity (https://www.linkedin.com/in/danielneil/). Company size: 51-200 (LinkedIn; ~199 associated members). Pain points: needs to scale an AI-native engineering org where agents and LLM-integrated systems ship into production across diligence, R&D and trial execution; wants a leader with a strong POV on how AI reshapes engineering teams. Challenges: building production-grade agent/LLM systems in a regulated pharma/techbio context; reliability of agent-driven R&D workflows. Must-haves: production reliability for AI-native/agent systems; AI-native engineering leadership and practices. Nice-to-haves: candidates who have built AI-native systems from scratch. ICP confidence: High (CTO, PhD in AI; AI-native company 51-200 emp actively building and shipping agents/LLM systems in production).
Dan Versoi — Technical Director & Software Architect, DevRev
Signal: 4 (ICP Director-level engineering/architecture leader at validated target company DevRev; sourced via LinkedIn People search, run 2026-08-01). Source: https://www.linkedin.com/in/dan-versoi/ . Company size: ~700-1,000 (DevRev — AI-native support/CRM platform shipping AI agents via AgentOS; in-range). Pain points (role/company-contextual): architecting reliable, cost-controlled agent systems in production; visibility into per-agent/per-run cost; guardrails for non-deterministic agent loops. Challenges: multi-tenant agent scalability & reliability; cost attribution across models/agents. Must-haves: per-agent cost & reliability observability; guardrails. Nice-to-haves: model routing, eval automation. ICP confidence: Medium-High (Technical Director/Architect at 50-2,000-emp AI-native agent-shipping company).
Signal: 2/4 (ICP engaging with agent content plus building production agents). Source: https://www.linkedin.com/in/danaletendre/ — headline Building Production AI Systems for Insurance Operations; recent comment on an agent post to the effect that it is all about the data or your AI agents cannot do anything (data-grounding/reliability signal). Company size: ~50-150 (COVU, AI-native insurtech, Series A, AI agents for insurance agencies and operations). Pain points: grounding agents in reliable data so they can act; production AI for insurance ops. Challenges: agent reliability/data-dependence, deploying agents into regulated insurance workflows. Must-haves: data-grounded, reliable production agents. Nice-to-haves: cost/observability per agent run. ICP confidence: Medium (Head of AI at AI-native insurtech building production agents; company size estimated ~50-150, Series A, borderline lower bound).
Daniel Hoske — CTO, Cresta
Signal: 1 (ICP writing publicly about agent cost/reliability in production). Source: https://cresta.com/blog/engineering-for-real-time-voice-agent-latency ; https://cresta.com/blog/building-and-deploying-production-grade-ai-agents-crestas-end-to-end-approach . Company size: ~500 employees (up from ~285 in 2024); note: Cresta is Series D — later stage than A–C but within the 50–2,000 headcount band. Real-time voice/chat AI agents + agent-assist for enterprise contact centers. Pain points: sub-second latency across ASR/LLM/TTS, "speculative triggering" causing unnecessary costs, production-grade reliability of voice agents at scale. Challenges: optimizing cost/latency tradeoff while keeping agents reliable in real-time customer conversations. Must-haves: cost control on speculative/parallel model calls; production reliability. Nice-to-haves: unified observability across the ASR/LLM/TTS pipeline. ICP confidence: High (caveat: Series D, upper end of size band).
Signal: Bucket 4 (ICP engineering leader at a company shipping clinical AI agents at scale).
Source: https://theorg.com/org/smarterdx#teams ; https://rocketreach.co/smarterdx-management_b7f64584c2ad3025 ; https://www.smarterdx.com/resources/smarterdx-raises-50m-to-bolster-hospital-revenue-integrity-and-quality-with-its-clinical-ai-solution
Company size: 368 employees as of 30 Apr 2026 (Tracxn). Series B — $50M led by Transformation Capital (May 2024), with Bessemer, Flare Capital, Floodgate; ~$71M total. NYC. Founded by Michael Gao MD (CEO) and Josh Geleris MD (Head of Product & Data Science). Reported ~$71M revenue.
Agent activity (verified): SmarterDx analyzes 100% of patient charts, captures missed revenue from incomplete documentation, and AUTOMATES DENIAL APPEALS — i.e. autonomous document-processing and drafting agents running over every chart in a hospital system. High-volume, long-context clinical documents.
Pain points (inferred from workload shape, not a quote): every-chart-for-every-patient coverage means token/context cost scales with hospital volume, not with seats — classic context/token waste and cost-per-run exposure. Clinical accuracy requirements make reliability non-negotiable.
Challenges: long clinical documents drive large context windows; multi-step review + appeal drafting chains; per-hospital cost attribution as they add health-system customers.
Must-haves: cost per chart / per appeal visibility; reliability and accuracy guardrails on clinical output; regression detection when models change.
Nice-to-haves: context pruning to cut token waste on long charts; routing cheap models for triage and expensive models only for escalation.
ICP confidence: Medium-High — "Head of Engineering" is explicitly in the ICP title list; 368 emp and Series B are both squarely in range; agent-style automation is the core product. Downgrade from High only because his title/tenure is sourced from org-chart aggregators (TheOrg/RocketReach) rather than a first-party page, and no LinkedIn URL confirmed. VERIFY TITLE before outreach.
Daniel Palmer — Co-Founder (technical), Relevance AI
Signal: 4 (technical co-founder at a qualifying agent-platform company)
Source: https://www.bvp.com/news/meet-the-founders-of-relevance-ai | https://relevanceai.com/blog/the-ai-workforce-revolution-24m-series-b-to-accelerate-our-mission
Company size: ~124 employees (Series B ~$24-37M led by Bessemer; Forbes 30U30 Asia founders)
Pain points (inferred): "AI workforce" platform where anyone builds & recruits teams of AI agents; ~40,000 agents created on the platform in a single month (Jan 2025) — scaling orchestration, reliability, and cost across a large, user-generated agent population.
Challenges: reliability & control of user-built agents at scale; cost management across many concurrent agents; guardrails for autonomous multi-agent runs.
Must-haves: reliable multi-agent orchestration; control/guardrails; scale.
Nice-to-haves: per-agent cost & observability surfaced to builders.
ICP confidence: Medium (technical co-founder — prior technical co-founder at Tribefire; exact title at Relevance unspecified; agent-platform decision-maker).
Daniel Rothman — Head of Engineering, Assort Health
Signal: Bucket 4 (ICP engineering leader at a company shipping agents — found via LinkedIn people search on Assort Health eng team). Source: https://www.linkedin.com/in/daniel-rothman-b90880a6/ ; company confirmation https://www.fiercehealthcare.com/ai-and-machine-learning/assort-health-scores-120m-series-c-scale-voice-ai-agent-platform-healthcare . Company size: ~100–150 and growing fast (public reporting: 15 employees in 2025 → crossing 250 by end of 2026); Series C $120M. Assort Health ships specialty-specific voice AI agents in production that answer patient calls, schedule appointments and verify insurance for healthcare providers. Head of Engineering = owns production reliability and scale. Pain points: agent reliability in a regulated/high-stakes healthcare setting, cost per patient call at scale, no clean visibility into per-run cost across specialties. Challenges: scaling agents across many clinics/specialties while keeping them reliable and compliant; controlling cost of high call volume. Must-haves: reliability guardrails + per-run cost/observability across a rapidly growing agent fleet. Nice-to-haves: standardized eval + operating layer as headcount and agent count scale 15→250. ICP confidence: High (Head of Engineering at a Series C company actively shipping voice agents in production).
Daniel Saks — Co-Founder & CEO (technical; co-founded AppDirect), Landbase
Signal: 4 (ICP building/shipping agents; publicly posts on agentic AI transforming GTM). Source: https://www.linkedin.com/in/danielsaks/ ; https://www.landbase.com/about ; https://www.linkedin.com/posts/landbase_daniel-saks-ceo-of-landbase-shares-insights-activity-7300213039482478592-tXhi. Company size: ~74 employees (Jan 2026 snapshot; ~50 earlier in 2026), Series A, $42.5M raised, founded 2024, San Francisco. Actively building AI agents: yes — GTM-1 Omni action model deploys multiple specialized agents that autonomously run the full go-to-market motion (trained on 40M+ campaigns). Pain points: coordinating/orchestrating multiple specialized agents end-to-end in production; proving ROI (conversion uplift) vs. cost of autonomous multi-agent GTM. Challenges: reliability and cost of multi-agent orchestration at scale; agents adapting on outcomes. Must-haves: coordinated agents that adjust based on what converts; ROI visibility. Nice-to-haves: cost-per-run and agent observability. ICP confidence: High-Medium (technical founder/CEO, Series A, ~74 employees, multi-agent product in production). Note: pain points inferred from product/role context, not a verbatim quote.
Daniel Sheard — Head of Enterprise Engineering, Lovable
Signal: 4 — ICP engineering leader at an agent-native company in an active enterprise push.
Source: https://www.linkedin.com/in/danielsheard/recent-activity/all/ (found via https://www.linkedin.com/company/lovable-dev/people/?keywords=head%20of%20engineering)
Company size: 51-200 employees (LinkedIn company page, verified this run). Lovable = AI app-building agent, Series C announced ~this month.
Observed this run: reposts Lovable company milestones (lovable.dev → lovable.com) and Morgan Jacobson (Head of Revenue Strategy @ Lovable), who writes: "Off the back of the series C announcement earlier this week... Lovable is investing heavily in our product capabilities for enterprise as our customers shift from using AI as a chatbot to non-technical employees shipping applications that run their internal operations."
Pain points: NOT expressed in his own words. Observed context: Lovable is moving agent workloads into enterprise environments post-Series C, where governance, reliability and predictable per-customer cost become buying criteria.
Challenges (inferred): enterprise-grade controls on agent execution; non-technical users generating unbounded agent work.
Must-haves (inferred): guardrails/budgets per enterprise tenant.
Nice-to-haves (inferred): per-customer cost attribution.
ICP confidence: Medium — correct title level (Head of Engineering scope) and a qualifying company (51-200, Series C, agents in production), but no first-person pain signal captured this run. Pair with Patrik Torstensson when working the Lovable account.
Daniel Siegel — Co-founder & Chief Product Officer, Synera
Signal: 4 (AI-focused product leader / technical co-founder at an ICP agent-shipping company). Source: https://www.synera.ai/about ; https://siliconangle.com/2026/04/14/german-startup-synera-lands-40-million-automate-engineering-workflows-ai-agents/ | Company size: ~102 employees (Tracxn, May 2026); Series B agentic-AI engineering platform, 60+ enterprise customers. | Pain points: (inferred) shipping agentic product that reliably executes autonomous engineering workflows; proving pilot→production value to industrial buyers; managing agent behavior/quality across 80+ integrated tools. | Challenges: product-level reliability and trust of autonomous agents; measuring outcomes/ROI per agent run. | Must-haves: production reliability, observability, per-run cost/outcome visibility. | Nice-to-haves: agent evaluation + iteration tooling. | ICP confidence: Medium — matches the "VP/Head of Product (AI-focused)" persona (CPO at an agentic-AI company); product (not eng) title, hence Medium.
Daniel Sikorskiy — Chief Architect, Wonderful
Signal: 1 (ICP author writing about agent reliability and self-verification in production).
Source: https://www.wonderful.ai/blog-articles/going-codeless (byline on wonderful.ai).
Company size: 350 employees as of 12 Mar 2026, scaling to ~900 by year-end — verbatim from Insight Partners Series B release. Series B, $150M at €1.7B.
Pain points: (verbatim) "Early on, models would confidently produce code that didn't work. They'd hand back something that looked right, and it was only when you ran it that you'd find out it wasn't... We built infrastructure that required models to test their own work before a task was considered complete."
Challenges: Confident-but-wrong agent output; failures only detectable at execution time; had to build verification infrastructure in-house.
Must-haves: Structured verification after every agent/tool step; run-time correctness signal before a task is marked complete.
Nice-to-haves: Verification as a platform primitive rather than bespoke per-agent code; regression detection as models change.
ICP confidence: Medium-High — Chief Architect is the technical decision seat below CTO, headcount verified from primary source, company ships enterprise agents at scale. Slightly below High only because "Chief Architect" is not a named ICP title, though it is functionally Director+ technical.
Account note: second Wonderful contact (with Ben Avitouv, Field CTO, and Roey Lalazar CTO already in library) — Wonderful is now a 4-contact account and should be treated as a priority multi-threading target.
Daniel Vassilev — Co-founder & Co-CEO (technical), Relevance AI
Signal: 4 (ICP technical co-founder building & shipping agents in production). Source: https://techcrunch.com/2025/05/06/relevance-ai-raises-24m-series-b-to-help-anyone-build-teams-of-ai-agents/ , https://relevanceai.com/blog/the-ai-workforce-revolution-24m-series-b..., https://www.bvp.com/news/meet-the-founders-of-relevance-ai , https://tracxn.com/d/companies/relevance-ai/. Company size: 124 employees (as of May 2026). Series B ~$38M total ($24M round led by Bessemer; King River, Insight, Peak XV); Sydney/SF. "AI Workforce" — an AI-agent operating system / no-code multi-agent builder ("Workforce", "Invent" text-to-agent); 40,000+ agents created in a single month. Vassilev is technical co-founder & co-CEO (with Jacky Koh; Daniel Palmer co-founder). Pain points: helping customers run large numbers of multi-agent teams reliably in production; cost/observability as agent volume explodes across a big customer base; quality/reliability of autonomously generated agents. Challenges: reliability and cost control across massive multi-tenant agent volume; per-run cost attribution to customer credits/pricing. Must-haves: reliability at scale, per-run cost visibility, observability. Nice-to-haves: behavior control/guardrails across tenants. ICP confidence: Medium — technical co-founder/co-CEO at a 124-person AI-native company shipping agents in production; note: Relevance is an agent-platform VENDOR (partner/competitor-adjacent, like Dust/Voiceflow already in library).
Daniel Whitston — Chief Technology Officer, AutogenAI
Signal: 4 (ICP shipping agents). Source: https://autogenai.com/uk/about/about-us/. Company size: 194 (Tracxn 2026). London UK with a dual New York HQ. Series A 2023 (Blossom/Spark Capital) plus Series B Dec 2023 (Salesforce Ventures) — cleanly inside the Series A-C band. Agent evidence: AutogenAI's product is a multi-step "language engine" (Genny-1) that autonomously drafts, reviews and refines full bid and proposal documents at scale; client Serco reports 6,000+ uses of the pilot functionality. Whitston is listed as current CTO on the official company leadership page. Pain points: NOT directly quoted from Whitston — inferred from the product surface, where high-volume autonomous document generation implies inference cost and reliability pressure. Recorded as inference, not as a claimed statement. Challenges: HQ framing has shifted to dual London/New York, so geography is ambiguous; London remains the EMEA base and product origin. Must-haves: unknown — no direct statement found. Nice-to-haves: unknown. ICP confidence: Medium — title, headcount and funding stage all verified cleanly, but no direct quote from Whitston about agent cost or reliability pain. Worth a discovery call rather than a pain-led cold open. Weakest evidentiary link among this run's Medium entries alongside Peter Hill.
Daniele Alfarone — Senior Director of Engineering, Dixa
Signal: 3 (ICP engaging with competitor content — named in a Humanloop customer case study; Humanloop is a direct adjacent in the eval/observability layer).
Source: https://humanloop.com/case-studies/dixa
Company size: ~158-161 employees (Crustdata / Tracxn, 2026). Series C, $105M raised. Copenhagen, Denmark — conversational customer-service platform shipping AI agents.
Pain points (from the case study): "Building AI products via traditional machine learning methods comes at an extremely high cost and slower development times." Dixa built dashboards specifically for "Compute Cost Monitoring: Track and analyze compute costs associated with AI model usage" and tied that data to pricing strategy.
Challenges: cross-team prompt iteration bottleneck between product, engineering and ML; GDPR/EU data-residency constraints on any observability vendor; regression risk when swapping models.
Must-haves: compute-cost visibility granular enough to feed product pricing decisions; regression testing across model swaps; EU-compliant deployment.
Nice-to-haves: gamified internal adoption tooling to get non-engineers iterating on prompts.
ICP confidence: High — Senior Director of Engineering (in band), headcount in band, Series C, and the company has explicitly built internal cost-monitoring dashboards. A team that built its own cost dashboard has already decided the problem is real and has already paid for it once in engineering time.
OUTREACH ANGLE: Dixa is one of very few prospects on record connecting agent compute cost to its own pricing strategy — cost-per-run is a revenue-margin question for them, not an infra hygiene question. Non-US (Denmark), which matters for the EU-residency story.
Danny Yates — Director of Engineering / Head of Engineering, Tray.ai
Signal: 4 (ICP leader at a company actively shipping AI agents). Found via LinkedIn company People-tab sweep of Tray.ai.
Source: https://www.linkedin.com/company/tray-ai/people/?keywords=engineering
Company size: 201-500 employees (per Tray.ai LinkedIn company page, London/SF). Tray.ai = integration/automation platform now shipping "Merlin" AI agents; Series C.
Pain points: NONE DIRECTLY OBSERVED THIS RUN — headline reads "Director of Engineering; Head of Engineering". Inferred only: agents that fan out across customer integrations multiply tool calls and token spend per workflow.
Challenges (inferred): cost of agentic workflows vs. deterministic automation runs — their existing pricing model is per-workflow-run, so agent token cost directly compresses margin.
Must-haves (inferred): cost-per-run instrumentation, budget caps per customer workspace.
Nice-to-haves (inferred): model routing between cheap/expensive models per step.
ICP confidence: Medium-High. Role and company both fit cleanly; the margin-compression angle is a strong hypothesis for a platform that sells runs. No expressed pain captured yet.
David Gildea — Vice President of AI Product, Druva
Signal: Bucket 3 — ICP engaging with competitor content (AWS Bedrock AgentCore). Found via web research; LinkedIn/Chrome unavailable this run.
Source: https://aws.amazon.com/solutions/case-studies/druva-agentcore-case-study/
Company size: ~1,370–1,386 employees (Tracxn, PitchBook, mid-2026). Data-protection SaaS.
Agent evidence: DruAI multi-agent product coordinating 8–10 specialized agents (data agents, help agents, action agents) in production; scaled to 3,000+ users and 17,500+ conversations in 4 months. Migrated OFF a custom LangChain-based stack onto a managed runtime.
Pain points: prior custom RAG/agent stack required "constant infrastructure management"; manual multi-person incident investigation process ("It was a very manual process that involved lots of people"); hallucination risk in a security-critical product (inferred from "no hallucinations" being the headline outcome).
Challenges: coordinating multi-agent workflows with scoped permissions and session isolation; holding reliability while scaling users 0→3,000 in one quarter.
Must-haves: managed/serverless agent runtime that removes infra burden; scoped-permission tool and API access; cross-session agent memory; guardrails against prompt injection.
Nice-to-haves: multimodal input (voice, screenshot analysis); MCP-based connectors into customer environments.
ICP confidence: High — VP AI title matches ICP exactly, headcount solidly in range, named on-the-record production multi-agent deployment, and an explicit build-to-buy migration off a self-built harness (the 1→5+ agent scale-wall pattern).
Profile URL: not found this run (do not fabricate).
David Hariri — Co-founder & Head of R&D, Ada
Signal: 1 (ICP technical co-founder writing publicly about agents; company at massive token scale). Source: https://www.ada.cx/blog/author/david-hariri/ and https://www.ada.cx/blog/q-and-a-with-david-hariri-ada-s-co-founder-and-head-of-r-and-d/ ; company scale per https://www.bvp.com/atlas/ada-architecting-fanatical-cx-loops-that-power-ai-agents . Company size: est. ~300-450 (Toronto, Series C; 350+ enterprise customers incl. Verizon, Pinterest, Monday.com, YETI). Now processing ~1.5 trillion tokens/month powering CX agents. Pain points: token cost/spend at enormous scale (1.5T tokens/month); reliability so agents "consistently outperform human teams"; feedback/CX loops to keep agents improving. Challenges: cost visibility & control at trillion-token scale; production reliability across 350+ enterprise deployments; per-customer agent quality compounding. Must-haves: cost attribution/control at scale; production reliability. Nice-to-haves: compounding improvement loops per customer. ICP confidence: Medium-High — technical co-founder & Head of R&D (decision-maker) at a ~350-person AI-native agent company operating at exactly the token scale where Alpha's cost+control wedge bites. (Note: Ada also reportedly appointed a new CTO, Jessica Popp — worth confirming as a second contact.)
David Hariri — Co-Founder & Head of R&D, Ada (Ada Support)
Signal: 4 (technical co-founder building & shipping customer-service AI agents in production). Source: https://www.ada.cx/blog/q-and-a-with-david-hariri-ada-s-co-founder-and-head-of-r-and-d/ ; https://www.bvp.com/atlas/ada-architecting-fanatical-cx-loops-that-power-ai-agents. Company size: ~250-400 (Ada, Toronto; Series C ~$130M raised; powers automation for 300+ companies incl. Meta, Verizon). Pain points: reliably automating complex multi-system workflows (a dozen+ actions per task) across 50+ channels/languages; autonomous-resolution accuracy; extending autonomy to generative voice. Challenges: reaching production-grade reliability across many channels; enterprise trust/compliance (SOC2, HIPAA, PCI). Must-haves: reliable autonomous resolution, orchestration across systems of record, visibility into agent actions. Nice-to-haves: cost efficiency per resolution. ICP confidence: High — technical co-founder / Head of R&D, Series C, 50-2000 employees, actively shipping AI agents in production. (Note: co-founder/CEO Mike Murchison already in brain.)
David Hsu — Founder & CEO (technical co-founder), Retool
Signal: 4 (technical founder at qualifying company actively shipping agents)
Source: https://retool.com/agents ; https://www.youtube.com/watch?v=zBliFZMJYyY (Replay 2026 - Retool agent orchestration on Temporal)
Company size: ~415 employees (Series C, ~$165M raised)
Pain points: extending workflow engine into reliable agent orchestration; governing autonomous agents doing real work
Challenges: production reliability & state management for long-running agents; scaling agents across customers
Must-haves: durable agent orchestration, observability, governance for agents in production
Nice-to-haves: cost visibility per agent run
ICP confidence: Medium-High (415 emp Series C, actively shipping Retool Agents)
Signal: 4 (also 3) — VP of AI building Cube D3 agentic analytics and the semantic layer that grounds agents; speaks/writes on improving AI agent accuracy via a semantic layer.
Source: https://theorg.com/org/cube-dev/org-chart/david-jayatillake + https://cube.dev/blog/author/david-jayatillake
Company size: ~72 employees (Series B, $30M in 2022, ~$49M total; Databricks Ventures investor).
Pain points: agent accuracy/reliability grounded in enterprise data; correctness of agentic analytics.
Challenges: making agents produce trustworthy outputs at scale without hallucination.
Must-haves: reliability + data grounding for agents.
Nice-to-haves: cost efficiency of agent queries.
ICP confidence: Medium-High — VP of AI at a ~72-person Series B building an agentic product.
Signal: 1/4 (ICP identified directly via LinkedIn people search for senior agent-leader titles; content-post buckets 1-3 this run were dominated by non-ICP content creators). Source: https://www.linkedin.com/in/dloker/
Company size: ~50-150 (Series A/B AI code-review agent startup)
Pain points (inferred from role at agent-native code-review company): running AI code-review agents across large codebases; per-PR/token inference cost; output consistency & reliability at scale; model routing as frontier models shift.
Challenges: keeping review quality high while controlling inference cost per run; eval/observability of agent behavior.
Must-haves: cost-per-run visibility, reliable agent output, model flexibility.
Nice-to-haves: automated eval/regression tracking across agent versions.
ICP confidence: High (VP of AI, agent-native SaaS, headcount in range). CodeRabbit already an account in Brain; Loker is a new contact there.
David Noel Ng — Head of AI Product & Engineering, yoummday
Signal: Bucket 1 — ICP writing publicly about inference-time cost/efficiency (getting more capability out of a frozen model without extra compute or training).
Source: LinkedIn people search (Germany geo) "Head of AI agents production platform"; signal verified on https://www.linkedin.com/in/dnhkng/recent-activity/all/ — post dated 31 Aug 2026 on inference-path optimization, layer duplication and recirculation, citing Inference-Path Optimization (Rost, Apr 2026), MACRO (Batorski et al., Aug 2026) and Recirculation (Mozer et al., Aug 2026), and linking his own write-up "LLM Neuroanatomy: How I Topped the LLM Leaderboard Without Changing a Single Weight" (dnhkng.github.io).
Company size: 201-500 employees (LinkedIn company page, Munich). yoummday positions itself as "the Enterprise AI Execution Layer for Customer Experience" — technology-native contact centre combining software with human talent, i.e. voice/CX agents in production.
Pain points: inference cost and efficiency — extracting more quality per unit of compute from pretrained models rather than buying bigger ones; behaviour differs per model architecture ("each architecture has its own neuroanatomy"), so tuning does not transfer.
Challenges: running an enterprise AI execution layer for CX where cost per interaction is the unit economics; making model-level efficiency gains durable across model families as vendors ship new architectures.
Must-haves: measurable inference efficiency per model; a way to keep quality while cutting compute spend on high-volume CX traffic.
Nice-to-haves: techniques that generalise across architectures instead of per-model retuning.
ICP confidence: Medium — title (Head of AI Product & Engineering), company size (201-500) and segment (CX platform shipping agents) all match. Caveat: his public writing is about model/inference-level efficiency rather than explicitly about agent cost-per-run or agent observability, so the pain is adjacent to Alpha's wedge rather than a direct statement of it. Ex-Munich Re enterprise AI, Max Planck alumni — technically deep, likely to respond to evidence over marketing.
David Paffenholz — Co-Founder & CEO, Juicebox (juicebox.ai)
Signal: 4 (co-founder/decision-maker at qualifying company shipping always-on agents)
Source: https://www.businesswire.com/news/home/20260520017045/en/Juicebox-Launches-AI-Agents-That-Continuously-Source-Top-Talent-Across-Every-Open-Role ; https://techfundingnews.com/slug-juicebox-80m-series-b-recruiting-ai/
Company size: ~65 employees (Series B, $80M at $850M valuation; 5,000 customers)
Pain points: running AI sourcing agents 24/7 across every open role; reliability & efficiency of always-on agents at scale
Challenges: scaling continuous agents across 5,000 customers; agent efficiency/cost as usage grows
Must-haves: reliable always-on agent execution and monitoring at scale
Nice-to-haves: per-role / per-agent cost & performance visibility
ICP confidence: Medium-High (65 emp Series B, shipping Juicebox Agents in production)
David Scheier — Head of AI Innovation, NiCE Cognigy (NiCE Labs) — Agentic AI, LLM Orchestration, MCP & Agents, NiCE Cognigy
Signal: 4 (ICP senior AI leader at a company actively shipping AI agents). Source: https://www.linkedin.com/in/dshire (found via NiCE Cognigy company People directory, filter 'AI'). Company size: 201-500 employees (LinkedIn company page verified); Cognigy builds 'AI Agents for Your Contact Center'. Pain points (inferred from stated remit — agentic AI, LLM orchestration, MCP & agents): production reliability of multi-agent flows, LLM/token cost at scale, orchestration complexity across many agents. Challenges: keeping agent quality consistent in production; controlling per-conversation cost as agent volume grows. Must-haves: observability into agent cost + reliability per run; orchestration and eval tooling. Nice-to-haves: automated cost optimization; standardized MCP/agent tooling. ICP confidence: High (explicit agentic-AI / LLM-orchestration leadership role at a verified 201-500 agent company).
David Slater — Co-Founder & Chief Architect, Armadin
Signal: 4 (ICP senior technical leader at agent-native company). NOTE: Found via web research (funding/press + company sources); LinkedIn/Chrome NOT connected this run. Source: https://www.prnewswire.com/news-releases/armadin-secures-record-breaking-189-9m-in-seed-and-series-a-funding-to-combat-the-era-of-ai-driven-hyperattacks-302709318.html and https://www.securityweek.com/kevin-mandias-armadin-launches-with-189-9-million-in-funding/. Company size: ~60+ (hired 60+ in first 6 months). Stage: Series A ($189.9M combined Seed+Series A led by Accel). Role context: co-founder and Chief Architect (senior technical leadership over agent platform architecture); other co-founders are Kevin Mandia (CEO), Travis Lanham (CTO), Evan Peña (Chief Offensive Security Officer). Actively building AI agents: autonomous AI security agents that continuously scan for and respond to threats; Fortune 100 customers. Pain points (inferred, not verbatim): reliable orchestration/architecture of autonomous agent fleets in production; controlling agent behavior at scale; observability/cost visibility across many agents. Challenges: architecting trustworthy multi-agent systems for high-stakes security. Must-haves: reliability, control, observability. Nice-to-haves: cost-per-run visibility. ICP confidence: Medium (Chief Architect = senior technical leader; agent-native security co, ~60 emp in-band, Series A).
David Toomey — Chief Technology Officer & Head of Product Development, Pico
Signal: 1 (ICP quoted publicly on agent cost visibility / per-agent cost attribution).
Source: LinkedIn post by TechIntelPro (past week, surfaced via content search "AI agents cost CTO", sortBy latest) announcing Pico's selection of XNODE Cortx as its enterprise AI control plane; quote attributed to David Toomey, CTO at Pico. LinkedIn profile confirmed separately: https://www.linkedin.com/in/david-toomey/ (CTO & Head of Product Development, Ireland; mutual connections include Danny Moore, Pico founder).
Company size: ~500 employees (LinkedIn company page, verified this run). Pico = global financial-markets technology / trading infrastructure provider. NOTE: not a Series A–C SaaS/AI-native company — this is the one ICP dimension that does not fit cleanly.
Pain points (verbatim from the quoted statement): "scaling AI without visibility is a massive risk"; needs a "single, centralized pane of glass to detect shadow AI, attribute per-agent costs, and maintain a well-documented audit trail inside our own network."
Challenges: no per-agent cost attribution; shadow/unmonitored AI agents appearing across the org; audit-trail and governance requirements in a trust/security-first regulated financial-markets environment; wants controls in place BEFORE broad agent rollout.
Must-haves: per-agent cost attribution; centralized visibility across all agent activity; audit trail retained inside their own network (data residency / self-hosted control plane).
Nice-to-haves: shadow-AI detection, unified control-plane UX, policy-driven governance.
ICP confidence: Medium — role (CTO) and headcount (~500) fit exactly, and the expressed pain is a near-verbatim match for the agent cost/visibility thesis; downgraded from High because Pico is PE-backed fintech infrastructure rather than Series A–C SaaS/AI-native, and the quote came via a third-party post rather than his own account.
David Villalon — Co-founder & CEO (technical; ex-Chief AI Officer), Maisa AI
Signal: 1 (ICP writing publicly about agent cost blowout). Source: https://www.linkedin.com/posts/davidvillalonpardo_a-single-ai-agent-hitting-an-api-can-burn-activity-7430607114743562240-dAAA ("A single AI agent hitting an API can burn $100K+/year"); company: https://techcrunch.com/2025/08/27/maisa-ai-gets-25m-to-fix-enterprise-ais-95-failure-rate/ , https://pulse2.com/maisa-profile-david-villalon-interview/. Company size: ~89 employees (grew ~5x from 35; Spain/Valencia; $25M seed backed by NVIDIA/big-tech angels). Enterprise Agentic Process Automation (APA) platform building & managing AI "Digital Workers" with full traceability. Pain points: agent API/token cost blowout at scale ($100K+/yr per agent); enterprise AI's ~95% failure rate; making agents accountable and reliable enough for production. Challenges: reliability/traceability of autonomous digital workers in enterprise; cost economics as agent usage scales. Must-haves: production reliability + full traceability/accountability of agent actions; cost control per agent run. Nice-to-haves: cost visibility/attribution per digital worker. ICP confidence: High — technical co-founder/CEO and decision-maker at an ~89-person AI-native company shipping agents in production, publicly vocal on exactly Alpha's cost + control wedge.
David Zeng — Co-Founder & Head of Engineering, Eve (eve.legal)
Signal: 4 (ICP technical decision-maker building/shipping agents in production). Source: https://www.lawnext.com/2026/06/eve-builds-on-ai-workforce-launch-with-eveos-an-ai-native-operational-platform-for-plaintiff-firms.html ; https://tracxn.com/d/companies/eve/. Company size: ~294 employees (May 2026); Series B $103M (Sep 2025), $1B+ valuation; total raised ~$164M (Lightspeed, Menlo, a16z, Spark). Eve builds an AI workforce + EveOS, an AI-native operational platform for plaintiff law firms — case intake, evaluation, medical overviews, discovery, drafting; trusted by 1,200+ firms, processes 200,000+ legal cases/year, $3.5B+ recovered. Pain points (inferred from product scale): reliability/accuracy of autonomous legal agents at 200k+ cases/yr where errors carry legal/financial risk; per-case/per-run cost as agent volume scales; observability into agent actions across a firm-wide agent workforce. Challenges: moving from single-task assist to a trusted always-on agent workforce (EveOS) firms rely on unattended. Must-haves: production reliability + auditability of agent outputs. Nice-to-haves: cost-per-case visibility, portability across models. ICP confidence: High (clear technical co-founder/eng decision-maker; AI-native company squarely in 50–2,000 band shipping multiple agents in production).
David Zhao — Co-Founder & CTO, LiveKit
Signal: 1 (technical founder posting/writing about real-time agent infra reliability, observability and cost-per-session). Source: https://livekit.com/blog/livekits-series-b ; Series C coverage https://voice-ai-newsletter.krisp.ai/p/ai-agents-running-on-webrtc-russ ; LinkedIn posts on LiveKit agent platform. Company size: ~100-150 (Series C $100M at $1B val, Jan 2026, Index Ventures; powers OpenAI voice mode + ~25% of US 911). Pain points: production-grade reliability for real-time voice agents; autoscaling; turn-by-turn observability; opaque cost per call at scale. Challenges: operating agents 24/7 with failover, context migration, load balancing. Must-haves: production reliability + observability + cost visibility per session. Nice-to-haves: unified build/deploy/observe platform. ICP confidence: Medium-High (Series C AI-native shipping voice agents at massive scale; note infra-vendor adjacency to an agent operating layer).
Davis Liang — Head of Machine Learning, Abridge
Signal: 4 (ICP technical leader found via LinkedIn people search at a named agent company). Source: https://www.linkedin.com/search/results/people/?keywords=Abridge%20director%20of%20engineering%20AI . Profile: https://www.linkedin.com/in/liangdavis/ . Company size: ~300-500 (Series D/E clinical AI; ambient documentation moving to agentic workflows). Pain points (inferred from role+company, no direct quote): reliability/accuracy of clinical AI at scale; inference cost of high-volume model calls. Challenges: eval + guardrails for safety-critical clinical outputs; cost/latency as usage scales. Must-haves: production reliability + cost visibility for high-volume model/agent calls. Nice-to-haves: observability + regression testing across model versions. ICP confidence: Medium (leads ML at a confirmed 50-2,000-emp clinical-AI company shipping increasingly agentic workflows; pain inferred from role+company). Found by scheduled task icp-prospect-signal-scanner run 2026-08-03.
Dean Bloembergen — Co-Founder & CTO, Owner.com
Signal: 4 — CTO at a Series C vertical SaaS actively shipping multiple production agents ('AI Executives' — marketing, finance and technology assistants for independent restaurants).
Source: https://www.owner.com/leadership ; https://www.restaurantbusinessonline.com/technology/tech-supplier-ownercom-raises-120m-giving-it-1b-valuation
Company size: ~500 employees (Series C, $120M raise, $1B valuation)
Pain points: running multiple customer-facing AI agents across a large SMB base; agent reliability/quality at scale
Challenges: multi-agent orchestration and consistency across marketing/finance/tech assistants
Must-haves: reliability and oversight across an agent fleet
Nice-to-haves: per-agent cost attribution
ICP confidence: High — technical C-suite (CTO), Series C, ~500 employees, AI-native vertical SaaS shipping 5+ agents
Signal: 1 (ICP authoring content on agent context/reliability failure modes). Source: https://www.anvilogic.com/learn/the-anvilogic-approach-to-the-agentic-ai-soc (bylined by him). Company size: 115 (Crustdata, 30 Jun 2026; corroborated by Tracxn). Stage: Series C — $45M led by Evolution Equity Partners, $85M total (https://www.anvilogic.com/learn/series-c). Pain points (quoted from his own piece): context starvation — "security teams struggle not just with noise, but with context, specifically the contextual glue needed to connect signals, surfaces, and systems into coherent threat narratives." Production hallucination — "an LLM has no direct visibility into enterprise data models, detections, and enrichment... This limitation often manifests as hallucinations or overly generic responses." Context-window discipline — grounding "improves reasoning accuracy and reduces hallucinations by narrowing the model's attention to relevant operational details"; agents "extract the most meaningful slices of the security graph." 1-to-N scaling — agents chain to other agents, customer-customisable via MCP, with an "aggressive roadmap" across detection/hunt/triage. Challenges: keeping a growing fleet of chained, customer-customisable agents grounded and non-hallucinatory across heterogeneous customer data lakes. Must-haves: context/retrieval quality instrumentation; hallucination and grounding-failure detection in production; per-agent-step tracing across chained agents. Nice-to-haves: context-window efficiency metrics (token waste on irrelevant retrieval); per-customer/per-run cost attribution as the agent roadmap expands. ICP confidence: High — exact title, verified Series C and headcount, agents "battle-tested in production environments", authored the architecture piece himself. NOTE: no dollar/token cost mention — cost interest is inferred; context + reliability pain is verbatim.
Debajyoti (Debo) Datta — Co-Founder & Director, Hippocratic AI
Signal: 4 (ICP technical leader found via LinkedIn people search at a named agent company). Source: https://www.linkedin.com/search/results/people/?keywords=Hippocratic%20AI%20engineering%20director%20VP . Profile: https://www.linkedin.com/in/dedatta/ . Company size: ~250-500 (Series C healthcare voice-AI agents; ~$500M+ raised, valuation >$1.6B). Pain points (inferred from role+company, no direct quote): reliability & safety of patient-facing voice agents at scale; cost + latency across millions of clinical agent calls. Challenges: guardrails/eval for safety-critical healthcare agents; scaling a fleet of specialized agents reliably. Must-haves: production reliability + safety guardrails; per-call cost & latency visibility. Nice-to-haves: unified observability across the agent fleet; regression/shadow testing. ICP confidence: High (technical co-founder + Director building healthcare LLM agents at a 50-2,000-emp agent company). Found by scheduled task icp-prospect-signal-scanner run 2026-08-03.
Signal: 4 (ICP AI-focused product co-founder at a qualifying agent company). Source: targeted ICP search of coding-agent companies; profile https://www.linkedin.com/in/dedy-kredo/ (headline 'Co-Founder and CPO at Qodo'). Company size: ~115 (Qodo, agentic code review/testing/quality operating in IDE/PR/CI-CD; Series B $70M). Pain points [INFERRED, not a verbatim quote]: reliability/trust of code-review & test agents inside production dev workflows; controlling LLM cost across many repos/PRs; enforcing agent 'code integrity' guardrails. Challenges: agents operating across IDE/PR/CI contexts at scale; per-run cost/latency. Must-haves: dependable agent output + cost control. Nice-to-haves: agent observability across workflows. ICP confidence: Medium (co-founder CPO = AI-focused product leader per ICP, at a ~115-emp agent company).
Deepak Bala — Co-Founder & CTO, Rocketlane
Signal: 4 (ICP at a company that just bet its category positioning on shipping agents). NOTE: LinkedIn/Chrome unavailable this run — found via web research; LinkedIn URL surfaced in search results, not opened/verified.
Source: https://www.rocketlane.com/authors/deepak-bala ; https://www.rocketlane.com/nitro-agents ; https://www.prnewswire.com/news-releases/rocketlane-launches-nitro-the-industrys-first-agentic-execution-platform-for-professional-services-302702084.html ; https://siliconangle.com/2026/03/25/rocketlane-bags-60m-investors-accelerate-professional-services-automation-ai-agents/
Company size: 303 as of 2026-06-30 (Tracxn — dated, precise). Grew ~2.3x from 133 (Indian entity, Dec 2024). Series C — $60M led by Insight Partners, March 2026; $105M total. Chennai/US. 750+ customers, 17 Forbes Cloud 100.
Agents in production: CONFIRMED + independently verified this run. "Nitro" agentic execution platform, GA (live free-trial signup, not a waitlist). Individually named agents that EXECUTE work: Configuration Agent ("owns customer environment setup, end to end"), Migration Agent, Documentation Agent ("joins calls, reads emails"), Workforce Agent. Named Nitro customers via SiliconANGLE: Glean and Notion — "reduce manual delivery effort by up to 50% in some projects." Whole platform nav restructured around agents (finance/projects/resourcing agents).
Pain points: [evidenced, company-level] The Nitro page markets precisely the primitives an agent operating layer sells: "Every action follows defined rules. Validations before execution. Approvals where it matters... Permissioned data access. Full audit logs. Complete visibility into every action." [inferred, strong] Agents execute BILLABLE customer work, so cost-per-agent-run lands directly on gross margin — their own pitch is "healthier margins" and "3x more projects without adding headcount." That is a live budget line, not a hypothetical.
Challenges: Multi-agent orchestration across config/migration/docs/resourcing; per-action auditability at enterprise compliance bars (Intercom, Gong, nCino, Worldpay, Vercel as logos); keeping delivery-effort reduction claims true as agent count grows.
Must-haves: Cost-per-run attribution tied to project margin; full per-action audit trail and visibility (they already say this is table stakes); reliability gates before agents touch customer environments.
Nice-to-haves: Agent-level margin reporting per project/customer; benchmarking autonomy rates across delivery types.
ICP confidence: High — verified independently this run: title confirmed (rocketlane.com + Crunchbase), headcount confirmed with a dated source, agents confirmed GA with named enterprise customers, fresh $60M Series C. NOTE: no public first-person commentary from him on operating agents — approach on product/economics, not on a quote.
Signal: 4 (ICP building/shipping AI agents in production). Source: https://www.tabs.com/blog/tabs-raises-55M-series-b-to-launch-the-first-ai-agents-for-billing-and-collections ; https://www.businesswire.com/news/home/20250915745370/en/ . Company size: ~192 employees (Series B $55M led by Lightspeed, Sept 2025; $91M total; founded 2023). Ships AI agents for billing, collections, revenue recognition & reporting (contract-to-cash) in the CFO's office; 200+ customers incl. Cursor & Statsig, automating $500M+ annual invoice volume. Pain points (INFERRED): reliability of collections/billing agents touching customer money; per-run/per-agent cost visibility as invoice volume scales; controlling agent behavior across many customer contracts. Challenges (INFERRED): scaling from a few workflows to a full agent fleet without cost/error blowup. Must-haves (INFERRED): per-run cost visibility + reliable, auditable agent actions. Nice-to-haves (INFERRED): unified observability across billing/collection agents. ICP confidence: High (technical co-founder/CTO ex-Latch Director of Software, Series B, ~192 emp in-band, agents in production at named enterprises). LinkedIn URL not surfaced — not fabricated.
Deepak Dutta — General Manager & Group VP (Business Agent AI), Uniphore
Signal: 1 (ICP posting/engaging on agent cost & value in the past month). Source: LinkedIn content search 'agentic AI cost control observability' / 'our AI agents production cost reliability' (past-month) + profile https://www.linkedin.com/in/deepakdutta1/. Company size: ~1,000-1,500 (Uniphore, enterprise conversational/agentic 'Business AI'; has acquired Orby AI, Autonom8, ActionIQ, Infoworks to build an agentic stack). Pain points (from his posts + role): value-per-dollar of AI as agents scale into production; agent spend outpacing business value; unifying many agents into a single 'Business Agent AI' fabric across CX/HR/Sales/Finance; outcome-based observability. Challenges: governing a multi-product agent stack, integrating acquired agent tech, executing agents across legacy systems. Must-haves: per-outcome cost/value visibility; orchestration & control across many agents. Nice-to-haves: cross-department agent governance. ICP confidence: Medium (VP/GM-level leader championing agentic AI at a 1,000+ emp agent company; role is GM/GVP rather than pure engineering, hence Medium).
Deepank Sharma — Field CTO, Cresta
Signal: 4 (senior applied-AI / Field CTO at an agent-native company; found via LinkedIn people search of Cresta).
Source: https://www.linkedin.com/search/results/people/?keywords=Cresta%20AI%20head%20of%20engineering%20director%20agents
Company size: ~500-700 employees; Cresta, Series D, contact-center AI agents. In-band on size; stage past Series C noted. ex-Ford, ex-Verizon.
Pain points (inferred): translating agent deployments into reliable, cost-controlled production for enterprise CX customers.
Challenges: agent reliability + cost transparency at enterprise scale.
Must-haves: reliability, cost visibility per run.
Nice-to-haves: governance/observability.
ICP confidence: Medium — Field CTO (senior technical, customer-facing) at agent-native company, in-band size.
Signal: 4 (Co-founder & CTO at a qualifying company building agentic platform in production)
Source: https://www.axios.com/2026/06/18/gradial-ai-agents-marketing | https://www.gradial.com/blog/gradial-65m-series-c
Company size: ~85 employees (Series C $65M led by Insight Partners, $675M valuation, $120M+ total)
Pain points (inferred): agents execute enterprise marketing work across dozens of tools (Adobe, Salesforce, ServiceNow, Databricks) — orchestration & reliability across many integrations; "system of work" for enterprise marketing implies long multi-tool agent runs.
Challenges: reliability of long, multi-tool agent workflows in production; enterprise-grade trust and control; scaling agentic orchestration layer.
Must-haves: reliable multi-step agent orchestration; control/guardrails over agent actions in real systems.
Nice-to-haves: cost visibility per workflow/tenant.
ICP confidence: High (Co-founder & CTO, ~85 emp, agentic platform shipping in production).
Signal: 4 (ICP C-suite technical leader at agent-native company; surfaced via Aisera leadership-appointments research). Source: https://aisera.com/blog/accelerating-agentic-ai-strategy/ ; https://theorg.com/org/aisera/org-chart/dan-deklich . Company size: ~250-500 (Aisera, agentic AI enterprise, Series D). Pain points (inferred, not quotes): scaling engineering + customer success for a fleet of production agents; reliability and cost control as agent usage grows; visibility into agent cost/quality per run. Challenges: owning end-to-end SDLC for agentic products at enterprise scale. Must-haves: production reliability, per-run cost visibility, 1->many agent scaling. Nice-to-haves: eval/observability tooling, spend optimization. ICP confidence: High (CDO overseeing software engineering at an in-range agent-native company; verified current).
Denis Leroy — Engineering Director, SoundHound AI
Signal: 4 (engineering leader at agent-native company). Source: https://www.linkedin.com/in/denis-leroy/ (LinkedIn people search "SoundHound AI director engineering agents"). Company size: ~750-950 (SoundHound AI). Ships agents: YES. Pain points (inferred from role): production reliability + cost of agent workloads. Challenges: scaling agents reliably. Must-haves: reliability + cost control. Nice-to-haves: agent observability. ICP confidence: Medium (title + company confirmed; found via company-scoped search).
Signal: 1 & 3 (CTO publicly speaking about agent token/context cost; moved internal systems away from MCP)
Source: https://www.agent-engineering.dev/article/why-perplexity-is-stepping-back-from-the-model-context-protocol-mcp-internally ; https://www.startuphub.ai/ai-news/artificial-intelligence/2026/perplexity-cto-on-gpt-5-5-efficiency
Company size: AI-native, est. several hundred employees (within 50-2000); caveat: frontier-adjacent
Pain points: context/token waste from agent tool schemas; high per-interaction token cost; latency of multi-tool agent calls
Challenges: controlling token consumption & context bloat as agents scale; cost/latency at scale
Must-haves: token/context efficiency; cost & latency control for agent interactions
Nice-to-haves: efficient multi-model routing/orchestration
ICP confidence: Medium (explicit on-ICP pain signal; caveat: large AI-native co may build in-house)
Signal: 4 (ICP writing about building/shipping agents in production). Source: https://www.linkedin.com/posts/dennis-cui-70a06340_at-decagon-weve-taken-a-radically-different-activity-7316192886675369984-Uyg2 (AOPs — "smarter, faster, secure AI agents") and Decagon Labs post activity-7442978285442981888. Ex-CTO/VP Eng at Cobalt Robotics. Company size: ~210-446 employees (2026), Series C, ~$250M raised, ~$4.5B valuation; builds conversational AI agents for enterprise CX resolving millions of inquiries across chat/email/voice (clearly 5+ agents in production). Pain points: making production agents faster, cheaper, and secure at enterprise scale; continuous improvement of agents in production. Challenges: scaling reliable custom agents across many enterprise customers, agent observability. Must-haves: production reliability + security guardrails, agent performance/cost visibility at scale. Nice-to-haves: self-improving agent tooling. ICP confidence: High (VP Engineering, 50-2000 emp, AI-native agent company shipping at scale).
Dennis Thompson — Senior Director, Software Engineering, Writer
Signal: 4 (senior technical leader at an agent-shipping company, found via LinkedIn currentCompany-facet People search). Source: https://www.linkedin.com/search/results/people/?currentCompany=%5B%2267088679%22%5D (WRITER). Company size: 201-500 (LinkedIn). Pain points (inferred from role + company stage): per-agent/per-run LLM cost visibility as Writer scales its enterprise agentic platform; reliability of non-deterministic agents in production; token/context waste. Challenges: scaling eng org while controlling aggregate model spend. Must-haves: cost-per-run observability + guardrails on runaway agents. Nice-to-haves: model routing / cost attribution by team. ICP confidence: High (clear technical Sr Director of Software Engineering at Writer, an enterprise AI-agent platform; 2nd-degree connection; title/company/profile verified on LinkedIn this run). Pain points INFERRED from role+stage, not fabricated quotes.
Denys Linkov — Head of ML, Voiceflow
Signal: 4 — leads ML / Enterprise AI at Voiceflow (AI agent building platform); frequent speaker (QCon SF, Toronto ML Summit, MLOps community) on putting LLM agents into production, agent behavior control, cost management, and reliability.
Source: https://www.voiceflow.com/contributors/denys-linkov + MLOps/QCon talks
Company size: 88 employees ($39.8M raised; ~$10.9M ARR 2026).
Pain points: agent behavior control, latency vs. response quality, cost management, memory management in production.
Challenges: hardening agents for enterprise-scale production reliability.
Must-haves: control + observability + cost management for agents.
Nice-to-haves: better memory management and performance benchmarking.
ICP confidence: Medium — Head of ML at an 88-person agent platform. Note: vendor-adjacent (agent-building platform).
Deon Nicholas — Co-Founder & Executive Chairman, Forethought
Signal: 1 (ICP writing/speaking about autonomous agent reliability and resolution rates). Source: https://www.unite.ai/deon-nicholas-co-founder-ceo-of-forethought-interview-series/ ; https://forethought.ai/team-member/deon-nicholas ; https://www.linkedin.com/in/deon-nicholas/ . Company size: ~100-150 (venture-backed agentic CX platform). Pain points: making support agents act autonomously and reliably (Autoflows reach 70-80% resolution vs 10-20% for RAG); integration blockers solved via Browser Agent (agent uses UIs without APIs); value gap between agents trained on proprietary vs generic data. Challenges: autonomous action-taking reliability; measuring/improving resolution KPIs; legacy-system integration. Must-haves: high autonomous resolution rate, reliability, action-taking (not just answers). Nice-to-haves: no-API/browser-based integration. ICP confidence: Medium (technical co-founder, ex-Palantir/Dropbox; now Executive Chairman rather than operational CEO and also running new venture Espa Labs — decision-maker influence at Forethought is partial).
Derek Ho — Co-Founder & COO (leads Engineering & Platform; ex-Palantir, Citadel), Distyl AI
Signal: 4 (ICP technical co-founder leading engineering/platform at agent-native company). NOTE: Found via web research (company/exec profiles + funding data); LinkedIn/Chrome NOT connected this run. Source: https://joinplank.com/research/distyl-ai and https://getlatka.com/companies/distyl.ai and https://www.channeldive.com/news/billion-dollar-ai-startup-distyl-ai-openai-azure-anthropic/802806/. Company size: ~106-159 (PitchBook ~106; Tracxn 159 as of May 2026). Stage: Series B ($202M+ raised; ~$1.8B valuation Sept 2025; ~$31M ARR reported). Role context: two-pillar founder structure — CEO Arjun Prakash leads field/strategy; Derek Ho (COO, ex-Palantir/Citadel) leads engineering, platform, and operations. Actively building AI agents: proprietary 'Distillery' agentic platform delivering production AI systems inside Fortune 500 within ~3 months on outcome-based, multi-year contracts; forward-deployed engineers. Pain points (inferred, not verbatim): reliability/accuracy of agents on enterprise-critical workflows; scaling agentic deployments across F500 with outcome-based SLAs; production-readiness speed. Challenges: governing/monitoring agents in regulated enterprise settings; orchestrating multi-agent workflows. Must-haves: reliability, production-readiness, governance. Nice-to-haves: cost-per-run visibility. ICP confidence: High (technical co-founder leading engineering & platform; Series B; ~150 emp in-band; agent-native, agents in production).
Signal: 4 (ICP CTO at company shipping conversational AI agents in production). Source: https://theorg.com/org/fabric-1/org-chart/derek-rockwell ; https://www.fabrichealth.com/press/fabric-secures-60m-series ; https://www.fiercehealthcare.com/ai-and-machine-learning/two-acquisitions-under-its-belt-and-80m-investors-fabric-building-tech-fix . Company size: est. ~150-300 employees (multiple acquisitions: Zipnosis, TeamHealth VirtualCare, conversational-AI startup). Series A ($60M led by General Catalyst; Thrive, GV, Salesforce Ventures; $80M total). Care-enablement platform with conversational-AI intake/triage/routing + AI assistant across 70 health systems, 3,800+ clinicians. CTO since Apr-2023 (ex-Zipnosis CTO/VP Eng/CISO). Pain points (INFERRED): reliability & clinical safety of patient-facing conversational agents; HIPAA auditability/control; cost & observability of agent runs across many health systems. Challenges: scaling reliable, compliant agents in a regulated, high-stakes setting. Must-haves: reliability + auditability/control; cost visibility per run. Nice-to-haves: compounding quality; guardrails. ICP confidence: MEDIUM-HIGH (CTO at Series A healthcare-AI company shipping agents in production; regulated/high-stakes).
Deven Panchal — Director of Engineering, Uniphore
Signal: 4 (ICP senior technical leader at an agent-native company; discovered via LinkedIn people-search of companies actively shipping AI agents) | Source: https://www.linkedin.com/in/devenpanchal | Headline: Director of Engineering @ Uniphore (enterprise conversational + agentic AI) | Company: Uniphore — actively building/shipping AI agents in production | Company size: ~800–1,000 (estimate) | Pain points (INFERRED from role/company context — not a verbatim quote observed this run): cost per conversation/agent run; reliability at enterprise scale; observability across deployed agents | Challenges: scaling deployed agents across enterprise customers cost-effectively and reliably | Must-haves: cost-per-run visibility; production reliability | Nice-to-haves: token optimization; cross-deployment observability | ICP confidence: High (senior technical/eng leadership at a 50–2,000-employee company shipping AI agents)
Signal: 1 (ICP writing about agent cost/control/reliability). Source: https://www.linkedin.com/in/dpandey/ (LinkedIn posts re: DevRev MetaHarness). Company size: ~940 employees (unicorn, ~$186M raised, Series A, $1.15B val). Technical co-founder (ex-Nutanix CEO), so a technical decision-maker per ICP. DevRev builds and ships AI agents (support/dev "MetaHarness" — repo-specific agent harnesses with task rules, tool policies, scoring, receipts, routing, verifiers, promotion gates). Pain points: making agents reliable, governable, repeatable, measurable, AND cheap in production; agents that stall past the demo. Challenges: enforcing agent behavior + cost across a large org and customer base; verification/promotion gates. Must-haves: reliability, governance, measurability, per-run cost control. Nice-to-haves: repeatability, automated verifiers/promotion gates. ICP confidence: High — explicit public framing around agent reliability+cost+measurability; company squarely in 50-2,000 band and shipping multiple production agents.
Signal: 4 (AI-focused VP of Product at an agentic-AI-native company; found via LinkedIn people search + web verification).
Source: https://www.linkedin.com/in/dhruvdhingra/ ; Fieldguide Series C coverage (SiliconANGLE/Yahoo Finance, Feb 2026).
Company: Fieldguide — agentic AI-native platform for audit & advisory firms; used by ~half of Top 100 US accounting firms (Baker Tilly, BDO, Grant Thornton, KPMG, RSM US).
Company size: 101-250 employees (Unify insights 2026).
Stage: Series C ($75M Feb 2026, $125M total, $700M valuation).
Pain points (INFERRED from role + company, NOT verbatim): shipping/scaling agentic workflows into enterprise audit; reliability & accuracy of agents on high-stakes documents; cost per agent run across large firm deployments.
Challenges (inferred): productizing multi-step audit agents at enterprise reliability.
Must-haves (inferred): reliability, auditability/observability of agent runs, cost control.
Nice-to-haves (inferred): eval/QA tooling, context/token optimization.
ICP confidence: High (VP Product, AI-focused, at 50-2,000 emp Series C agent-native company).
Dhruv Mahajan — Chief AI Scientist, Resolve AI (resolve.ai)
Signal: 4 (ICP senior technical AI leader at agent-native company shipping in production; found via web research pivot — LinkedIn search was blocked this run by a login wall, so used agent-startup web research + verification as in prior successful runs). Source: https://resolve.ai/news/Series-A-extension-and-Resolve-AI-Labs ; https://resolve.ai/news/resolveai-raises-125-million-series-a . Company size: 161 employees (Jun 2026). Stage/funding: Series A $125M + $40M extension at $1.5B valuation (Lightspeed, DST, Salesforce Ventures); >$190M total. Actively building/shipping agents: autonomous Site Reliability Engineer (SRE) agents that join on-call rotations, investigate production incidents, find root cause, and execute remediations. Background: joined from Meta where he led post-training for large-scale Llama foundation models; leads Resolve AI Labs. NOTE: this is a DIFFERENT person from existing Resolve AI records (Mayank Agarwal CTO id 129, Spiros Xanthos CEO id 130). Pain points (role-inferred, not verbatim): economics of continuously-running agents at enterprise scale; agent reliability for autonomous incident resolution; per-run inference cost/token spend as agents run 24/7. Challenges: trustworthy autonomous action in high-stakes production; controlling cost of always-on agents. Must-haves: reliable + cost-efficient agent operation at scale; observability into agent actions/cost. Nice-to-haves: inference cost optimization / model routing. ICP confidence: High (Chief AI Scientist / Head-of-AI-level technical leader at 161-emp AI-native company whose product is production agents).
Signal: 4 (ICP CTO at a company shipping healthcare AI agents). Found via web research + verification — LinkedIn content/people search UNAVAILABLE this run (session not authenticated / login wall). Source: https://www.commure.com/agents ; https://www.fiercehealthcare.com/ai-and-machine-learning/fresh-200m-raise-commure-rolls-out-ai-agents (Commure rolls out AI agents) ; https://www.linkedin.com/in/dhruv-parthasarathy-0b29b658/. Company size: ~1,200-1,566 employees per Tracxn/PitchBook/SV Investclub (2026). SIZE CAVEAT: one source (Revelio Labs) reports ~2,373, at/above the 2,000 upper bound; majority of sources place it in-band. $7B valuation, General Catalyst-backed; $70M raise May 2026. Actively shipping agents: launched Commure Agents — AI "colleagues" that automate physician/front-office workflows, patient outreach and billing; 500+ health orgs, 500K+ clinicians, $25B+ annual claims. Dhruv Parthasarathy is CTO of the combined Commure/Athelas (ex-CTO Athelas; ex-Director of AI at Udacity). Pain points (INFERRED): reliability of agents automating clinical/billing workflows; per-run cost/observability at very high interaction volume; governance across a broad agent portfolio. Challenges: scaling reliable, cost-efficient, governable agents across a large customer base. Must-haves: reliability, control/governance, per-run cost visibility. Nice-to-haves: observability + cost optimization. ICP confidence: High on role/agent-fit; Medium overall due to size ambiguity (some sources put headcount >2,000).
Diego Baugh — Co-founder & Chief Product Officer (AI-focused), Tennr
Signal: 4 (AI-focused product co-founder at a company shipping AI agents in production; web research this run). Source: https://techfundingnews.com/top-10-us-ai-agents-2026-fastest-scaling-category-52b-by-2030/ ; https://theorg.com/org/tennr/org-chart/trey-holterman ; https://www.linkedin.com/in/diego-baugh/ . Company size: 205 employees (NYC; Series C $101M at $605M valuation). Tennr ships RaeLM vision-language agents for healthcare referral/authorization documentation in production. Pain points (inferred; no verbatim quote this run): product reliability at volume, agent accuracy on edge-case clinical scenarios, measurable outcomes (approval rates). Challenges: owning the product roadmap for agents that must be reliable/auditable in a regulated domain. Must-haves: reliability, accuracy, outcome measurement. Nice-to-haves: cost/latency visibility per workflow. ICP confidence: Medium (co-founder & CPO = AI-focused product leader, explicitly in ICP as 'VP of Product (AI-focused)'; 205-emp Series C agent company). Verified LinkedIn slug captured from search result.
Signal: 1 (ICP speaking publicly about agent reliability in production — named in an Anthropic customer case study).
Source: https://claude.com/customers/vambe
Company size: ~80+ employees (grew from 22 to 80+ per Series A press coverage). Vambe, Santiago, Chile. Series A, $14M, Dec 2025, led by Monashees. WhatsApp-first AI agents automating the full commercial cycle (sales, retention, support) for B2C companies.
Pain points: (1) unreliable function/tool calling — "When an AI model misunderstands a function call or executes it incorrectly, it doesn't just break the conversation—it costs the customer money or damages their reputation"; (2) large reliability variance between models on action-taking (not just conversational) agents; (3) agents sit on a revenue-critical path, so failures are financially material.
Quote: "The main difference was function calling reliability. Vambe's platform needs to execute real business actions."
Challenges: executing agentic actions reliably at scale in a revenue path; choosing and continuously re-validating which model is most reliable per task.
Must-haves: high, measurable function-calling reliability; production evaluation of action-taking agents; fast detection when an agent starts executing incorrectly.
Nice-to-haves: cross-model comparison tooling to route each task to the most reliable model.
ICP confidence: High — CTO and technical co-founder, headcount and Series A stage both squarely in range, AI-native agent company, and a specific first-person reliability pain quote. LATAM geography adds coverage the library is thin on.
Verification note: headcount is from Series A press coverage, not a first-party source. LinkedIn URL was seen directly in search results.
Diego Comas — Senior Director of Engineering, Sourcegraph
Signal: 4 (ICP eng leader at an agent-shipping company; LinkedIn people-search 'Sourcegraph engineering director agents'). Source: https://www.linkedin.com/in/diegocomas/ . Company size: ~200-400 (Sourcegraph; code-intelligence platform shipping the Amp + Cody coding agents to enterprise dev teams). Profile headline: 'Senior Director of Engineering — leading Security, Platform & Services; building scalable foundations for AI & enterprise products.' Pain points (INFERRED from role + company; no verbatim quote observed this run): reliability & governance of coding agents in production; platform scalability and cost control as agent usage grows; security of autonomous coding agents. Challenges: making agent-primary developer workflows reliable and safe at enterprise scale. Must-haves: reliability guardrails + governance for coding agents. Nice-to-haves: per-run cost/observability. ICP confidence: High (Senior Director of Engineering at a 50-2,000-emp devtools company actively shipping coding agents). Note: brain already has Sourcegraph CTO Beyang Liu — Comas is net-new.
Diego S. Burgos — Chief Technology Officer (Founding Team Member), Pomelo
Signal: 4 (ICP publicly presenting on shipping agents). Found via web research; title VERIFIED live on LinkedIn people search 2026-08-31 — headline reads "CTO @ Pomelo | Fintech Infrastructure | AI-native engineering | Building from Latin America", Greater Buenos Aires.
Source: https://www.tipranks.com/news/private-companies/pomelo-highlights-agentic-ai-architecture-at-ignite-evolution-2026
Company size: ~386-399 (multiple aggregators, Mar 2026). Series C. Card-issuing / fintech infrastructure, Argentina. Squarely in the 50-2,000 band and the Series A-C stage band.
Pain points: Cost scaling with volume. Pomelo publicly described an in-development agentic system explicitly intended to "support higher transaction volumes and client growth without a linear rise in operating costs" — i.e. they are already framing agents as a unit-economics problem.
Challenges: Making agentic architecture hold up under regulated, high-volume card processing; his own LinkedIn positioning is "AI-native engineering," suggesting an active org-wide agent build-out.
Must-haves: Decoupling agent operating cost from transaction volume; reliability under regulated fintech load.
Nice-to-haves: Per-client / per-workflow cost attribution.
ICP confidence: Medium — right title, right size, right stage, and an explicit cost-scaling thesis at company level. Downgraded from High because the on-stage remarks at Ignite Evolution 2026 were delivered by Tech Ops Lead Ivan Raitman, not by Burgos personally; no first-person quote from Burgos located yet.
Signal: 4 — ICP-title technical AI leader at an agent-shipping company (found via people search: 'Head of Engineering agentic AI'). Source: https://www.linkedin.com/search/results/people/?keywords=Head%20of%20Engineering%20agentic%20AI%20startup . Company size: ~50-150 (est). Satisfi Labs = conversational-AI / answer-engine agents for sports & entertainment brands. Pain points (INFERRED): running conversational agents in production reliably & cost-efficiently across many brand deployments. Challenges: scaling agent deployments across clients while controlling cost. Must-haves: production observability + cost control per agent. ICP confidence: Medium — Head of AI Eng at in-range agent-native company; pains inferred from role.
Dimitri Masin — Co-founder & CEO (technical; ex-Monzo, led Data/ML), Gradient Labs
Signal: 4 (ICP building/shipping specialist agents in production; strong cost/reliability pain match). Source: https://gradient-labs.ai/about; funding coverage (Sifted, Crowdfund Insider) June 2026. Company size: "more than 40 employees" per reporting — BORDERLINE at the 50-employee floor; fast-growing (revenue ~10x YoY, 32M+ end users, ~$42.6M raised, Series A + extension, Redpoint/Octopus/CommerzVentures). Product: domain-specific multi-agent system for financial services (lending, disputes, KYB, fraud, claims) with strict compliance; agents share context/memory/tasks. Pain points: reliability vs. latency tradeoff (median 15–20s "thinking" time chosen because a wrong answer is worse than a slow one); cost accountability — prices at ~30% of equivalent human cost and only charges on successful resolution. Challenges: maintaining strict compliance + accuracy across long-running multi-agent workflows in a regulated domain. Must-haves: reliability/accuracy guarantees, cost-per-resolution economics, compliance/audit. Nice-to-haves: shared context/memory across agent fleet. ICP confidence: Medium — excellent pain-signal fit; flag headcount is right at/near the 50 floor (verify before prioritizing).
Dimitrios Chavouzis — Head of AI Engagements — Agents, Evals & RL Environments, Pareto AI
Signal: 4 (ICP whose stated remit is agents + evals + RL environments; posts on agent evaluation/safety findings). Source: https://www.linkedin.com/in/dimitrios-chavouzis/ — headline "Head of AI Engagements | Agents, Evals & RL Envs @ Pareto.ai"; amplified an AI Security Institute study showing safety prompting had no measurable effect on harmful-advice rates (i.e. prompt-level controls don't hold at run time). Found via LinkedIn people search "Head of AI agents evals production SaaS" (2026-08-29). Company size: ~540–573 employees (Revelio Apr 2026 540; ~573 Mar 2026), AI-native (human-data + RL environments for agent training/eval), ~$14–18M raised. Pain points: evaluating agent behaviour at scale — prompt-level guardrails demonstrably fail, so evaluation has to happen on real runs. Challenges: running large volumes of agent/RL-environment executions across a distributed workforce. Must-haves: run-level evaluation and evidence rather than prompt-level assurances. Nice-to-haves: cost/throughput visibility per evaluation run. ICP confidence: Medium — Head-of level, agents/evals remit, correct headcount band; his public signal is evaluation and safety rather than cost, and Pareto is a data/RL-env vendor rather than a classic agent-shipping SaaS.
Signal: 4 (ICP product co-founder at agent-native company shipping agents in production; found via web research — LinkedIn search blocked this run). Source: https://techcrunch.com/2026/03/12/sales-automation-startup-rox-ai-hits-1-2b-valuation-sources-say/ ; https://sequoiacap.com/article/partnering-with-rox-every-seller-needs-an-agent-swarm/ ; https://theorg.com/org/rox . Company size: ~109 employees. Stage/funding: Seed+Series A ($50M+; Sequoia, General Catalyst, GV); ~$1.2B valuation (2026). Actively building/shipping agents: Rox deploys "agent swarms" — hundreds of AI sales agents that monitor accounts, research prospects, and update CRM (built on Amazon Bedrock + Perplexity API). Role: co-founder & Product Lead = head of product / VP Product (AI-focused). NOTE: DIFFERENT person from existing Rox records (Avanika Narayan; Shriram Sridharan) — net-new. Pain points (role-inferred): reliability/quality of large fleets ("swarms") of agents in production; cost per agent run across hundreds of concurrent agents; visibility into what each agent costs and produces. Challenges: scaling from few to hundreds of agents while controlling cost & quality; product-level observability. Must-haves: per-run cost visibility, agent reliability at swarm scale. Nice-to-haves: cost optimization across the swarm. ICP confidence: Medium (product co-founder rather than pure engineering; strong company fit — AI-native, right size, agents in production; title maps to VP Product AI-focused).
Dion Almaer — Chief Technology Officer, Augment Code
Signal: 1/4 — Public writing (Medium 'Ben and Dion' / Augment blog) on building autonomous coding agents (Auggie) that run for days; enterprise Context Engine indexing 400K+ files.
Source: https://www.augmentcode.com/blog/augment-inc-raises-227-million ; https://www.linkedin.com/in/dalmaer/
Company size: ~188 (Series B, ~$1.4B val)
Pain points: reliability of long-running autonomous coding agents; context/token management at 400–500K file scale; cost of extended agent runs
Challenges: keeping agents reliable across days-long tasks in large enterprise codebases
Must-haves: context efficiency, agent reliability, cost/token visibility per run
Nice-to-haves: multi-repo orchestration
ICP confidence: High — CTO, Series B, coding agents in production at NVIDIA/EY/Adobe/Palo Alto Networks
Signal: 2 (non-ICP TechInfluencer Evan Kirstel LinkedIn Live/video post featuring Doug Marquis, CTO of Zywave, discussing shipping agentic AI in regulated insurance workflows).
Source: https://www.linkedin.com/search/results/content/?keywords=shipping%20AI%20agents%20production%20CTO (Evan Kirstel post, ~3 days ago, 2026-07 LinkedIn content search).
Company size: ~893-977 employees (insurtech SaaS, Milwaukee).
Pain points: shipping agentic AI reliably in regulated insurance workflows; explicitly named the "AI harness" needed around agents — observability, testing, COST MANAGEMENT, and explainability; moving from pilots to production.
Challenges: encoding domain knowledge/producer workflows into artifacts agents can use; compliance, security, licensing and data-access controls.
Must-haves: agent observability, testing, cost management, explainability.
Nice-to-haves: insurance MCP server integration for specialized contextual guidance.
ICP confidence: High — CTO at ~950-emp SaaS actively shipping agentic AI; directly voiced cost-management + observability pain, which is the core ICP pain signal.
Douwe Kiela — Co-founder & CEO (technical co-founder; pioneered RAG; Stanford adjunct professor), Contextual AI
Signal: 1 (ICP technical co-founder writing/speaking about deploying agents in production, context engineering and cost; found via web research — LinkedIn/Chrome not connected this run). Source: https://venturebeat.com/technology/contextual-ai-launches-agent-composer-to-turn-enterprise-rag-into-production ; https://contextual.ai/blog/platform-ga-press-release ; https://www.datacamp.com/podcast/rag-2-and-the-new-era-of-rag-agents (talk: "10 Lessons for Deploying RAG Agents in Production"). Company size: ~93 employees; Series A ($80M, total ~$100M raised; Greycroft, Bain Capital Ventures, Lightspeed); enterprise platform for building specialized RAG agents (Agent Composer turns enterprise RAG into production-ready agents), customers incl. HSBC and Qualcomm. Pain points: getting RAG agents to production-grade reliability/accuracy; context engineering and token/context waste; grounding/attribution for expert knowledge work. Challenges: reliability and explainability of agentic RAG at enterprise scale; controlling cost per run while keeping accuracy high. Must-haves: reliability/accuracy, context efficiency, governance. Nice-to-haves: cost-per-run visibility, model-level efficiency. ICP confidence: High (deeply technical co-founder — ex-Head of Research at Hugging Face / FAIR, creator of RAG — at a ~93-emp AI-native company shipping enterprise agents; note title is CEO but fits the 'technical co-founder' ICP; CTO Amanpreet Singh already in the People Library, this is a distinct net-new person).
Dr. Allen Badeau — Chief AI Officer, DigitalNet.ai
Signal: 4 (senior technical AI leader at a qualifying company actively shipping agents). Source: https://www.linkedin.com/in/allenbadeau/ (found via LinkedIn people search on DigitalNet.ai). Company size: ~1,100-1,150 employees incl. 300+ AI/data specialists (DigitalNet.ai, Bethesda MD; formed from Harmonic AI, Zillion Technologies, Axis Group; ~$200M in acquisitions). Confirmed 50-2,000 band and running 2,000+ trained AI agents on 50+ specialized models across fraud detection, cybersecurity, government service optimization, pharma and consumer domains (source: company profile / press). Role: top technical AI decision-maker (CAIO) — owns the agent platform strategy. Pain points (inferred from role/company, no direct quote this run): governing and observing 2,000+ production agents; per-run/per-agent cost visibility at fleet scale; reliability and compliance for regulated/government agent workloads. Challenges: scaling an enterprise agent fleet reliably and cost-efficiently across many verticals. Must-haves: fleet-level observability, cost control per agent/run, reliability & governance. Nice-to-haves: cross-vertical agent reuse, model right-sizing. ICP confidence: High (Chief AI Officer / senior technical AI leader at an in-range company operating one of the largest named production agent fleets; distinct from Vikas Salaria at same company). profile_url captured here in Source in case API drops it.
Dr. David Noel Ng — Head of AI Product & Engineering, yoummday
Signal: 4 (ICP writing/working on shipping agents — surfaced via LinkedIn PEOPLE search, not content search).
Source: https://www.linkedin.com/search/results/people/?keywords=Head%20of%20AI%20agents%20in%20production%20cost ; profile + activity at https://www.linkedin.com/in/dnhkng/recent-activity/all/ ; company page https://www.linkedin.com/company/yoummday/
Company size: 201-500 employees (verified on yoummday's LinkedIn company page, 28 Aug 2026). HQ Munich, Germany. Positions itself as "the Enterprise AI Execution Layer for Customer Operations" / "the first technology-native contact center" with "the world's smartest, auto-tuning AI" plus multilingual human talent — i.e. AI agents running against live customer interactions, outcome-priced.
Pain points: NOT DIRECTLY EXPRESSED — do not attribute quotes to him. His public LinkedIn activity is technical/research in nature (e.g. a study feeding 64 semantically matched sentences across 8 languages through four architectures and extracting hidden-state representations per layer). Inferred, unverified pain from role + company model: yoummday sells outcome-based CX, so per-interaction agent cost directly compresses gross margin; multilingual auto-tuning agents imply repeated model routing and eval spend.
Challenges: owns both AI product AND engineering at a 200-500 person company — thin org, wide scope. Enterprise CX buyers demand reliability/auditability that agent stacks do not give out of the box.
Must-haves (hypothesis to test in discovery, not stated): per-interaction / per-agent-run cost attribution, since pricing is tied to outcomes rather than seats.
Nice-to-haves (hypothesis): model routing and eval tooling across languages.
ICP confidence: Medium-High — title, level and company size are a clean fit and the company is demonstrably an agent operator (not a vendor of agent infra); downgraded from High only because no pain signal was captured in his own words. NET-NEW company for the People Library — yoummday returned zero matches on dedup.
Verification notes: headcount read directly off LinkedIn (201-500 band). Profile URL confirmed by DOM href extraction, not guessed.
Dr. Sanjay Saini — Director of AI, TNS (Transaction Network Services)
Signal: LinkedIn People search (ICP title + agents), closest to Bucket 4. Source: https://www.linkedin.com/in/dr-sanjay-saini-5b656519/ (found via people search '"Director of AI" agents production'). Company size: ~1,300-2,066 employees (varying sources; most cite ~1,300 — in-band 50-2000). TNS = payments/telecom/financial infrastructure (IaaS). Agent evidence is self-reported in his headline: 'Leading production GenAI systems at scale (LLMs, Agents, ML).' Pain points: running production GenAI/agents at scale (reliability, observability, cost at scale). Challenges: operationalizing agents inside a large payments/telecom infra org; governance & scale. Must-haves: reliability, scalability of production agents. Nice-to-haves: cost-per-run visibility, evals. ICP confidence: Medium. Caveats: TNS is not AI-native (payments/telecom IaaS) and company-level agent-shipping is evidenced only by his own profile; upper headcount estimate (2,066) is near the 2,000 ceiling.
Dror Weiss — Co-founder & CEO, Tabnine
Signal: 4 (co-founder building & shipping AI coding agents in production). Source: https://www.tabnine.com/blog/author/dror/ ; https://www.crunchbase.com/person/dror-weiss ; https://en.wikipedia.org/wiki/Tabnine . Company size: 68 employees (May 2026), Series B ($25M). Tabnine ships AI coding agents/assistants used by ~1M developers, 10M+ installs. Technical co-founder/CEO (CTO is Eran Yahav, already in brain). Pain points: reliability & control of coding agents in enterprise; inference cost of code-gen at scale; governance/privacy for regulated buyers. Challenges: keeping agent behavior controllable and cost-predictable across large enterprise deployments. Must-haves: governance/control over agent behavior + cost efficiency. Nice-to-haves: observability into agent outputs/quality. ICP confidence: Medium-High — technical co-founder/CEO at in-range company shipping coding agents in production.
Signal: 4 (ADAPTED — public conference speaker listing, not LinkedIn; LinkedIn was unavailable this run).
Source: https://www.ai.engineer/worldsfair/schedule — AI Engineer World's Fair 2026, Expo Stage 1 NE, Day 4 (Thu 7/2).
Talk (verbatim): "Harness Engineering: The New Core Skill for Agentic Developers".
Company size: 59 employees (Tracxn, 31 May 2026). $125M raised across 2 rounds; Series A reported by TechCrunch at a $500M valuation (Nov 2024). At the very bottom of the 50–2,000 band — re-verify before outreach.
Pain points: INFERRED FROM TALK TITLE, NOT A VERBATIM QUOTE — "harness engineering" is the same primitive we sell: the loop around the agent, not the model. Strongest thesis alignment of anyone found this run.
Challenges: making agentic developer workflows repeatable rather than one-shot.
Must-haves: a controllable harness/loop around agent execution.
Nice-to-haves: cost and reliability telemetry per loop iteration.
ICP confidence: Medium — AI-native Series A company shipping agentic tooling and a Head-of-Product title that sits inside the ICP's "VP of Product (AI-focused)" lane; downgraded because 59 employees is borderline and product rather than engineering leadership.
NO LinkedIn URL captured — do not fabricate one.
Duo Ding — Engineering & Research Leader (AI agent workflows, agent reasoning & evaluation), Cresta
Signal: 4 (ICP engineering/research leader owning agent reasoning + evaluation at an agent-native company; found via LinkedIn people search "Cresta AI head of engineering agents" this run).
Source: https://www.linkedin.com/search/results/people/?keywords=Cresta%20AI%20head%20of%20engineering%20agents — profile summary: "Engineering & Research Leader at Cresta since January 2026, building AI agent workflows, RAG, agent reasoning and evaluation systems." Ex-Apple. Menlo Park, CA.
Company size: 501–1,000 employees (LinkedIn company page, verified this run). Cresta = AI-native contact-centre platform running real-time AI agents across voice, chat and email in production for large enterprises.
Pain points: NO public pain post captured this run — do not treat as expressed. Qualification is role + company based.
Challenges (inferred from role scope, flagged as inference): explicitly owns agent reasoning AND evaluation systems — the function that absorbs eval cost, non-determinism and regression risk as agent count grows. Real-time voice agents add hard latency + per-minute inference cost constraints.
Must-haves: unverified this run.
Nice-to-haves: unverified this run.
ICP confidence: Medium — clearly senior ("Engineering & Research Leader", ex-Apple) at a qualifying company and squarely on the agent-eval/reasoning surface, but the public title does not state Director/VP explicitly, so seniority is not fully confirmed. Cresta is also later-stage than the Series A–C band. Several other Cresta contacts already exist in the People Library (Tim Shi, Daniel Hoske, Xiangru Chen, Jove Zhong, etc.); Duo Ding is net-new.
Signal: 4 (technical co-founder/CEO at an agent-native company shipping agents in production; found via 2026 agent-startup funding research + LinkedIn verification).
Source: https://www.linkedin.com/in/dvirginzburg/ ; company context https://techcrunch.com/2026/07/29/encore-ai-raises-30m-to-build-ai-agents-that-learn-from-customer-calls/ , https://www.prnewswire.com/news-releases/encore-ai-raises-30m-to-deploy-the-only-enterprise-ai-platform-built-to-generate-revenue-from-customer-interactions-302837694.html
Company size: ~50 employees (Tel Aviv / New York; formerly Insait); Series A $30M (July 2026, led by Team8); 40+ enterprise customers (mostly banks/insurers). Agentic customer-interaction platform (voice/CX agents built to drive revenue).
Pain points (inferred from role/company, not quoted): production voice/CX agents at 40+ enterprises must be reliable and cost-controlled per interaction; scaling agent volume across regulated financial customers; visibility into cost per run/interaction.
Challenges: predictable per-interaction agent cost; reliability/compliance at bank-grade scale; scaling from pilots to broad production.
Must-haves: cost-per-run visibility and control; reliability/observability for customer-facing agents.
Nice-to-haves: token/context waste reduction; per-customer cost attribution.
ICP confidence: Medium-High — technical Co-Founder & CEO (PhD in deep learning, ex-Microsoft) of a Series A agent-native company; headcount ~50 sits at the lower ICP boundary.
Signal: 1 (ICP engaging on agent reliability/observability/evals). Source: https://softwareengineeringdaily.com/2025/04/22/agentic-ai-at-glean-with-eddie-zhou/ and Arize Observe fireside "AI Agent Evals, Permissions, and Production Trust" (https://www.youtube.com/watch?v=NY8gftpN8q4). Company size: ~1,000-1,500 (under 2,000 ceiling; ~$7.2B valuation enterprise-search/agents company). Pain points: evaluating dynamic agents that "can take multiple paths," optimizing for deterministic outputs, tool orchestration, building enterprise agents people "can actually trust." Challenges: agent evals/observability, non-determinism, production trust and permissions at enterprise scale. Must-haves: agent evaluation + observability, deterministic/controllable behavior. Nice-to-haves: cost visibility on multi-path agent runs. ICP confidence: Medium — deeply influential technical voice on agent reliability/evals at an in-range company; caveat: title is Founding Engineer (senior IC) rather than a formally titled decision-maker.
Edward Wu — Founder & CEO (technical founder; ex-ExtraHop lead architect), Dropzone AI
Signal: 1 & 4 — public interviews/commentary on autonomous AI security (SOC) agents in production. Source: https://www.fastcompany.com/91550806/dropzone-ais-edward-wu-wants-ai-agents-to-fight-cyberattacks ; https://www.dropzone.ai/press-release/dropzone-ai-raises-16-85-million-series-a-to-equip-cyber-defenders-with-24-7-generative-ai-powered-autonomous-investigations. Company size: ~50-56 employees (PitchBook/Crunchbase, Jan 2026); Series A ($16.85M; $20M+ total). Building/shipping: autonomous SOC agents deployed across 6+ production customer environments. Pain points: reliability & trust of autonomous agents making security-investigation decisions; consistent agent behavior on high-stakes alerts; scaling agents across many prod environments. Challenges: proving agent accuracy to SOC teams; observability into agent reasoning and tool-calls. Must-haves: reliability + auditability of agent decisions in production. Nice-to-haves: cost visibility per investigation/run. ICP confidence: Medium-High (technical founder/decision-maker; 50+ emp Series A; 6+ agents in prod).
Signal: 1 (ICP speaking about agents breaking in production / control at scale). Source: Gong "Mission Big Dipper" / Revenue Harness launch, June 2026 (https://www.prnewswire.com/news-releases/gong-launches-mission-big-dipper-unveils-industry-first-revenue-harness-302808677.html) + Gong Agents blog (https://www.gong.io/blog/announcing-gong-agents-for-revenue-teams). Company size: ~1,000–1,200 (within 50–2,000). Pain points: generic AI tools "break in production because they lack revenue context, governance, and human-in-the-loop control"; orchestrating many governed agents across the revenue cycle. Challenges: agent reliability, governance and permissioning at enterprise scale. Must-haves: governance/orchestration/control layer for agents with human-in-the-loop. Nice-to-haves: cost attribution across agents. ICP confidence: Medium — technical co-founder shipping a multi-agent execution layer in production; headcount fits; caveats are late stage and CPO (vs CTO) title.
Signal: 1 (ICP writing/speaking about agent + inference cost). Source: The Twenty Minute VC (https://www.thetwentyminutevc.com/eiso-kant) & Air Street "There is no scaling wall" (https://press.airstreet.com/p/there-is-no-scaling-wall-in-discussion). Company size: ~150 (Dec 2025), AI-native coding-agent / foundation-model lab, ~$3B val. Pain points: inference cost as the core economic constraint of running coding agents at scale; original Chinchilla scaling laws ignored inference cost that they now have to pay to run models/agents. Challenges: unit economics of running capable software agents; keeping cost-per-run down as agents scale. Must-haves: cost-efficient inference for agentic workloads. Nice-to-haves: better visibility into agent/token economics. ICP confidence: High — technical co-founder/CTO at a 150-person AI-native company building coding agents, publicly vocal on agent/inference cost.
Eli Ben-Joseph — Co-Founder & CEO (technical), Regard
Signal: 4 (ICP technical co-founder/CEO at AI-native company). Source: https://www.fiercehealthcare.com/ai-and-machine-learning/regard-picks-61m-build-out-ai-powered-clinical-insights-research-llms ; Tracxn (137 employees as of May 2026). Company size: 137. Series B ($61M, Oak HC/FT). Building AI agents: clinical AI that reads EHR data to autonomously surface diagnoses and generate care plans for clinicians across hospital systems. Pain points (inferred from domain + market signal, NOT a verbatim quote): production reliability/accuracy of clinical agents; scaling across hospital systems; cost per patient-encounter run; auditability. Challenges: clinical accuracy + regulatory trust at scale. Must-haves: reliability + clinical accuracy + auditability. Nice-to-haves: cost-per-run visibility. ICP confidence: Medium (technical co-founder/CEO; Series B; 137 employees; agentic clinical decision support — the "5+ agents in production" framing is an inference from the product, not a stated fact).
Eli Brosh — VP AI, Papaya Global
Signal: LinkedIn People search (ICP title + agents), closest to Bucket 4. Source: https://www.linkedin.com/in/eli-brosh-058989/ (found via people search '"VP of AI" agents platform'). Company size: ~683 employees (verified via web/Revelio, 2026). Papaya Global (workforce payments/payroll SaaS) is actively shipping AI agents in production: AI Payroll Data Validation Agent (99.7% accuracy across 160+ countries), 'OneData' global workforce AI agent, 'All-in Data Agents', and a 2026 Reindeer partnership automating high-volume finance workflows — multiple production agents. Pain points: accuracy & reliability of finance/payroll validation agents, compliance across 160+ jurisdictions, scaling multiple production agents. Challenges: enterprise-grade reliability & governance for money-movement agents; automating high-volume workflows without errors. Must-haves: accuracy, compliance, reliability at scale. Nice-to-haves: cost-per-run visibility, observability across agents. ICP confidence: Medium-High (VP AI = VP-level AI leader; in-band size confirmed; multiple production agents confirmed). Company is payments/HR SaaS (not pure AI-native) but clearly building/shipping agents.
Eli Cohen — Director of Technology Incubation, Snyk
Signal: 4 (ADAPTED — public conference speaker listing, not LinkedIn; LinkedIn was unavailable this run).
Source: https://www.ai.engineer/worldsfair/schedule — AI Engineer World's Fair 2026, Expo Stage 1 NE, Day 3 (Wed 7/1).
Talk (verbatim): "Continuous Offensive Security the only approach in an agent-first world".
Company size: ~1,870 employees (Crustdata, Aug 2026; Tracxn 1,854 as of 30 June 2026; PitchBook 1,550). Within the 50–2,000 band. Late-stage private, not Series A–C.
Pain points: INFERRED FROM TALK TITLE, NOT A VERBATIM QUOTE — an "agent-first world" changes the control and verification surface; continuous rather than point-in-time assurance is required.
Challenges: verifying autonomous agent behaviour continuously rather than at release gates.
Must-haves: runtime control and verification over what agents actually do.
Nice-to-haves: cost/observability tie-in per agent run.
ICP confidence: Medium — Director-level technical leader at a qualifying-headcount company, but the expressed signal is security/control rather than cost blowout, and Snyk is a security vendor rather than an agent-shipping SaaS in the classic sense.
NO LinkedIn URL captured — do not fabricate one.
Ellie Zhou — Sr Director of Applied AI, Ironclad
Signal: 4 (applied-AI leader at a company shipping agents in production). Source: LinkedIn (Ironclad team post naming AI leadership — Mingsheng Hong (VP of AI), Alvin Dias (VP of Engineering), Ellie Zhou (Sr Director of Applied AI), Hongmin Fan — heading to AI Engineer World's Fair 2026). Company size: ~600 (legal AI, $200M+ ARR; applied-AI org building agents). Pain points: (company-level) scaling reliable, cost-controlled agents in production legal workflows. Challenges: agent reliability/quality and cost at enterprise scale. Must-haves: agent cost control + output reliability. Nice-to-haves: eval/observability for applied-AI agents. ICP confidence: Medium — Sr Director of Applied AI (decision-influencer) at an active agent-builder; signal is company/team-level (Signal 4). Profile URL not verified.
Elliot Trabac — Director of Engineering, Gorgias
Signal: 4 — ICP-title engineering leader inside an agent-shipping org (Gorgias AI Agent for ecommerce support).
Source: https://www.linkedin.com/company/gorgias/people/?keywords=vp%20engineering — headline read verbatim this run: "Director of Engineering @Gorgias". His recent-activity page was blocked/unreadable this run, so no post-level signal was captured.
Company size: 201-500 employees (LinkedIn company page, verified this run). Gorgias = ecommerce CX platform, Series C, ships AI Agent handling support tickets autonomously.
Pain points: none captured directly — do not claim any in outreach.
Challenges (inferred from org context only): Gorgias runs autonomous support agents per merchant across a large merchant base; peers there include Victor Duprez (Head of Engineering – AI Agent, already in the brain) and Pierre-Alexandre Masse (SVP Eng, already in the brain).
Must-haves / nice-to-haves: unknown.
ICP confidence: Medium — title and company both qualify cleanly (Director of Engineering, 201-500 emp, Series C, agents in production) but zero first-person signal was observable this run. Best used as a second contact on the Gorgias account behind Victor Duprez.
Signal: 4 (ICP senior product leader at agent-shipping company). Source: https://www.linkedin.com/company/gleanwork/people/?keywords=head%20of. Company size: 501–1K employees (per Glean LinkedIn). Company context: Glean is a Work AI platform (AI-native) that builds and ships AI agents/assistants across enterprise data; actively shipping Glean Agents in production. Pain points (inferred from role/ICP): heading product for an agent platform — reliability of agents across many enterprise data sources, and cost/token efficiency of agent runs at large deployment scale. Challenges: scaling from a few to many production agents while keeping cost and behavior predictable. Must-haves: per-run cost visibility and reliability/observability for agents. Nice-to-haves: optimization to cut token waste without degrading quality. ICP confidence: Medium-High — Head of Product (AI-focused) at an AI-native agent platform in target size band; product leadership rather than pure engineering.
Signal: 1 (ICP CTO publicly writing/speaking about agent reliability & cost in production). Source: https://stackoverflow.blog/2026/02/04/code-smells-for-ai-agents-q-and-a-with-eno-reyes-of-factory/ and https://www.youtube.com/watch?v=7lhWqnTfxlE ("How can we trust deploying AI coding agents at scale?") + ZenML LLMOps case study "Building Reliable Agentic Systems in Production". Company size: est. ~50-200 (Series B, NEA-led; 200% QoQ growth; enterprise customers incl. MongoDB, EY, Zapier, Bayer, Clari). AI-native, shipping autonomous coding agents (Droid) in production. Pain points: agents "work great in demos but fail unreliably in production"; 10-20% failure rates at scale = large volume of failed tasks; trusting agent deployment at enterprise scale; making codebases "agent-ready". Challenges: production-grade reliability of multi-step coding agents; observability/trust for autonomous code changes; cost of agents eating into large engineering budgets. Must-haves: reliability at scale, visibility/trust into what agents do before deploy. Nice-to-haves: cost efficiency of iterative agent loops. ICP confidence: High — technical co-founder/CTO and decision-maker at an AI-native company actively shipping agents in production, vocal exactly on Alpha's reliability+control thesis.
Eoin Hinchy — Co-Founder & CEO (technical founder; ex-security engineer / Director of Security, DocuSign and eBay), Tines
Signal: 4 (ICP-matching technical founder publicly writing/speaking about shipping and governing agents in production). Source: https://siliconangle.com/2026/07/28/tines-launches-3b-help-enterprises-govern-ai-built-apps-agents/ and https://www.prnewswire.com/news-releases/tines-launches-3b-to-help-enterprises-govern-ai-workflows-apps-and-agents-302836517.html (Tines 3B launch, Jul 2026); https://siliconangle.com/2026/01/15/tines-launches-ai-interaction-layer-unify-agents-copilots-workflows/ (AI in Tines, Jan 2026); https://www.tines.com/platform/agents/; https://www.linkedin.com/posts/eoinhinchy_fin-x-tines-approaches-to-building-ai-enabled-activity-7360967105963323392-h4EK. Company size: ~470-600 employees (Revelio 470 Mar 2026; Tracxn 548; other trackers ~601 Jul 2026) — within band. Series C, $125M Feb 2025 at $1.125B valuation, ~$272M raised (Goldman Sachs, SoftBank, Felicis, Accel, CrowdStrike). Ships agents, copilots, workflow automation to security/IT/ops teams; Tines 3B is explicitly a build/run/govern layer for AI-generated workflows, apps and agents. Pain points (his own framing): "every organization is experimenting with AI, but most are struggling to operationalize it"; landscape of "isolated copilots" and "'black box' agents that IT leaders are afraid to trust"; the hard part is no longer building but "knowing what's running, whether you can trust it, and who's responsible for it"; "wild code" AI sprawl. Challenges: governance and ownership of agent-generated work; trust and auditability at enterprise scale; connecting agents to real systems safely. Must-haves: runtime visibility into what agents are executing, ownership/accountability per agent, governance and guardrails. Nice-to-haves: unified interaction layer across agents/copilots/workflows; cost attribution per agent run. ICP confidence: High (technical co-founder/CEO at a ~500-person Series C company shipping multiple agent products; his public pain language maps almost 1:1 to the agent-visibility/control wedge). Note: title is CEO but he is a hands-on technical founder, which the ICP explicitly allows.
Signal: 4 (ICP technical leader at company shipping multiple AI agents in production). Source: https://pulse2.com/tabnine-profile-eran-yahav-interview/ , https://devops.com/tabnine-adds-agents-capable-of-automating-workflows-to-ai-coding-platform/ , https://www.crunchbase.com/organization/tabnine . Company size: ~68 employees (May 2026), Series B, total funding ~$102M. Named Visionary in 2026 Gartner MQ for Enterprise AI Coding Agents. Pain points: running multiple autonomous coding agents across the SDLC (code gen, testing, docs, review, bug fixing) — reliability and correctness of agent output in enterprise codebases; enterprise-grade governance/control over agent behavior. Challenges: scaling multi-agent workflows enterprise-wide while keeping outputs trustworthy and auditable; on-prem/air-gapped enterprise deployment constraints add cost/complexity. Must-haves: reliability and control of agents at scale, visibility into what agents do. Nice-to-haves: cost attribution per agent run, model routing to control LLM spend. ICP confidence: High — technical co-founder/CTO at an AI-native ~68-person Series B company whose core product is production AI agents. (Note: profileUrl is company page; personal LinkedIn not verified in this run.)
Eric Grigson — Director of Developer Experience, Culture Amp
Signal: 4 (ICP speaking publicly about running agents at scale). Title verified live on LinkedIn people search 2026-09-02: "Current: Director of Developer Experience at Culture Amp", Melbourne, Australia. Headline: "Engineering Leader | Developer Experience & AI Transformation | Data-Driven Insights".
Source: https://webdirections.org/ai-engineer-melbourne-26/speakers/eric-grigson-paul-hughes.php (AI Engineer Melbourne 2026, Leadership track). LinkedIn URL href-verified from search results.
Company size: 501-1K (Culture Amp LinkedIn company page, read live 2026-09-02); ~928 Revelio Labs Dec 2025 / ~954 LeadIQ May 2026. Speaker bio cites a 200-person engineering org. Melbourne, Australia.
Pain points: Ran a six-month research program across 88 engineers measuring DORA metrics, adoption telemetry and sentiment. Candid about the regressions: MTTR increased post-rollout, decentralised messaging created confusion, and out-of-hours commits rose. Explicitly driving "a deliberate shift toward agentic AI across their entire product organisation" — so agent count is going up while reliability metrics are already moving the wrong way.
Challenges: measuring agent ROI against reliability regression; MTTR degradation after AI rollout; no single place to see agent usage, cost and outcome across 200 engineers.
Must-haves: unified adoption + reliability telemetry for agents, attributable to team and repo; evidence to defend or kill the programme.
Nice-to-haves: benchmarking against industry; guardrails that work without top-down mandates; spend-per-team reporting.
ICP confidence: Medium-High — Director (in band), 501-1K employees (in band), agentic rollout across the whole product org, and unusually well-instrumented, quotable evidence of reliability pain. Discount: the agent estate is internal developer productivity, not 5+ customer-facing production agents. Net-new name AND net-new company.
Signal: 4 (ICP writing publicly about shipping agents). Found via web research — LinkedIn/Chrome unavailable this run.
Source: https://www.shift-technology.com/en-gb/resources/reports-and-insights/four-questions-with-eric-sibony-shifts-agents (21 Apr 2026)
Company size: ~627–656 employees (Revelio Labs 656 worldwide as of Mar 2026; Tracxn 627 as of 30 Jun 2026). Paris/Boston. Insurance AI — claims, fraud, subrogation, underwriting. Late-stage (well past Series C) — stage is the ICP stretch.
Agent evidence: describes Shift's move from "AI as core" to "agents as product" — they "injected agentic reasoning into the engine and introduced an integration/packaging layer so agents behave as callable services and can orchestrate tools such as entity reconstruction, APIs, or external browsing." A fraud capability "becomes a 'fraud agent' that assesses a claim, runs investigations with multiple tools, escalates to humans when needed, and can be invoked by other agents in a broader claims orchestration flow." Shipped Shift Claims (agentic AI) Sept 2025; renewed 5-year AXA partnership Mar 2026 across 15 countries.
Verbatim: "prompt engineering alone is simply not enough"; on model churn — "migrating and re-optimising systems is now part of ongoing product work, not a one-time project."
Pain points: model/tooling churn making migration a permanent cost line; long-horizon state management across multi-day tasks; getting agents to the >99% accuracy insurance flows require.
Challenges: building a data architecture that feeds agents correctly; tool integrations; drift monitoring; making abstention ("I don't know") a first-class, risk-adjusted output rather than forcing binary answers; standing up dedicated R&D that continuously adapts to rapidly changing models.
Must-haves: multi-agent review/orchestration layer; calibrated confidence + human escalation paths; structured access to curated context and decision logic; auditable, explainable outputs; agents packaged as callable, composable services.
Nice-to-haves: interoperable agent-ecosystem protocols so agents chain work across vendors; browsing/external tool access; persistent state over multi-day/multi-month tasks.
ICP confidence: Medium — title, vertical, agent-shipping fit and headcount band are all clean; company stage (late-stage, past Series C) is the only miss. Strong "agents-as-callable-services + model churn cost" narrative that maps directly to Alpha's operating-layer pitch.
Eric Steinberger — Co-founder & CEO, Magic.dev (Magic AI)
Signal: 4 (technical founder building coding agents / long-context models for code). Source: https://tracxn.com/d/companies/magicai/ and https://www.linkedin.com/in/ericsteinb/. Company size: 109 employees (Apr 2026); Series C, $466M raised over 4 rounds. Building AI code generation/completion + long-context coding agents. Pain points: reliability of long-horizon coding agents, context/token efficiency at very long context windows, cost of running large-context agent inference. Challenges: making autonomous coding agents dependable over long tasks. Must-haves: context/token waste reduction, cost-per-run visibility, agent reliability at scale. Nice-to-haves: routing between models per task. ICP confidence: High (technical decision-maker, AI-native agent company, 109 employees, Series C).
Signal: 1 (ICP writing about agent cost blowout). Source: https://www.linkedin.com/in/erikpeterson + QCon AI Boston 2026 session; shares postmortems incl. a $47K eleven-day agent loop and an engineering team whose annual AI budget was gone in weeks; "aggregate AI spend up 320%, AI billing structurally different from cloud billing." Company size: ~150-250 employees, Series C; CloudZero = cloud/AI cost intelligence, building agentic cost-intelligence features. Pain points: runaway agent cost/loops, no per-run cost visibility, AI budgets exhausted fast. Challenges: attributing agent spend, structurally new billing model for agents. Must-haves: real-time per-agent/per-run cost visibility, spend guardrails. Nice-to-haves: forecasting. ICP confidence: Medium (technical founder/CTO, 50-2000 emp; strong pain match, but CloudZero is cost-observability adjacent — potential competitor to Alpha's cost-control layer; validate positioning before outreach).
Erika Rice Scherpelz — Head of Engineering, Sourcegraph
Signal: 4 (ICP technical leader at agent/coding-AI company). Source: https://www.linkedin.com/in/erikars/ ; theorg.com & Bloomberg confirm Head of Engineering. Company size: ~150-250 emp (Sourcegraph; ships Cody AI coding assistant + Amp coding agent). Pain points (inferred from role + company, labeled as such): reliability/accuracy of AI coding assistant at enterprise scale; eval of coding agents; cost of LLM calls across code-search + Cody; platform/infra/security across product lines. Challenges: leading ~60 engineers across Enterprise Code Search, Cody, and Platform/Infra/Security; making agentic coding reliable and cost-effective for enterprise. Must-haves: cost-per-run visibility + eval/reliability for coding agents. Nice-to-haves: routing/model-selection to control LLM spend. ICP confidence: High (Head of Engineering at in-range coding-agent company; verified via multiple sources).
Ershad Ali Mohammad — SVP Engineering, Kore.ai
Signal: 4 (net-new ICP senior technical leader at an agent platform company). Source: https://www.linkedin.com/in/ershadalim/ ; LinkedIn people search (Kore.ai engineering), 2026-07-27 run. Headline: 'SVP Engineering @ Kore.ai | Building Enterprise Agentic AI Platform - Multi-Agent, Voice, RAG | Scaled 120+ Eng Orgs | Ex-Harness, HighRadius, Bloomberg'. Company size: ~1,000 employees (mature; Series D-stage agent platform) - NOTE larger/later-stage than the Series A-C guideline but within the 50-2,000 band and squarely agent-native. Kore.ai ships an enterprise multi-agent / voice / RAG platform in production. Pain points (INFERRED from role+platform context, not verbatim): cost + reliability across a multi-agent enterprise platform; inference/token spend at scale; per-agent-run observability. Challenges: scaling 1 -> many agents; controlling cost/reliability across enterprise deployments. Must-haves: cost-per-run + reliability observability/control. Nice-to-haves: vendor-neutral routing; replay/simulation. ICP confidence: High (SVP Engineering, agent platform; note large/mature company).
Ertan Dogrultan — Director of Engineering, Platform (headline: Platform Engineering), Replit
Signal: 4 (ICP technical leader at agent-native company, posting about agent infrastructure & cost). Source: https://www.linkedin.com/in/ertand/ (LinkedIn posts, May-Jul 2026); title per theorg.com (Director of Engineering). Company size: ~200-400 emp (Replit; ships Replit Agent). Pain points: running tens of thousands of agents in parallel (~4x normal load on free days); agent infra economics ('a day without thinking about cost'); durable execution + sandbox protocol for agents. Challenges: scaling agent platform reliably at spiky global load; cost of free/high-volume agent runs. Must-haves: agent-run cost visibility & control, reliable agent infra at scale. Nice-to-haves: event-streaming/durable-execution efficiency, benchmarking on real data. ICP confidence: Medium-High (Director-level platform/agent-infra leader at agent company; verified current at Replit; headline generic but external source confirms Director title).
Esha Manideep — Co-Founder & CTO, Giga (GigaML)
Signal: Bucket 1 (ICP building/shipping agents — found via 2026 agent-startup research + LinkedIn verification). Source: https://www.linkedin.com/in/esha-manideep/ ; company confirmation https://finance.yahoo.com/news/giga-raises-61m-funding-reinvent-171500172.html . Company size: ~66 employees (Tracxn, Jan 2026); Series A $61M (Redpoint, YC S23, Nexus). Giga ships enterprise voice + chat support/ops agents in production (DoorDash, Zepto), 90%+ real-world resolution. CTO / technical co-founder (IIT Kharagpur, Forbes 30U30) — owns the agent architecture. Pain points: production reliability at scale, token/context waste in multi-tool agent runs, cost-per-run visibility as agent volume grows. Challenges: engineering agents that hold up in production across many enterprise customers and reasoning steps; keeping per-conversation cost down while maintaining resolution quality. Must-haves: observability into what each agent run costs and why it fails; runaway-loop / budget guardrails. Nice-to-haves: an agent operating layer to standardize eval, cost control and reliability across the fleet. ICP confidence: High (CTO at a ~66-person Series A company actively shipping agents in production — direct buyer persona).
Signal: 4 (ICP technical co-founder/CTO at agentic procurement company shipping agents in production). Source: The Next Web / fintech.global (Pivot $40M Series B, May 2026), Tracxn headcount, LinkedIn (pivothq). Company size: 116 employees (Tracxn, Jun 2026; grew from 94 in May) — confirmed. Pivot = agentic-AI procurement operating system (source-to-pay) deploying specialized multi-agents across sourcing, approvals, purchasing, invoicing, reporting; $40M Series B (Forestay Capital + Notion Capital co-lead; $70M total); Paris HQ with offices in London/Berlin/NY/Tel Aviv/São Paulo; customers DoorDash, Lemonade, Flix. Pain points: reliability/accuracy of multi-agent procurement automation in the enterprise; cutting procurement cost + cycle time; deep ERP integration where legacy software fails. Challenges: end-to-end agent reliability across the full procurement lifecycle. Must-haves: production reliability, control, integrations. Nice-to-haves: per-run cost visibility, spend analytics. ICP confidence: High (Co-Founder & CTO at 116-person agent-native Series B company).
Signal: 4 (ICP at a company shipping agent-facing products). Source: role verified https://aiconference.com/speakers/etienne-dilocker-2/; Weaviate shipped Agent Skills (Feb 2026) for coding agents, plus Weaviate Agents. Company size: ~74-104 (GetLatka ~104; PitchBook 99; Tracxn 74 as of 30 Apr 2026) — above the 50 floor on all three but near the bottom of the band. Stage: Series B — $50M led by Index Ventures; plus strategic investment from RICOH Innovation Fund (Mar 2026). No Series C found. Pain points: INFERRED from role/company context. The only first-person material verifiable is his 2024 AI Conference talk on large-scale vector search discussing "cost-effective, tailored solutions" for infrastructure sizing — cost-conscious but pre-agent. NO direct 2025-26 statement from him on agent cost, reliability, or context waste was found. Challenges (inferred): agent memory and retrieval at scale; cost-effective index configuration; serving agentic query patterns. Must-haves (inferred): efficient agent memory/retrieval, predictable cost at scale. Nice-to-haves (inferred): per-agent retrieval cost visibility. ICP confidence: Medium — role, stage and headcount all verify cleanly and the company is demonstrably shipping agent products, but zero verified personal pain signal. Treat as a cold-but-qualified name, not a warm one. LOWEST-priority of this run's additions.
Signal: 4 (ICP technical leader at an agent-native company absent from the brain). Source: gap-company research after brain dedup (Commure not in the current list). Commure is an AI-native healthcare RCM & ambient platform; 'Commure Agents' ships multiple production agents — AI Call Center, Scheduling (Sherpa), Patient Outreach, Referral Management, Prior Authorization, and Discharge Planning (6+ named agents live). Company size: ~1,600 employees (2026), 300%+ YoY growth two years running. Eugene Kuznetsov is Co-Founder & CTO (ex-Salesforce VP Product IoT Cloud, ex-Veracode, MIT). Pain points: reliability/accuracy of many healthcare agents in production; cost per run at scale; observability + governance across a large agent fleet and enterprise deployments; latency. Must-haves: reliability, per-run cost visibility, governance/compliance across agents. ICP confidence: High (Co-founder & CTO; agent-native; headcount within 50-2,000). Note: pains inferred from role/company profile, not a personal quote.
Eugene Mann — Co-Founder & Chief Product Officer, Maven AGI
Signal: 4 (ICP technical/product co-founder at a qualifying agent company). Source: targeted ICP search of enterprise AI-agent CX platforms; profile https://www.linkedin.com/in/eugenemann/ (role Co-Founder & CPO, also per theorg.com; ex-Stripe Applied ML product lead, ex-Google). Company size: ~100-150 (Maven AGI, enterprise AI-agent platform for CX, Series B; resolves up to ~93% of support queries autonomously). Pain points [INFERRED from role+company, not a verbatim quote]: reliability/accuracy of support agents at high autonomous-resolution rates in production; controlling per-resolution LLM cost at enterprise scale; eval/guardrails for autonomous resolution. Challenges: many FDE-heavy enterprise deployments, keeping agents dependable across tenants. Must-haves: production reliability + cost visibility per agent/run. Nice-to-haves: agent analytics for product decisions. ICP confidence: Medium (co-founder CPO = 'VP of Product (AI-focused)' per ICP, at an agent company).
Everett Berry — Head of GTM Engineering, Clay
Signal: 4 (ADAPTED — public conference speaker listing, not LinkedIn; LinkedIn was unavailable this run).
Source: https://www.ai.engineer/worldsfair/schedule — AI Engineer World's Fair 2026, "AI in GTM" track, Day 4 (Thu 7/2).
Talk (verbatim): "GTM Engineering: The Technical Bits".
Company size: ~1,167 employees (Tracxn, 28 Feb 2026; ~1,000–1,100 across other sources in 2026). Series C: $100M at $3.1B valuation, Aug 2025, led by CapitalG. Fits both the headcount band and the Series A–C window.
Pain points: INFERRED FROM ROLE AND COMPANY, NOT A VERBATIM QUOTE — Clay's core product runs very high-volume LLM/agent enrichment workflows on behalf of customers, which is exactly the shape where per-run token economics dominate gross margin.
Challenges: unit economics of agent runs at customer scale; keeping enrichment workflows reliable across many models and providers.
Must-haves: cost per run visibility at workflow granularity.
Nice-to-haves: model routing and context pruning to protect margin.
ICP confidence: Medium — Head-level and genuinely technical, at a Series C company inside the headcount band that runs agents at scale; downgraded from High because the function is GTM engineering rather than core platform/AI engineering, so budget authority over agent infrastructure is unconfirmed.
NO LinkedIn URL captured — do not fabricate one.
Eyal Ben Barouch — Head of Data & AI, Tavily
Signal: 1 (ICP speaking about data/search infrastructure for AI agents — LangTalks AI Engineering Conference 2026 speaker; Head of Data & AI at Tavily, "the search engine for AI agents").
Source: https://langtalks.ai/conference ; https://theorg.com/org/tavily?person=eyal-ben-barouch ; MongoDB case study "How Tavily uses MongoDB to enhance agentic workflows"
Company size: ~91 employees (Tavily; LinkedIn lists 51–200; Series A ~$20–25M; NY / Tel Aviv / Abu Dhabi). Within 50–2,000.
Pain points: powering reliable agentic workflows with clean, relevant web data at scale; retrieval/search reliability for AI agents; agent data quality feeding production agents.
Challenges: scaling data/AI infrastructure that many agents depend on; keeping agentic workflows reliable and grounded.
Must-haves: reliable retrieval and scalable agent infrastructure.
Nice-to-haves: cost-efficient agent data/inference pipelines.
ICP confidence: Medium — Head-of-AI-level leader at an AI-native ~91-person Series A company built for AI agents; role is data-focused and company is infrastructure-for-agents rather than shipping end-user agents.
NOTE: No verified LinkedIn URL captured (profile_url left blank — search-returned URL appeared malformed). Discovered via web/conference fallback — Claude-in-Chrome/LinkedIn browsing was not connected during this scheduled run.
Eyal Peleg — Co-Founder & CTO, Sedric AI
Signal: 3 (financial-services compliance / governance-adjacent — ICP building agentic monitoring). Source: https://www.sedric.ai/blog/sedric-ai-raises-18-5-million-series-a ; https://www.fintechfutures.com/regulations-compliance/llm-powered-compliance-platform-sedric-ai-secures-18-5m-series-a . Company size: 51-200 employees (Tel Aviv; Series A $18.5M led by Foundation Capital, $22M total; founded 2020). Ships an LLM/agent-based compliance platform for banks, lenders, trading platforms & insurers — first dedicated compliance LLM, agentic monitoring across every customer touchpoint. Pain points (INFERRED from domain): keeping compliance-monitoring agents reliable/auditable across high call/chat volume; cost of running LLM monitoring at scale; explainability of agent decisions to regulators. Challenges (INFERRED): scaling agentic monitoring across new financial verticals without accuracy/cost blowup. Must-haves (INFERRED): per-run cost visibility + defensible, auditable agent behavior. Nice-to-haves (INFERRED): unified observability across their agent fleet. ICP confidence: Medium-High (senior technical co-founder, in-band headcount & Series A stage, ships agents; compliance-vertical fit slightly narrower than core cost/reliability ICP). LinkedIn URL not surfaced in unauthenticated search — not fabricated.
Ezra Tanzer — Director, Product Management (Agentic Development Security), Snyk
Signal: 4 (ICP presenting on agent governance at a company shipping agents).
Source: AI Engineer World's Fair 2026 speaker record — https://ai.engineer/worldsfair/2026/speakers.json ; session "Agentic Development Security".
Company size: ~1,870 employees (Aug 2026) — inside the 50–2,000 band but near the ceiling. Snyk's Evo Agentic Development Security launched Jun 2026; agents are central to the roadmap.
Pain points: session abstract is empty in the source dataset — NOT verified. Inferred from his product remit (Agentic Development Security): needing to know what agents are doing inside the SDLC and to constrain them.
Challenges (inferred, unverified): visibility and policy enforcement over agents that generate and ship code.
Must-haves (inferred, unverified): per-agent action visibility and guardrails.
Nice-to-haves (inferred, unverified): cost/usage attribution per agent.
ICP confidence: Medium — meets the Director floor and the AI-focused product criterion, company in the headcount band and actively shipping agents. Downgraded because: no first-party pain signal captured, Snyk is far past Series A–C, and this is a product rather than engineering-leadership seat. Second contact at Snyk this run (with Manoj Nair) — treat as one account, not two independent signals.
Signal: 4 (adapted — LinkedIn auth unavailable this run; identified via web research + verification). Source: https://www.forbes.com/profile/fabian-hedin/ ; https://en.wikipedia.org/wiki/Lovable_(company) ; https://research.contrary.com/company/lovable. Company size: ~1,000+ employees as of mid-2026 (within 50–2,000); Lovable (Swedish "vibe coding" app builder) hit ~$100M ARR in 8 months, valued ~$6.6B (Dec 2025). Co-founder & CTO. Product generates full-stack apps via AI coding agents at very high volume. Pain points (inferred from product/scale): token/compute cost of code-generation agents at consumer scale; reliability of generated output; per-run cost visibility across millions of generations. Challenges: controlling compounding LLM spend as free+paid usage scales; keeping agent output reliable. Must-haves: cost-per-run/token visibility and control; reliability guardrails. Nice-to-haves: observability across the agent generation pipeline. ICP confidence: High — technical co-founder/CTO, agent-native company in size range. LinkedIn URL not captured this run.
Fatih Yildiz — Founder & CTO, Edge Delta
Signal: 1 (company publicly exposing agent cost/model-mix telemetry; ICP owns it). Source: https://edgedelta.com/company/about (live leadership page + real-time agent fleet telemetry panel). Company size: ~100 with plans to double (GeekWire funding coverage) — WEAKEST headcount datapoint of this run, press-sourced not live count; verify before outreach. Stage: Series B — $82M led by Quiet Capital; investors include ServiceNow, Cisco Investments, Menlo. Pain points (read from their own live dashboard, not a personal quote): the About page publishes real-time agent operating stats — "In the last 24 hours, 810B events were used by AI teammates on watch" — plus a models-powering-investigations cost mix (Claude Opus 4.6 49%, GPT 5.4 16%, Gemini 3.5 Flash 15%, GPT 5.6 Sol 9%, Gemini 3.6 Flash 5%, Claude Opus 4.8 3%, GPT 5.5 3%) and integration mix (Kubernetes 34%, GitHub 21%, custom MCP 16%). A 7-model, 4-vendor production routing mix at 810B events/day is a cost-control problem by construction. They also ship a "Data Tiering for AI Teammates" solution page — explicit token/context economics work. Challenges: always-on autonomous agents against streaming telemetry at petabyte scale; agents that "investigate without waiting for a prompt" (unbounded run initiation); multi-model routing across four vendors; avoiding lock-in while keeping cost predictable. Must-haves: cost-aware model routing across a heterogeneous model fleet; data tiering to keep agent context affordable; customer-environment data residency. Nice-to-haves: per-investigation cost attribution; agent-level budget guardrails; multi-agent collaboration across batch/stream. ICP confidence: Medium — High on role and pain fit; downgraded because headcount (~100) rests on a single press source and needs a second confirmation.
Fergal Reid — VP of AI / Chief AI Officer, Intercom (Fin)
Signal: 1 (ICP writing publicly about shipping/scaling LLM agents in production). Source: https://www.linkedin.com/in/fergalreid/ , https://vux.world/why-ai-is-the-future-of-customer-service-with-fergal-reid-chief-ai-officer-at-intercom/ , https://www.intercom.com/blog/author/fergal_reid/ | Company size: ~1,200–1,800 employees (2026; Tracxn ~1,229, others ~1.4–1.8K — within 50–2,000). Intercom's Fin is an autonomous customer-service agent deployed at large scale across thousands of customers. Reid (PhD) leads applied AI / Fin and posts regularly on LLM implementation, agent reliability, and AI pricing/economics. Pain points: shipping LLM-enabled agents that are reliable enough for production customer support; connecting model cost to per-resolution economics (Fin is priced per resolution, so token/inference cost directly hits margin); measuring answer quality at scale. Challenges: moving from impressive demos to dependable production behavior; controlling hallucination/quality; cost-per-resolution as volume scales. Must-haves: production reliability, quality measurement/eval, per-run/per-resolution cost visibility. Nice-to-haves: agent observability and tracing. ICP confidence: High (VP/Chief AI decision-maker at a 50–2,000-employee AI-native company running an agent in production at scale).
Signal: 1 (ICP publicly expressing agent cost + reliability pain). Source: CNBC "users shift from tokenmaxxing to efficiency" (https://www.cnbc.com/2026/06/26/openai-anthropic-new-ai-spending-reality-as-users-shift-to-efficiency.html) + Lindy 3.0 LinkedIn announcement. Company size: ~50 (Tracxn lists 52 as of May 2026; one June 2026 report cited ~25 — BORDERLINE, treated as ~50, size caveat noted). Pain points: "unsustainable" AI/token costs (company was built on the bet that token cost would fall dramatically); moved 100% of traffic off Anthropic Claude to DeepSeek purely for cost; agent reliability "degrades off the happy path" (public 14-item bug list from a paying customer). Challenges: agent cost economics + reliability across 1,600+ app integrations serving 400k+ users. Must-haves: dramatically lower agent run cost without sacrificing reliability. Nice-to-haves: per-agent cost visibility. ICP confidence: Medium — exceptionally strong cost/reliability signal; company headcount borderline ~50 (size caveat).
Florian Neumeier — Vice President Product, Data & Engineering, yoummday
Signal: 4 (ICP at an agent-operating company — surfaced via LinkedIn PEOPLE search, colleague traversal from Dr. David Noel Ng).
Source: https://www.linkedin.com/search/results/people/?keywords=yoummday ; company page https://www.linkedin.com/company/yoummday/
Company size: 201-500 employees (verified on yoummday's LinkedIn company page, 28 Aug 2026). HQ Munich, Germany.
LinkedIn headline (verbatim): "Vice President Product, Data & Engineering at yoummday (AI-powered digitalization of the whole CX value chain)".
Pain points: NOT DIRECTLY EXPRESSED — no pain quote captured; do not fabricate one. He is recorded here as the budget-holding engineering counterpart to David Ng at the same account.
Challenges: single VP owning Product + Data + Engineering at a 200-500 person company running AI agents against live enterprise customer interactions — classic profile for hitting a wall when agent count goes from 1 to 5+.
Must-haves (hypothesis, unverified): cost and reliability visibility across the CX agent fleet, since yoummday prices on business outcomes rather than seats.
Nice-to-haves (hypothesis): consolidated observability across the AI and human-talent sides of the same workflow.
ICP confidence: Medium — title and company size are a clean fit; no expressed pain, so this is a fit-based lead requiring discovery. Pair with David Ng (ID 991) as a two-threaded entry into the same account.
Verification notes: profile URL confirmed by DOM href extraction. Note there are two similarly-named Florian Neumeiers on LinkedIn — the correct one is /in/florianneumeier/ (the CFA one is a different person).
Signal: 4 (ICP eng leader at an agent-shipping company; LinkedIn people-search 'Cresta AI director engineering'). Source: https://www.linkedin.com/in/florinszilagyi/ . Company size: ~400-600 (Cresta; Series C/D CX-AI company shipping a unified human+AI agent platform for contact centers). Pain points (INFERRED from role + company; no verbatim quote observed this run): reliability + cost-per-run control for CX/voice agents at scale; production observability across a large agent fleet. Challenges: scaling a reliable multi-agent CX platform. Must-haves: per-run cost visibility + reliability guardrails. Nice-to-haves: routing/cost optimization. ICP confidence: Medium (Head of a regional R&D/engineering center — genuine eng leadership though regional in scope; Cresta well within the 50-2,000-emp band and actively shipping agents). Note: brain already has Cresta leaders Tim Shi, Daniel Hoske (CTO), Jove Zhong (Head of FDE), Ming Yin, Ping Wu, Deepank Sharma — Szilagyi is net-new.
Gabe Pereyra — President & Co-Founder (technical; ex-DeepMind research scientist), Harvey
Signal: 4 (technical co-founder at company shipping agents at scale). Source: https://www.linkedin.com/in/gabepereyra , https://www.harvey.ai/blog/harvey-raises-at-dollar11-billion-valuation-to-scale-agents-across-law-firms-and-enterprises. Company size: ~350 employees; $11B valuation; scaling agents across law firms and enterprises. Technical co-founder (ex-research scientist DeepMind, ex-ML engineer Meta) directing research/technical roadmap. 25,000+ custom agents in production. Pain points: scaling reliable agents across many enterprise customers, keeping legal-grade accuracy, controlling agent behavior. Challenges: research-to-production reliability; agent evaluation and control at enterprise scale. Must-haves: reliability, eval/observability, behavior control. Nice-to-haves: per-run cost visibility, model-routing efficiency. ICP confidence: High (technical co-founder and decision-maker at an agent-native company in the size band). Note: second Harvey contact captured alongside CTO Siva Gurumurthy — both distinct decision-makers.
Gabor Melli — VP of Artificial Intelligence, LegalOn Technologies
Signal: 4 (ICP shipping agents). NOTE: Claude-in-Chrome/LinkedIn was NOT connected this run (2 retries), so the prescribed LinkedIn post/comment searches could not be run. Found instead via web research + primary-source verification. Source: https://www.legalontech.com/post/we-are-legalon-meet-gabor-melli-vp-of-artificial-intelligence ; appointment PR https://www.legalontech.com/press-releases/legalon-taps-gabor-melli-to-lead-investment-in-ai ; LinkedIn headline reads "VP of AI at LegalOn | PhD | Agentic AI | LLMs" (https://www.linkedin.com/in/melli/). Company size: ~596-754 employees (Revelio Labs 754 as of Mar 2026; PitchBook 596). SoftBank-backed, $50M Series E. Multi-agent shipping confirmed: LegalOn announced five new AI agents for in-house legal teams executing specialized legal tasks inside its platform. Pain points (INFERRED from company-published material, NOT personal quotes — no verbatim quote captured): running 5+ specialized legal agents in a regulated, accuracy-critical domain; per-review cost of long contract context; agent output must be defensible to lawyers. Challenges: scaling from single-purpose contract review to a fleet of agents; steering a multi-million dollar AI investment with measurable unit economics; cross-jurisdiction (US/UK/Japan) model behavior. Must-haves: per-agent and per-review cost visibility; reliability/accuracy guardrails in production. Nice-to-haves: cross-agent context reuse to cut token waste; per-customer cost attribution. ICP confidence: High — exact-match title (VP of AI), Director+ level, company in 50-2,000 band, 5+ agents shipping in production. UNVERIFIED: personal pain statements, current tenure not re-confirmed beyond LinkedIn headline.
Signal: 1/4 (co-founder/CEO of a qualifying enterprise agent platform, public commentator on agent adoption & economics)
Source: https://finance.yahoo.com/sectors/technology/articles/dust-raises-40m-ai-multiplayer-133000633.html | https://www.latent.space/p/dust | https://sequoiacap.com/founder/gabriel-hubert/
Company size: ~144 employees (raised $40M in 2026; platform used by ~51,000 workers at 3,000+ companies)
Pain points (inferred): horizontal enterprise agent platform ("multiplayer AI") deploying custom agents connected to internal data/tools; focus on agent usefulness, reliability and adoption (88% DAU in some deployments) and the economics of running agents at enterprise scale.
Challenges: enterprise agent reliability; safe connection of agents to internal tools/data; driving adoption; cost at scale across many customer agents.
Must-haves: reliable enterprise agent infra; observability/control; safe tool/data connectivity.
Nice-to-haves: cost transparency per agent/workflow.
ICP confidence: Medium (co-founder & CEO / equivalent decision-maker; more product- than eng-oriented — technical co-founder Stanislas Polu already in brain).
Signal: 1 (ICP speaking about agentic AI in production — panelist at LangTalks AI Engineering Conference 2026; VP Engineering at an AI-native company purpose-built around AI agents).
Source: https://langtalks.ai/conference ; company news https://fintech.global/2026/08/04/zenity-lands-125m-as-ai-agent-security-race-heats-up/
Company size: ~230 employees (Series C; closed $125M led by Norwest, Aug 2026; R&D Tel Aviv, commercial NY). Within 50–2,000.
Pain points: securing and governing AI agents in production; agent sprawl as enterprises scale from 1 → many agents; no visibility into what agents are doing/costing across the enterprise; agent reliability and attack surface.
Challenges: scaling agent security/governance as customers deploy more agents; opaque agent behavior.
Must-haves: agent observability, governance controls, guardrails, model/identity policy.
Nice-to-haves: cost visibility per agent run.
ICP confidence: Medium-High — VP Eng at a 230-person Series C AI-native company centered on AI agents; agent-security focus (secures agents vs. shipping its own end-user agents), so not rated Higher.
NOTE: Discovered via web/conference fallback — Claude-in-Chrome/LinkedIn browsing was not connected during this scheduled run.
Signal: 1 (ICP publicly writing/speaking about deploying and controlling agent workflows in production; hosts LangTalks, a podcast on building agentic apps for real production use; conference talk "From Tool Calling to CodeAct: Practical Lessons for Deploying Executable Agent Workflows"). Source: https://home.mlops.community/public/videos/from-tool-calling-to-codeact-practical-lessons-for-deploying-executable-agent-workflows-gal-peretz-agents-in-production-2025-2025-08-04 ; https://www.linkedin.com/in/gal-peretz/ . Company size: ~185-239 employees; was Series D; acquired by Axon Enterprise Feb 2026 (NOTE acquisition/late stage). Prev Head of AI & Data at Torq. Pain points: reliability of executable agent workflows in life-critical (emergency) settings; controlling agent behavior (tool-calling vs CodeAct); moving agents from demo to reliable unattended production. Challenges: agent observability; safety/guardrails; production reliability. Must-haves: control over agent behavior, reliability, observability. Nice-to-haves: latency/cost optimization. ICP confidence: Medium (Head of AI actively deploying agents, ~200 emp; company now part of large Axon).
Signal: 1 (ICP publicly writing about agent context/token cost and compounding spend). NOTE: Claude-in-Chrome/LinkedIn not connected this run (2 retries) — found via web research + primary-source verification instead of the prescribed LinkedIn post search.
Source: https://www.cortex.io/post/context-engineering (Cortex engineering blog, "Context Engineering")
Company size: 104 employees (Tracxn, Jun 2026). Series C — $60M Sep 2024, $198M total. Internal developer portal / engineering intelligence platform, now shipping AI agents over the service catalog.
What he said (verbatim, short): "Every token an agent processes costs money, and every token adds latency."
Pain points: context rot/bloat inflating token spend and latency without improving output quality; teams unable to compare token spend against engineer-time saved; agents that "cost more than they save".
Challenges: engineering "Minimal Relevant Context" rather than dumping everything into the window; keeping agent context fresh as the underlying service catalog drifts.
Must-haves: cost-per-successful-task tracking; per-agent context/token spend visibility; a way to prove ROI on agent features.
Nice-to-haves: automated staleness/drift detection feeding agent context; context pruning policies.
ICP confidence: High — CTO/co-founder, 104 emp, Series C, and an entire first-party engineering post on exactly the agent-cost/context thesis thealpha.ai sells against.
Signal: 4 (senior technical leader at an in-band company actively shipping AI agents; identified via LinkedIn people search "Glean AI agents director engineering"). Source: https://www.linkedin.com/in/garvitjuniwal/ . Company size: ~900 employees (in-band; LinkedIn lists Glean as large-but-<2000); Work AI / enterprise search company that ships Glean Agents (agentic AI) in production. Pain points (inferred from role; no direct post/quote read this run): running an enterprise agent platform means unpredictable per-query LLM cost, token waste on large-context retrieval, and reliability/accuracy at scale. Challenges: controlling agent compute spend while keeping answer quality/latency; cost attribution per agent/customer. Must-haves: per-run cost visibility, guardrails, production observability for a fleet of agents. Nice-to-haves: model routing / prompt-cost optimization across agents. ICP confidence: High (CTO-level technical leader; company in-band and shipping agents; pains inferred not quoted).
Signal: 4 (LinkedIn people search — ICP title "SVP Engineering (AI Agents)" + agents-in-production keywords). Source: https://www.linkedin.com/search/results/people/?keywords=VP%20Engineering%20AI%20agents%20production (profile: https://www.linkedin.com/in/gauravnarasimhan/). Company size: ~50–220 (SEO SaaS; LinkedIn lists 51–100, broader Search Atlas Group ~220). Leads AI agents for SEO, content, paid search and social — multi-agent product surface in production. Pain points (inferred from role/company): scaling multiple production agents across SEO/content/ads/social; per-run cost and reliability of many concurrent agents. Challenges: keeping agent output reliable at customer scale; controlling LLM spend as agent count grows. Must-haves: visibility into cost per agent/per run; reliability guardrails in production. Nice-to-haves: model routing / right-sizing to cut token spend. ICP confidence: High (SVP Eng, agent-native SaaS shipping 5+ agents, 50–2,000 emp).
Signal: 4 (ICP with an explicit agentic-AI mandate, shipping an agent platform). Found via web research — LinkedIn/Chrome unavailable this run. Independently surfaced by TWO separate research passes this run, which is a corroboration signal.
Source: https://www.cognite.com/en/company/newsroom/cognite-announces-new-leaders-to-continue-driving-ai-innovation-and-accelerate-industrial-value ; https://itbrief.news/story/cognite-appoints-ai-leaders-to-drive-innovation-push
Company size: ~898–900 employees (Tracxn 900 as of 30 Jun 2026; Revelio Labs 898 as of Mar 2026, up 44.4% from 622 in 2023; company reported 800+ globally Jan 2026). Lysaker/Oslo, Norway. Growth-stage rather than Series A–C — stage is the ICP stretch.
Agent evidence: Cognite ships Cognite Atlas AI, a low-code industrial agent builder/workbench delivering "specialized industrial agents, virtual employees" for workflow automation and decision support in manufacturing, energy and power generation. Also published an LLM/SLM benchmark report and a "Definitive Guide to Industrial Agents." Background: ex-Google senior SWE on AI for advertising systems; founded Snapsale (acquired by Schibsted); IOI silver medalist.
Verbatim: "I am thrilled to lead the charge for this next phase of AI innovation. Being at the forefront and co-innovating with the global hyperscalers and LLM providers will unlock more value from agentic AI and all industrial data that are uniquely accessible in our platform." His mandate is described as returning "to his disruptive founder roots to extend Cognite's AI technology to deliver more real-time intelligence from any data, on any hyperscaler, and leveraging any LLM at scale."
Pain points: industrial data siloed and not AI-ready so agents cannot reason without it; enterprises generate enormous operational data volumes but struggle to extract insight (per co-executive James Sirota).
Challenges: getting agents to acceptable accuracy on safety-critical industrial operations; making agents work across any hyperscaler and any LLM at scale (model portability/benchmarking); scaling agentic operations for asset-heavy customers with strict reliability requirements; keeping pace with hyperscaler/LLM roadmaps.
Must-haves: AI-ready contextualized industrial data layer under the agents; multi-LLM/multi-cloud flexibility; low-code agent authoring so domain experts can build agents; benchmarked model selection; demonstrable near-term outcomes.
Nice-to-haves: agents embedded in field-ops, maintenance and robotics applications; real-time reasoning over time-series + document + 3D context; human + autonomous agent collaboration models.
Secondary contact at same account: James Sirota, Global Head of Engineering (owns worldwide engineering for Data Fusion and Atlas AI) — worth adding on a future run.
ICP confidence: Medium — textbook title and agent-shipping fit at ~900 employees (in band); growth-stage rather than Series A–C, and the confirming leadership source is 2025 rather than 2026. The "any LLM at scale / model portability" framing is a direct match for Alpha's routing-and-cost story.
George He — Head of Platform Engineering, LlamaIndex
Signal: 1 (ICP engineering leader at an agent-infra company speaking about long-running agent reliability/context). Source: AI Engineer World's Fair 2026 (speaker/session data: https://www.ai.engineer/worldsfair/2026/speakers.json) — talk "Everyone talks about document search, but what about results?" (long-running agents need something more durable than one-off retrieval — reusable work: structured outputs, citations, extracted entities). Company size: ~78–102 employees (2026), Series A — AI-native framework/platform for building agents. Pain points: one-off retrieval is fragile for long-running agents; agents need durable, reusable context rather than repeated work. Challenges: making agent workflows reliable and durable over long horizons; context/token efficiency. Must-haves: durable reusable agent state/context and reliability. Nice-to-haves: structured outputs/citations infra. ICP confidence: High (Head of Platform Engineering at a ~100-person Series A agent-infrastructure company; directly building/scaling agents). LinkedIn not present in source dataset.
George Sivulka — Founder & CEO, Hebbia
Signal: 4 (ICP building/shipping agents, writing about the multi-agent future). Source: https://www.linkedin.com/in/sivulka/ plus podcast "Knowledge Work 2.0: The Company Creating The Multi-Agent Future". Company size: ~150-250 employees (NYC; publicly announced adding ~100 across eng/product/design), Series B, ~$700M+ val (a16z, Thiel). Technical founder (former Stanford PhD, applied math). Hebbia ships AI agents used by BlackRock, Carlyle, Centerview and ~40% of largest asset managers; some jobs process 30B tokens with line of sight to 100B-token jobs in 2026. Pain points: enormous token consumption per job, inference cost + latency at scale, accuracy/trust in high-stakes finance/legal. Challenges: keeping cost and latency down at massive token scale (moved to Baseten stack for cited 10x lower cost, 4x lower latency); guaranteeing accuracy/citations. Must-haves: low-cost reliable inference at 10B-100B token scale, per-job cost visibility, accuracy/grounding. Nice-to-haves: multi-model flexibility. ICP confidence: High — extreme token/cost exposure and production agent scale; clean fit on size and "5+ agents in production".
Gerad Suyderhoud — Sr Director of AI & Automation, Gladly
Signal: Bucket 4-adjacent (ICP AI leader at agent-native CX company). Sourced via LinkedIn people search "Gladly Sidekick AI" after Signal 1-4 engagement searches were saturated with non-ICP/already-captured results. Source: https://www.linkedin.com/in/geradsuyderhoud/ . Company size: ~300-400 employees (Gladly, customer-service platform, ~$1B valuation, Series C/D). Agent-native: owns Gladly Sidekick, an AI customer-service agent shipped to enterprise CX teams; "AI & Automation" org runs multiple production agents. Pain points (inferred from role/company; no public post captured this run): cost per resolution/run of CX agents at scale, reliability/containment of autonomous support agents, measuring agent ROI vs human agents. Challenges: scaling from pilot to many production agents; per-run cost visibility. Must-haves: cost-per-run observability, reliability guardrails. Nice-to-haves: provider portability, agent performance analytics. ICP confidence: Medium-High (title=Sr Director of AI ✓, size ✓, agent-native ✓; no direct engagement signal observed this run).
Giedrius Steimantas — Director of Scraping Engineering, Oxylabs
Signal: 4 (ADAPTED — public conference speaker listing, not LinkedIn; LinkedIn was unavailable this run).
Source: https://www.ai.engineer/worldsfair/schedule — AI Engineer World's Fair 2026, Expo Stage 1 NE, Day 4 (Thu 7/2).
Talk (verbatim): "The Missing Layer in Agentic AI".
Company size: ~482 employees (Tracxn, 30 April 2026; PitchBook 500; GetLatka ~397). Comfortably inside the 50–2,000 band.
Pain points: INFERRED FROM TALK TITLE, NOT A VERBATIM QUOTE — the framing that agentic AI stacks are missing an infrastructure layer maps directly onto our "no operating layer between the agent and the model" thesis.
Challenges: agents that need reliable, high-volume external data access without runaway cost per run.
Must-haves: an infrastructure layer that sits under the agent rather than inside the framework.
Nice-to-haves: cost attribution across the data + inference path.
ICP confidence: Medium — Director of Engineering at a qualifying-headcount company speaking directly to the missing-layer thesis; downgraded because Oxylabs is a data-collection vendor and the talk was an expo-stage slot, so it likely carries vendor positioning rather than a first-person operator pain.
NO LinkedIn URL captured — do not fabricate one.
Gigi Yuen-Reed — Chief Data & AI Officer, Cohere Health
Signal: 4 (ICP AI decision-maker shipping agentic AI in production). Source: https://www.prnewswire.com/news-releases/cohere-health-chief-data--ai-officer-gigi-yuen-reed-phd-named-to-100-women-in-ai-2026-list-302790855.html ; https://www.coherehealth.com/news/fortune-enterprise-ai-healthcare-administration-cohere-health. Company size: ~500–1,000 employees; Cohere Health (Boston; NOT the model lab Cohere) — clinical-intelligence + "clinically trained, agentic AI" for prior authorization / utilization management / payment integrity. 85% of PA requests approved in real time; ~50% administrative cost savings; cited by Fortune (May 2026) as a high-performing enterprise agentic-AI deployment. Note: distinct from Cohere (model lab) already in the People Library — no overlap. Head of AI persona: Gigi Yuen-Reed leads responsible, domain-specific AI systems (20+ yrs healthcare/enterprise AI). Pain points (inferred): reliability + clinical accuracy of agentic decisions under strict regulatory bar; cost/ROI at health-plan scale; auditability/explainability of every agent action for compliance. Challenges: scaling agentic clinical reasoning across payers/providers with trust and governance. Must-haves: reliability + explainability/governance. Nice-to-haves: per-run cost visibility, model routing. ICP confidence: Medium-High (named AI decision-maker; 500–1,000 emp in band; agentic AI in production. Buyer-intent caveat: mature AI org may build in-house).
Gil Barak — Co-founder & CEO (technical/security co-founder, ex-Secdo co-founder), BlinkOps
Signal: 1 (agentic security automation — found via web research; LinkedIn/Chrome unavailable this run; pain inferred from company positioning, no personal quote). Source: https://www.businesswire.com/news/home/20250728640162/en/ ; https://fintech.global/2025/07/29/ai-cybersecurity-startup-blinkops-raises-50m-series-b/ . Company size: 100+ employees. Funding: Series B $50M (2025), ~$90M total. Background: previously co-founded Secdo (EDR, acquired by Palo Alto Networks) — technical/security founding background. Product: agentic security automation / "cybersecurity micro-agents." Pain points (inferred): reliability & trust of agents in production; scaling agent fleets; control/governance. Challenges: enterprise-grade reliability for autonomous security agents. Must-haves: trust, control, governance. Nice-to-haves: per-run cost/observability. ICP confidence: Medium (CEO but technical/security C-suite co-founder; company clearly qualifies).
Gil Feig — Co-Founder & CTO, Merge
Signal: 4 (ICP speaking at AI Engineer World's Fair 2026; Merge positions as 'connective infrastructure for production AI'). Source: https://www.ai.engineer/worldsfair/schedule ; size verified via Tracxn. Company size: ~141 employees (Series B, $75M total, Accel/NEA). Merge ships 'Agent Handler' to govern how agents/employees connect Claude, ChatGPT and other tools to third-party systems. Pain points (inferred): governing and observing agent actions across many integrations, reliability of agent tool-calls, controlling what agents can do. Challenges: making agent<>SaaS connectivity safe and reliable at scale. Must-haves: governance, observability, and control of agent behavior. Nice-to-haves: cost visibility across agent tool usage. ICP confidence: Medium — technical co-founder/CTO at a 141-person Series B expanding into agent infrastructure; strong on size/stage, agent-shipping is emerging vs. core.
Gila Hayat — Co-Founder & CTO, Darrow
Signal: 4 (ICP at a company running always-on production agents at corpus scale). Source: https://www.lawnext.com/2026/05/darrow-the-ai-lab-for-legal-risk-launches-a-platform-to-let-plaintiffs-firms-manage-litigation-like-an-investment-portfolio.html; role confirmed on https://www.darrow.ai/company/about. Company size: ~170 people, stated by CEO Evya Ben Artzi in a May 2026 briefing. Stage: Series B — $35M led by Georgian, Sept 2023; ~$60M total; company says profitable for three years; no Series C as of Aug 2026. Pain points: INFERRED — no first-person statement from Hayat found. What IS verified: "its AI agents continuously analyze publicly available data across industries to surface litigation exposure"; agents test corporations against 164 compliance "legal weaknesses"; ~200,000 plan sponsors and 60,000 funds analysed; direct Anthropic and OpenAI integrations. Critically, Darrow prices on usage — "usage scales with the number of cases, exposures and scans a customer runs" — so usage-based pricing on continuously-running agents structurally forces per-scan unit-cost accounting, and gross margin depends directly on cost per scan. Challenges: continuous always-on agent scanning at corpus scale where margin is a direct function of cost per run. Must-haves: per-run/per-scan cost attribution tied to customer billing; cost control on always-on background agents. Nice-to-haves: model routing across Anthropic/OpenAI by task cost; drift detection across 164 detection templates. ICP confidence: Medium — title, headcount, stage and production agents all verified from primary sources, and the usage-based-pricing + always-on-agents combination is a textbook unit-economics setup; but no first-person pain quote, so the hook is inferred.
Giovanni Casinelli — Co-Founder, CTO & President, Aspire
Signal: 4 (ICP interviewed about shipping agents into production). Found via web research; title VERIFIED live on LinkedIn people search 2026-08-31 — headline reads "Co-Founder, CTO & President @ Aspire (YC W18)".
Source: https://www.frontier-enterprise.com/aspires-early-engineering-bets-on-scale/ (Mar 2026 interview)
Company size: ~1,100+ mid-2026 (Revelio/PitchBook aggregation). Singapore-HQ B2B fintech SaaS, Series C. In band on headcount and stage.
Pain points: Agent sprawl across the finance stack. Describes agents live in production via MCP into ChatGPT/Claude for spend queries, invoice approvals and cash-reserve checks, plus AI spanning onboarding checks through financial-crime triage. Stresses building observability and testing "from day one."
Challenges: Operating a growing number of distinct agents across regulated financial workflows in multiple SEA markets; observability was an explicit early engineering bet, implying he has already felt the visibility gap.
Must-haves: Observability across a multi-agent production estate; testability of agent behaviour before it touches money movement.
Nice-to-haves: Cost attribution per agent/workflow.
ICP confidence: Medium — right person, right company, right stage, agents demonstrably in production. Downgraded from High because no cost-blowout or token-waste quote was found from him; the evidence is production-agent breadth and an observability emphasis, not cost pain specifically.
Signal: 4 (net-new ICP senior technical leader at an agent platform company). Source: https://www.linkedin.com/in/girishahankari/ ; LinkedIn people search (Kore.ai engineering), 2026-07-27 run. Headline: 'EVP Engineering | Agentic AI & ML Expert | Generative AI | SearchAI RAG | AI For Work | 20+ Years'. Company size: ~1,000 employees (mature Series D-stage agent platform) - NOTE larger/later-stage than Series A-C but within 50-2,000 and agent-native. Pain points (INFERRED, not verbatim): platform-wide agent cost + reliability; token/inference spend; per-agent observability. Challenges: scaling and governing many agents across enterprise customers. Must-haves: cost-per-run + control/observability across the platform. Nice-to-haves: vendor-neutral control plane. ICP confidence: Medium-High (EVP Engineering at agentic AI platform; large/mature company).
Signal: 4 (ICP building/shipping agents in production). Source: https://www.linkedin.com/in/gopaldpatel/ ; https://www.qualified.com/ai-sdr ; https://en.wikipedia.org/wiki/Qualified_(company)
Company size: ~160 employees (grew from 17 to 160+); Series C ($95M, 2022 — Sapphire, Tiger Global, Salesforce Ventures).
Company context: Qualified ships Piper, an autonomous AI SDR agent that engages inbound leads and books meetings at scale — a production agent handling high-volume interactions.
Pain points (inferred from role/product): running an autonomous customer-facing agent (Piper) in production at scale; keeping per-conversation cost predictable across many inbound sessions; reliability/quality of autonomous actions.
Challenges: scaling an agent that acts on behalf of customers reliably; visibility into cost-per-run as inbound volume grows.
Must-haves: production reliability and guardrails for an autonomous SDR agent; cost visibility per conversation/run.
Nice-to-haves: agent observability and continuous improvement over time.
ICP confidence: High — technical co-founder & CTO (clear decision-maker), 50–2,000 headcount, actively shipping a flagship production AI agent.
Gopinath Polavarapu — Chief Data & AI Officer / CPTO, JAGGAER
Signal: 1 (ICP author). LinkedIn post ~3 weeks old, 32 reactions.
Source: https://www.linkedin.com/search/results/content/?keywords=%22cost%20per%20agent%22%20production%20engineering&datePosted=past-month
Company size: 1,503 as of March 2026 (Revelio Labs; up from 1,432 in 2023). Other sources range 967–1,503 — all inside the 50–2,000 band. Procurement SaaS, Durham NC. PE-backed, mature (NOT Series A–C — noted as the one deviation).
Pain points (his own words): "Defaulting every request to a frontier model isn't rigor; it's the over-engineering tax — and it's part of the AI Reality Gap between what pilots promise and what production economics allow." / "route every request to the right model, escalate only where the cost of error justifies the premium, and own the routing layer itself." / "how we run this in production at JAGGAER."
Challenges: Production AI economics not matching pilot promises; the majority of production traffic is routine (classification, extraction, routing) yet defaults to frontier models; vendor lock-in constraining cost/performance optimisation.
Must-haves: Owning the routing layer in-house; a model recommendation matrix and routing decision tree; cost, sovereignty and control treated as first-class architecture concerns.
Nice-to-haves: Security framework for evaluating models of any origin; hybrid open-weight/frontier operation.
ICP confidence: Medium — C-level AI owner with the right pain and in-band headcount at a SaaS actively running agents in production, but JAGGAER is a mature PE-backed SaaS rather than Series A–C. BYOK / zero-markup and portable-trace positioning should land well given his explicit "own the routing layer" and sovereignty stance.
Gordon Gibson — Director, Applied Machine Learning, Ada
Signal: 4 (ICP AI/ML leader at agent-shipping company). Source: https://www.linkedin.com/company/ada-cx/people/?keywords=machine%20learning. Company size: 201–500 employees (per Ada/ada CX LinkedIn). Company context: Ada builds and ships customer-service AI agents (conversational AI) in production for enterprises. Pain points (inferred from role/ICP): leading applied ML for production LLM agents — model/agent reliability, evals, and controlling token/inference cost of agents serving high ticket volumes. Challenges: keeping production LLM agent systems reliable and cost-efficient as usage scales. Must-haves: production evals, reliability, and cost/token telemetry per agent run. Nice-to-haves: automated routing/optimization to reduce spend without hurting resolution quality. ICP confidence: High — Director of Applied ML owning production agent ML at an ICP agent company in size band.
Grant Oviatt — Co-Founder & VP of Product, Prophet Security
Signal: 1 (ICP-adjacent co-founder speaking publicly about agent observability/audit trails). Source: https://cisoseries.com/elevating-the-soc-with-prophet-security/ . Company size: ~81 employees (Tracxn 2026), $30M Series A led by Accel — in range. Prophet Security builds an agentic AI SOC (autonomous alert triage/investigation). Pain points: emphasizes an "audit trail that lets you trace every query, piece of evidence, and reasoning step an agent used"; deliberate use of frontier models vs training their own. Challenges: full traceability/observability and reliability of autonomous security agents. Must-haves: agent observability + reasoning-step audit trail. Nice-to-haves: cost/model-choice tuning. ICP confidence: Medium (co-founder/technical decision-maker with a strong observability pain signal; title is VP Product rather than a pure eng/AI title).
Signal: 4 (CTO at an agent-active company; net-new individual — brain already had co-founder Adam Guthrie, this is a different, recently-appointed leader). Source: https://www.linkedin.com/in/pelander/ and https://www.luminance.com/press/greg-pelander-joins-luminance-as-chief-technology-officer-expanding-executive-bench-and-u-s-expertise-at-enterprise-giant/. Company size: ~350-600 (Luminance, Legal-Grade AI for contract analysis + agentic negotiation; Series C). Pain points [INFERRED from role+company, not a verbatim quote]: reliability/accuracy of high-stakes legal agents; governance & compliance for autonomous contract workflows; cost visibility as agent usage scales across enterprise clients. Challenges: scaling engineering + AI org while keeping agents reliable and governable; proving ROI/cost per completed legal task. Must-haves: reliability guardrails + governance + per-run cost visibility. Nice-to-haves: model routing on cost-per-completed-task. ICP confidence: Medium-High (CTO, ex-ClickUp VP Eng Product & AI; agent-active legal company, right size, Series C — clean stage fit).
Guillaume Lample — Co-Founder & Chief Scientist, Mistral AI
Signal: 4 (ICP building/shipping agents) — found via web research; LinkedIn/Chrome extension NOT connected this run; verified via Mistral site, Crunchbase, Agentic List 2026.
Source: https://mistral.ai/about/ ; https://www.crunchbase.com/person/guillaume-lample ; https://www.agentconference.com/agenticlist/2026 (Agent Development Platforms)
Company size: ~500–1,000 employees (Paris; €2B Series C Sept 2025, ~€12B valuation; founded 2023). Ships Le Chat + Mistral AI Studio — a production agent platform with observability, judge models, runtime guardrails, versioning, and hybrid/on-prem deployment.
Role note: Co-founder leading model research (h-index ~41; LLaMA/Mistral/Mixtral lineage; ex-Meta FAIR). Distinct from co-founder/CTO Timothée Lacroix (already in library).
Pain points (INFERRED from domain/role, not verbatim quotes): right-sizing models (SLMs) for cost-efficient agentic production; token/context waste and latency in multi-step agents; giving enterprise builders cost & reliability control when shipping agents.
Challenges: cost-efficient inference at agentic scale; production reliability/observability for customers' agents.
Must-haves: cost-efficient model routing, production observability, guardrails.
Nice-to-haves: per-run cost/latency telemetry across agent steps.
ICP confidence: Medium — technical co-founder / Chief Scientist (Head-of-AI-equivalent) at a 50–2,000-emp AI-native company; a frontier lab, but ships an agent platform (AI Studio) fitting the ICP.
Gurpreet Singh — Director of Solutions Engineering, Agentic AI, Kore.ai
Signal: 4 (company-scoped ICP people search on Kore.ai; found via LinkedIn people search, NOT an observed post — pains INFERRED from role+company). Source: https://www.linkedin.com/search/results/people/?keywords=Kore.ai%20Director%20engineering%20AI%20agents | Company size: ~900 (Kore.ai, enterprise agentic-AI platform — multi-agent, voice, RAG; Series D, NVIDIA/FTV-backed. CAVEAT: Series D, later-stage than core A-C profile, but under 2,000 emp and strongly agent-native). Pain points (inferred): deploying multi-agent systems for enterprise customers; cost/reliability of agents in customer production environments. Challenges (inferred): scaling reliable agents across enterprise deployments; cost attribution per customer/run. Must-haves (inferred): production reliability + per-run cost visibility. Nice-to-haves (inferred): context/token-waste reduction. ICP confidence: Medium (Director level at strongly agent-native company; role is Solutions Engineering = more customer-facing than core product eng; Series D/~900 emp caveat; verified title+company, pains inferred).
Gurtej Gill — VP of AI Operations & Adoption, Canary Technologies
Signal: 1 (ICP at company shipping agent fleet). Source: LinkedIn people search "Canary Technologies VP Engineering" run live this session — result reads "Gurtej Gill — VP of AI Operations & Adoption at Canary Technologies", San Francisco CA, 5K followers. Company size: ~371-374 (Revelio Labs 2026). Series C, $80M June 2025. NOTE: found directly on live LinkedIn this run, not via web research — exact profile URL not captured, left blank rather than guessed. Agent evidence: Canary launched "Canary AI Agent Studio", a hospitality-specific AI agent builder for hotels. Pain points: role title itself ("AI Operations & Adoption") is the strongest possible buying signal for an agent operating layer — this is a dedicated AI-ops owner at a company shipping a multi-agent product. Challenges: driving adoption and operational reliability of agents across a large hotel customer base; no direct quote captured. Must-haves: per-agent operational visibility, cost attribution, reliability monitoring. Nice-to-haves: adoption/usage analytics per agent. ICP confidence: High — VP-level, AI-operations-specific mandate at a Series C company with a shipped agent-builder product. Highest-priority contact at Canary alongside Blake VanLandingham. Also noted at Canary: Ian Clark, Director of Engineering (UK) — secondary target, not added this run.
Gus Iwanaga — VP of Product, Commerce (Founder & GM of mosAIc, agentic AI), commercetools
Signal: 4 (ICP building/shipping agents; AI Engineer World's Fair 2026 speaker). Source: https://www.ai.engineer/worldsfair/schedule — talk "The End of the Static Screen: Architecting Intent-Driven UX with Agentic Orchestration"; also runs commercetools' agentic AI product line. Company size: ~505-588 employees (Series C-stage, ~$1.9B valuation), composable/agentic commerce platform actively building agentic commerce (mosAIc). Pain points: orchestrating multiple commerce agents reliably, intent-driven agentic UX, making retail catalogs agent-readable/buyable-by-AI. Challenges: moving from static screens to agentic orchestration across a large enterprise commerce stack; agent reliability for transactional (purchase) flows. Must-haves: dependable multi-agent orchestration, controllable/observable agent behavior in commerce transactions. Nice-to-haves: cost visibility per agentic workflow, standardized agent-readable catalog layer. ICP confidence: Medium-High — VP of Product (AI-focused) leading the company's agentic AI product at a right-sized Series C commerce company. Note: web research + verification; LinkedIn not connected this run.
Guy Gur-Ari — Co-founder (technical) / Chief Scientist, Augment Code
Signal: 4 (ICP building/shipping agents in production — found via 2026 research on AI coding-agent companies). Source: https://softwareengineeringdaily.com/2025/10/09/scaling-ai-in-enterprise-codebases-with-guy-gur-ari/ , https://www.augmentcode.com/blog/production-ready-ai-remote-agents-now-available-for-all-augment-code-users , https://www.linkedin.com/in/guy-gur-ari (posts). Company size: 176 employees (May 2026); Series B ($252M raised, Sutter Hill / Evolution Equity). Co-founders: Igor Ostrovsky (ex-Pure Storage chief architect) and Guy Gur-Ari (ex-Google AI researcher, PhD physics). Actively building AI agents: Augment ships coding agents incl. cloud "Remote Agents" that run multiple dev tasks concurrently. Pain points (inferred, not verbatim): running coding agents reliably over large enterprise codebases; concurrent cloud agents drive compute/cost scaling; context management. Challenges: agent reliability across the SDLC; long-context over big repos. Must-haves: production reliability, robust context/retrieval. Nice-to-haves: per-task cost visibility. ICP confidence: High (technical co-founder, 176-person Series B whose product is coding agents in production).
Guy Kronenthal — Chief R&D Officer, Papaya Global
Signal: 4 (ICP speaking publicly about coding agents / AI-SDLC). Title verified live on LinkedIn people search 2026-09-02: "Chief R&D Officer | Building cross border payment solutions @Papaya Global", Israel.
Source: https://langtalks.ai/conference — panellist on the "AI-Coding / AI-SDLC" panel. The Org profile confirms he leads a ~200-person engineering org.
Company size: DISPUTED — LinkedIn company page band reads 1K-5K (read live 2026-09-02), while Crunchbase gives 501-1,000 and Unify gives ~818 (Mar 2026). Two of three independent sources put it inside the 50-2,000 band; the LinkedIn band is inconclusive at its upper end. VERIFY HEADCOUNT BEFORE ANY OUTREACH — if actual headcount is above 2,000 this record falls out of ICP.
Pain points: Publicly working the coding-agent / AI-SDLC problem across a ~200-person R&D org — the classic 1->N agent sprawl where many teams adopt agents independently and nobody owns the aggregate spend or telemetry.
Challenges: agent spend spread across many dev teams; proving the agent programme pays back to an exec audience; policy and guardrails at org scale.
Must-haves: org-wide agent spend and usage attribution; policy controls per team.
Nice-to-haves: ROI reporting to the exec team; benchmarking across teams.
ICP confidence: Medium — Chief R&D Officer (well above Director), strong active agent programme, right region. Held at Medium rather than High purely because of the headcount-band conflict noted above, and because Papaya is fintech/HR-payroll rather than AI-native. Net-new name (Papaya Global already in library).
Signal: 4 (ICP building/shipping agents) — found via web research; LinkedIn/Chrome extension NOT connected this run; verified via CTech, SecurityWeek, company site.
Source: https://www.calcalistech.com/ctechnews/article/ryb30q8bge ; https://www.securityweek.com/mate-security-raises-35-million-for-agentic-soc/
Company size: ~48 employees (Tel Aviv; ~$50M total raised incl. $35M Series A 2026; founded 2025; plans to roughly double headcount by year-end). Open platform for "agentic security operations" — AI agents that detect, investigate, respond to, and hunt threats.
Role note: Co-founder & CTO; founding team are Wiz/Microsoft/Meta veterans (CEO Asaf Wiener, CPO Oren Saban).
Pain points (INFERRED from domain/role, not verbatim quotes): reliability/trust of autonomous SOC agents acting on security events; cost of running investigation/hunting agents continuously at alert scale; avoiding runaway multi-step agent loops.
Challenges: production reliability & correctness of autonomous security actions; per-run cost control as agent volume scales with alerts.
Must-haves: production reliability, guardrails, cost predictability per run.
Nice-to-haves: per-agent/per-run cost & action observability.
ICP confidence: Medium — CTO/technical co-founder actively building AI agents; headcount (~48) sits just below the 50-emp floor but is growing fast (doubling planned), so flagged rather than excluded.
Guy Podjarny — Founder & CEO (technical; ex-Snyk founder, ex-Akamai CTO), Tessl
Signal: 4 (ICP building agents) + Signal 1 (writes/speaks publicly about agents, context and cost). Source: https://tessl.io/blog/announcing-tessls-products-to-unlock-the-power-of-agents/ ; https://www.gv.com/news/guy-podjarny-interview-tessl-ai-software-development
Company size: ~60 employees (May 2026, growing ~50% YoY); Series A ($125M at $750M valuation — Index, Accel, GV, Boldstart).
Company context: Tessl builds an "Agent Enablement Platform" / AI-native, spec-centric development — giving coding agents structured, versioned context so autonomous agents can build and evolve software reliably. Guy hosts the "AI Native Dev" podcast and speaks/writes extensively about agents.
Pain points (inferred from product thesis): agents waste tokens/context without structured grounding; agents are unreliable without versioned context; hard to make autonomous agents production-worthy.
Challenges: giving agents the right context efficiently; controlling agent behavior at scale; reliability of agent-generated software.
Must-haves: structured context/grounding for agents; reliability and control over autonomous agent behavior.
Nice-to-haves: cost efficiency of agent context; observability into what agents do.
ICP confidence: Medium-High — highly technical founder/CEO and decision-maker, 50–2,000 headcount, company's entire thesis is agent enablement (context/reliability/cost). Note: profile_url is company page (personal LinkedIn not confirmed).
Signal: 3 (ICP engaging with competitor/observability content — Guy featured in a Braintrust-related LinkedIn post on AI Gateway observability; authored Lucidworks blog 'When AI Agents Fly Blind'). Source: https://www.linkedin.com/in/guysperry/ ; https://lucidworks.com/blog/when-ai-agents-fly-blind-why-your-agentic-platform-needs-precision-search. Company size: ~250-500 employees (enterprise search/AI; NOTE stage is beyond Series A-C). Pain points: agents 'flying blind' without precision search/grounding; agent reliability & accuracy in production; grounding agents in enterprise knowledge. Challenges: getting agentic platforms to retrieve the right knowledge; agent precision at enterprise scale. Must-haves: agent grounding/observability, retrieval quality. Nice-to-haves: commerce/merchandising agent tuning. ICP confidence: Medium (CTO at right-sized company actively shipping AI agents; company stage beyond Series A-C). [Supersedes placeholder row id 371.]
Haixun Wang — VP of Engineering & Head of AI, EvenUp
Signal: 4 (ICP leading agent development shipping into production). Source: https://www.evenuplaw.com/blog/document-ai (Building Trustworthy and Scalable Document AI for Legal Tech) + https://www.evenuplaw.com/blog/ai-communication-agents-capacity-growth/ (AI Communication Agents expanded operational capacity 2.5x in first 90 days). Company size: ~300-500 (est.; Series E $150M Oct 2025, $2B valuation). Pain points: shipping trustworthy/scalable document + communication agents across the full case lifecycle (claims initiation, liability verification, treatment monitoring, record retrieval, balance confirmation); reliability and trust for legal-grade output. Challenges: scaling multiple agent workflows reliably in a regulated domain; document understanding + generation quality. Must-haves: trustworthiness/auditability, reliability at scale, systems of intelligence + action. Nice-to-haves: proactive case progression, compounding operational leverage. ICP confidence: High (VP Eng / Head of AI at a mid-size company actively shipping 5+ agent workflows in production). Note: personal LinkedIn URL not confirmed in search; profile_url points to authoritative company engineering blog authored under his org.
Signal: 3/4 (ICP technical AI leader at company shipping an agent-building platform + adding cost controls — overlaps competitor/observability space; found via LinkedIn People search). Source: https://www.linkedin.com/search/results/people/?keywords=VP%20of%20AI%20shipping%20agents%20startup ; verification: https://www.algolia.com/products/ai/agent-studio and https://www.hpcwire.com/bigdatawire/this-just-in/algolia-adds-governance-cost-controls-to-agent-studio/. Company size: 500+ employees, Palo Alto (AI-native search infra; later stage / ~Series D — caveat vs A-C band). Actively shipping agents (CONFIRMED): Algolia Agent Studio — "build production-ready AI agents in weeks"; July 2026 added governance + COST CONTROLS to Agent Studio; orchestrates ~1.75T queries/yr. Hamish leads AI + search infrastructure (founder of Sajari, acquired by Algolia). Pain points (directly relevant to thealpha): cost control + governance for agentic experiences at scale; making agents production-ready fast; visibility into what agents cost. Challenges: scaling reliable, cost-governed agentic search across enterprise customers. Must-haves: per-run cost visibility/controls, governance, reliability. Nice-to-haves: deeper observability/optimization. ICP confidence: Medium (VP AI at 500+ emp AI-native co shipping an agent platform with explicit cost-control need; stage later than Series A-C). LinkedIn profile URL not captured this run — do not fabricate.
Hamza Sayah — Co-Founder & CTO, Qevlar AI
Signal: 4 (web-research proxy — LinkedIn/Chrome unavailable this run; found via Mar 2026 funding announcement + company profile). Source: https://www.unite.ai/qevlar-ai-raises-30m-to-transform-security-operations-centers-with-autonomous-ai/ ; https://www.eu-startups.com/2026/03/paris-based-qevlar-ai-raises-e25-8-million/ ; headcount via https://tracxn.com/d/companies/qevlar-ai . Company size: ~70 employees (Tracxn, May 2026; PitchBook ~63); Paris; ~$30M raised (Partech, Forgepoint, EQT); founded 2023. Co-founder Ahmed Achchak (CEO, ex-Datadog Cloud Security PM); both founders are AI engineers.
Pain points (from company positioning): autonomous SOC agents must investigate EVERY alert at tier-1 depth and correctly decide malicious vs benign — reliability/accuracy of agents in production is existential; investigations cut to ~3 min.
Challenges: agent accuracy at scale (false pos/neg), trust in autonomous decisions, cost/throughput per investigation.
Must-haves: reliable, auditable autonomous agent decisions; scale to full alert volume.
Nice-to-haves: cost-per-investigation visibility.
ICP confidence: High — clear CTO/technical co-founder, confirmed 50-2,000 headcount, shipping autonomous agents in production (security).
Hannes Hapke — Director, 575 Lab (Open Source Office), Dataiku
Signal: 1 (ICP speaking/writing about agent reliability & explainability — QCon AI Boston 2026 speaker on "opening the black box" of AI agent tool selection; directs Dataiku's 575 Lab, which is releasing open-source Agent Explainability Tools to trace decision-making across multi-step agent workflows).
Source: https://infoq.com/news/2026/03/qconai-boston-2026-talks ; Dataiku 575 Lab launch https://www.dataiku.com/company/news/dataiku-launches-575-lab
Company size: est. ~1,000–1,200 employees (Dataiku). Within 50–2,000.
Pain points: agent explainability; tracing/understanding why agents select specific tools; failures propagating downstream through multi-step agent workflows; governance/trust of agentic AI.
Challenges: visibility into agent decision paths (not just output logs); reliability and governance as agentic workflows grow more complex/autonomous.
Must-haves: agent observability, explainability, governance/trust tooling.
Nice-to-haves: standardized open-source trust infrastructure and protocol support (e.g., MCP); Agentic AI Foundation member.
ICP confidence: Medium — Director-level AI leader at a qualifying-size company actively building agentic-AI tooling; Dataiku is a more established enterprise data-science platform (agent-adjacent tooling), hence Medium.
NOTE: No verified LinkedIn URL captured (profile_url left blank). Discovered via web/conference fallback — Claude-in-Chrome/LinkedIn browsing was not connected during this scheduled run.
Hao Liu — Director of Engineering, Decagon
Signal: 4 (net-new ICP senior technical leader at an agent-native company; company founders already in brain, this is a distinct Director-level add). Source: https://www.linkedin.com/in/haoericliu/ ; LinkedIn people search (Decagon engineering), 2026-07-27 run. Headline: 'Director of Engineering @ Decagon | Ex-Block, Lyft, AWS, MSFT'. Company size: ~200-400 employees; Decagon raised ~$481M (Series D, $250M Series D Jan 2026) — NOTE stage is past the Series A-C guideline, retained on size + agent-native fit. Decagon = enterprise CX agents across chat/voice/email/SMS in production. Pain points (INFERRED from role+company, not verbatim): per-conversation / per-agent cost at high volume; production reliability/accuracy of CX agents; retry and token waste. Challenges: scaling many concurrent agents reliably; measuring cost per resolved conversation. Must-haves: cost-per-run visibility + reliability guardrails. Nice-to-haves: honest model comparison on cost-per-completed-task. ICP confidence: Medium-High (Director of Engineering, right size, agent-native; stage past Series C).
Hari Poludasu — Vice President, Engineering, Kore.ai
Signal: Targeted ICP people-search — VP Engineering at an agent-native platform (Kore.ai) not yet represented by this individual (brain has other Kore.ai leaders; Hari is distinct). Source: https://www.linkedin.com/in/hari-poludasu-86b07154/ (LinkedIn people search). Company size: ~600-1000 (within 50-2000). Pain points (ICP-inferred; no direct post captured this run): scaling many enterprise agents reliably in production; cost/token control across large multi-tenant deployments; per-agent-run observability. Challenges: reliability + governance at enterprise scale; visibility into agent cost per run. Must-haves: production observability, cost visibility, reliability. Nice-to-haves: automated cost guards / anomaly detection. ICP confidence: High (VP Engineering, agent-native platform, 50-2000).
Signal: 4 (ICP AI leader building/shipping AI at an agent-adjacent, AI-native company; found via LinkedIn people search for ICP titles + 'agents production'). Source: https://www.linkedin.com/in/hariprasad-p-s-162b0778/ ; company headcount verified via Revelio Labs/Tracxn (https://www.reveliolabs.com/companies/hyperverge/employees). Company size: ~400 employees (401-402 as of mid-2026; grew from 232 in 2023). AI-native identity verification (KYC/KYB) SaaS, has processed 1B+ identities; moving into agentic onboarding/automation. Pain points (inferred from role/company): scaling AI-driven document/identity workflows reliably in production; controlling model/inference cost as volume grows to billions of checks; accuracy + latency of AI decisioning at scale. Challenges: keeping fraud-detection AI reliable and explainable across geographies; per-transaction unit economics. Must-haves: production-grade reliability, cost visibility per run/transaction, monitoring of AI decisions. Nice-to-haves: cross-model routing, automated evals. ICP confidence: Medium-High (clear Head-of-AI target title; AI-native company in 50-2,000 range; agent-shipping is emerging/agentic-automation rather than fully confirmed 5+ discrete agents).
Harjot Gill — CEO & Co-Founder (technical; ex-Netsil founder), CodeRabbit
Signal: 4 (technical co-founder building/shipping agents) + 1 (podcast guest on RAG-for-code-review architecture). Source: https://www.linkedin.com/in/harjotsgill/ ; https://softwareengineeringdaily.com/2025/06/24/coderabbit-and-rag-for-codereview-with-harjot-gill/ ; https://www.businesswire.com/news/home/20250916401011/en/ (Series B) ; https://businesshonor.com/2026/04/coderabbit-ai-agent-sdlc. Company size: ~213 employees (May 2026); Series B $60M (Sep 2025, Scale Venture Partners + NVIDIA), $88M total; 20k+ customers. CodeRabbit ships AI code-review agents and is expanding to multi-agent SDLC operations. Pain points: agent token cost at very high volume (code review across 20k customers / large repos); reliability/accuracy of review agents (reducing false positives/hallucinations); orchestrating multiple specialized agents across the SDLC. Challenges: scaling multi-agent workflows cost-effectively; deep repo context (RAG) at scale. Must-haves: cost control per review/agent, reliable agent output. Nice-to-haves: cross-agent observability. ICP confidence: High — technical co-founder/CEO, ~213 emp Series B, clear agent product at large production volume; direct cost + reliability pain.
Signal: 4 (adapted method — found via LinkedIn company People-page directory of an in-band agent-native company; LinkedIn content/post search was low-yield this run, consistent with prior runs). Source: https://www.linkedin.com/company/perplexity-ai/people/ and https://www.linkedin.com/in/harrisonwong/. Company size: 201-500 employees (LinkedIn header). In-band, agent-native (ships answer engine, Comet browser agent, Deep Research agent in production). Pain points (INFERRED from role+company, no quotes): per-query/per-agent-run LLM cost at consumer scale (agentic answer engine + Comet agentic browser + Deep Research agent, millions of runs); token/context efficiency; reliability & citation accuracy of agent outputs; latency; near-zero per-run/per-agent cost visibility. Challenges: keeping unit economics sane as agent usage compounds; orchestrating multi-step search->browse->research agents reliably at scale. Must-haves: per-run/per-agent cost observability; reliability guardrails at massive scale. Nice-to-haves: model routing/cost optimization; agent governance. ICP confidence: High (VP Eng at in-band agent-native company).
Harshil Shah — Head of Agentic AI, Rush Street Interactive (RSI)
Signal: 4 (ICP leading agent development; LinkedIn headline: 'Head of Agentic AI @ RSI | AI Agents for SaaS | AI Agents for business functions'). Source: https://www.linkedin.com/in/harshil-shah-76b24a146/ (found via LinkedIn people search 'Head of AI agents production'). Company size: ~912 employees (public online gaming/sports-betting operator, NYSE:RSI; actively building internal AI agents across SaaS & business functions). Pain points (inferred from role/headline, not a verbatim quote): standing up multiple production agents across business functions; reliability & governance of agents in a regulated environment. Challenges: scaling from pilots to many production agents; cost/latency control across business-function agents. Must-haves: visibility into agent performance and cost per run; reliability guardrails. Nice-to-haves: centralized agent observability/orchestration. ICP confidence: Medium (title and agent-building mandate are strong; company is a gaming operator building internal agents rather than an agent-native product vendor).
Haseeb Khan — VP of Engineering, Platform & AI, Conga
Signal: 4 (LinkedIn people search — VP of Engineering + Agentic AI, agents-in-production). Source: https://www.linkedin.com/search/results/people/?keywords=VP%20Engineering%20AI%20agents%20production (profile: https://www.linkedin.com/in/haseebkhan11/). Company size: ~1,800 (LeadIQ 1,001–5,000; RocketReach ~2,153 — near the 2,000 ceiling, flag). Conga = CPQ/CLM/Document Automation SaaS, "AI-powered innovation"; leads AI/ML and Agentic AI platform across AWS + Azure. Pain points (inferred from role/company): standing up agentic AI across a large SaaS platform on multi-cloud; reliability + cost of agents embedded in CPQ/CLM workflows. Challenges: reliability and spend control as agentic features roll out to 10k+ customers. Must-haves: production observability + cost-per-run for platform agents; multi-cloud consistency. Nice-to-haves: model routing / spend optimization. ICP confidence: Medium (VP Eng Platform & AI, SaaS building agentic AI; headcount sits at the upper edge of the 50–2,000 range).
LinkedIn: https://www.linkedin.com/in/hassaanraza. Signal: 4 (ICP building/shipping conversational video AI agents; API-first CVI in production). Source: https://sacra.com/research/hassan-raza-tavus-ai-avatar-developer-platform/ ; https://tracxn.com/d/companies/tavus/. Company size: ~91 employees (May 2026; $64.2M raised, Sequoia/Scale/YC). Pain points: running real-time, low-latency face-to-face conversational agents reliably at scale; emotionally-intelligent adaptive interactions that stay consistent. Challenges: reliability and cost-efficiency of real-time multimodal agents delivered via API to many products. Must-haves: low-latency reliability, per-conversation cost visibility. Nice-to-haves: observability into multimodal agent behavior. ICP confidence: Medium-High (technical co-founder/CEO, ~91-employee company shipping agents in production; some pain points inferred from product/company signals rather than a direct quote).
Signal: 4 (web-research pivot — Chrome/LinkedIn NOT connected this run, so ICP technical leaders sourced via funding announcements + primary-source verification). Source: https://techcrunch.com/2026/06/15/malaysias-respond-io-raises-62-5m-eyes-acquisitions-in-north-america-and-europe/ ; https://www.bizpreneurme.com/company-spotlight-hassan-ahmed-cto-at-respond-io/ . Company size: ~196 employees (PitchBook/Tracxn, mid-2026); Series B, $62.5M raised Jun 2026; KL, Malaysia. Actively shipping AI agents: customer-conversation AI agents that qualify leads, update CRMs, route chats and recommend products across WhatsApp, TikTok, Instagram, Facebook and voice for 10,000+ brands in 180+ countries. Pain points (inferred from product + CTO role): reliability/consistency of agents at very high conversation volume across many channels; per-conversation LLM cost control at 10,000+ brand scale; security/stability with every release. Challenges: maintaining production reliability across multi-channel, multi-language deployments. Must-haves: production reliability, predictable cost per conversation. Nice-to-haves: unified cross-channel observability. ICP confidence: High (Co-founder/CTO at a 50-2,000-employee company actively shipping customer-facing AI agents).
Signal: 4 (ICP VP-Eng at a company building/shipping AI automation agents). NOTE ON METHOD: LinkedIn browsing unavailable this run (Chrome extension not connected); found/verified via web research (May 2025 appointment press release). Pain points INFERRED from Quandri's domain, not verbatim quotes.
Source: https://markets.financialcontent.com/minstercommunitypost/article/bizwire-2025-5-15-quandri-appoints-heather-natour-as-vp-of-engineering-to-drive-product-innovation
Company size: ~50-90 (Series A AI-automation platform building digital-worker/agent bots for insurance-brokerage personal-lines renewals). NOTE: near the lower 50-emp bound — confirm headcount before outreach.
Pain points (inferred): scaling automation/agent bots reliably across brokerage workflows; accelerating AI development investment.
Challenges (inferred): moving from single automations to a broader set of production agents; reliability of unattended automation.
Must-haves (inferred): reliability + monitoring for production automation agents.
Nice-to-haves (inferred): cost/throughput visibility per automation run.
ICP confidence: Medium — clean VP-of-Engineering ICP title at an agent/automation company; main caveat is company size near the 50-emp floor.
Heeren Pathak — Chief Technology Officer, Gradient AI
Signal: 4 (CTO at an insurance AI company scaling automated underwriting/claims decisioning). Source: https://www.gradientai.com/leadership-team/ (site updated Jul 2026) ; funding: https://www.gradientai.com/press-gradient-ai-secures-56-million-in-series-c-funding-to-expand-ai-powered-insurance-solutions
Company size: 101-250 (Crunchbase); 51-200 (LeadIQ); ~120 estimated Mar 2026 — in range. Series C, $56.1M led by Centana Growth Partners (Jul 2024).
Profile URL: not found — do not fabricate.
Pain points: Accuracy-versus-speed trade-off in automated underwriting and claims decisioning; long implementation cycles ("from start to finish takes about 90 days") for getting automated decisioning live at an insurer. Stated goal is predictions that make underwriting "better, faster and more accurate."
Challenges: Delivering explainable, regulator-defensible automated decisions at scale for insurers; shortening time-to-production for each new deployment.
Must-haves: Reliable, explainable model/agent outputs in a regulated setting; faster deployment loops per customer.
Nice-to-haves: Cost-per-decision visibility; automated drift/regression monitoring across insurer deployments.
ICP confidence: Medium — CTO title and headcount verified and in range; downgraded because (a) the pain evidence is decisioning/ML-framed rather than explicitly agent-cost or agent-reliability, and (b) PitchBook flags an unconfirmed "acquisition" event dated 2026-03-03 with no named acquirer. RE-VERIFY company independence and his current role before any outreach.
Helen Greul — SVP Engineering, PolyAI
Signal: 4 (senior engineering leader at an AI-native company shipping agents; found via LinkedIn people search of PolyAI).
Source: https://www.linkedin.com/search/results/people/?keywords=PolyAI%20head%20of%20engineering%20director%20AI
Company size: ~250-350 employees; PolyAI raised $200M+ (Series D, Dec 2025, $750M valuation, ~$40M ARR); enterprise voice AI agents. NOTE: technically Series D (slightly past the A-C guideline) but firmly in the 50-2,000 band and AI-native, shipping voice agents at scale.
Pain points (inferred from role/company context; no direct post captured this run): scaling AI/platform/infra & security teams, agent reliability and per-call cost at production voice volume.
Challenges: infra scaling, developer experience across an agent fleet, reliability.
Must-haves: production reliability, cost control at call volume.
Nice-to-haves: better observability / DevEx.
ICP confidence: Medium-High — SVP Engineering (ICP), in-band size, agent-native; stage past Series C noted.
Signal: 1/4 (adapted — LinkedIn auth unavailable this run; found via web research/verification). Source: https://tracxn.com/d/companies/sully/ ; https://builtin.com/job/head-engineering-scribe/ ; https://www.sully.ai/. Company size: ~60–70 employees. Head of Engineering at Sully.ai (previously founded Diagnosis AI, acquired by NASA 2016). Sully ships multiple production healthcare agents (scribe, receptionist, voice) integrated to EHRs. Pain points (inferred from role + product): owning reliability and scaling of a multi-agent production stack in healthcare; controlling LLM/token spend on high-volume transcription and voice workloads. Challenges: agent reliability at clinical bar; observability into what each agent costs per encounter as clinician count grows (20k+ clinicians reported across the category). Must-haves: production reliability + cost-per-run visibility. Nice-to-haves: standardized eval/guardrail tooling across agents. ICP confidence: Medium-High — Head of Engineering title (senior technical, above Director), agent-native company in size range. LinkedIn URL not captured this run.
Henry Ehrenberg — Co-Founder (focused on technical strategy and engineering), Snorkel AI
Signal: 3 (Snorkel AI is a Portkey customer with a documented multi-agent debugging case study) + 4 (leads public work on agent evaluation).
Source: name, title, bio and LinkedIn hyperlink verified on Snorkel's own author page https://snorkel.ai/author/henry-ehrenberg/ ; agent-eval work https://snorkel.ai/blog/senior-swe-bench-evaluating-coding-agents-like-senior-engineers/ ; company agent positioning https://snorkel.ai/company/ and https://snorkel.ai/blog/enterprise-environments-ai-agents/ ; Portkey case study https://portkey.ai/case-studies/snorkel-ai-multi-agent-debugging
Company size: Sources conflict badly. TrueUp/Workforce.ai: ~700 (+43% YoY, https://www.trueup.io/co/snorkel-ai). Revelio Labs: 797 (2025) / 776 (Jul 2026). BUT Snorkel laid off 31 people described as 13% of workforce in Sept 2025 (Business Insider), which implies a core workforce of ~240 — the tracker figures likely include contracted expert/annotator networks. Every reading is inside 50–2,000, but treat ~240 core employees as the more defensible number. Other context: $1.3B valuation Aug 2025, latest secondary valuation ~$0.7B (down 46%).
Pain points: No verbatim quote from Ehrenberg. Strongest documented pain is the Portkey case study on Snorkel's own multi-agent system: "Prior to Portkey, Snorkel debugged LLM applications with a fragmented patchwork of logs in s3, DB queries, and manual visualization." It also documents a runaway-cost incident — the Planner "received a chart with no question, decided to search the web for context, and spiraled into 38 calls chasing information that didn't exist." That is a textbook agent cost-blowout with no per-run visibility. NOTE: the Portkey case study was surfaced by one research pass and NOT independently re-verified in the verification pass — confirm before quoting in outreach.
Challenges: Runs a Planner → Code Agent → Verdict multi-agent architecture with sub-agents and tool calls, processing thousands of completions per week (Multi-Agent Question-Answer Validator). Also leads Senior SWE-Bench, an open benchmark for evaluating coding agents on realistic senior-engineering work — so agent evaluation is his personal remit.
Must-haves: Consolidated agent tracing to replace the s3-logs-plus-DB-queries patchwork; runaway-loop detection and cost ceilings per agent run; evaluation infrastructure for multi-step agents.
Nice-to-haves: Per-agent cost attribution across the internal agent fleet; reduced debugging time on non-deterministic multi-agent failures.
ICP confidence: Medium — technical co-founder is a valid title target and the multi-agent cost-blowout story is the single sharpest wedge found this run. Downgraded from High on two counts: (1) verification found NO source showing Snorkel itself operates customer-facing production agents — their product is the data/eval layer underneath other people's agents (Snorkel Evaluate, Expert Data-as-a-Service), so the internal agent fleet is the real hook, not the product; (2) headcount is genuinely ambiguous (240 vs 700–800) and the company is in a down-round/layoff posture, which may suppress new vendor spend. Use Snorkel's own wording "Co-founder" — theorg.com's "Co-Founder & Head of Engineering" is LinkedIn-derived and uncorroborated.
Signal: 4 (technical decision-maker at a company building/shipping agents in production). Source: https://www.linkedin.com/in/henrytpeter/ and https://ushur.ai/ai-agents. Company size: ~213-240 employees (Series C, $108M raised, founded 2014). Actively building AI agents: yes — Ushur's platform deploys AI Agents for regulated CX automation (healthcare, insurance, finance). Pain points: securing and controlling autonomous agents (he is also CISO) while keeping them reliable in regulated production environments. Challenges: governing agent behavior at scale; balancing security, reliability, and cost of agents in production. Must-haves: control + observability over what agents do; secure, compliant agent execution. Nice-to-haves: per-run cost visibility and guardrails. ICP confidence: High (co-founder/CTO, Series C, ~220 employees, agent-native platform). Note: pain points inferred from role (CTO+CISO) and company context, not a verbatim quote.
Signal: 4 (ICP writing/speaking about shipping agents) + 3 (LangSmith competitor engagement).
Source: https://www.langchain.com/blog/how-apollo-rebuilt-its-ai-assistant-on-deep-agents-to-power-the-full-gtm-loop and https://interrupt.langchain.com/nyc/agenda (Interrupt NYC, 24 Sep 2026 — talk: "From Supervisor to Deep Agents: Re-Architecting a Production GTM Assistant Without Breaking Trust").
Company size: ~800 (Revelio Labs 2026 ~800; RocketReach 826; note LeadIQ shows 147 — sources disagree, treat ~800 as best estimate). Series D GTM SaaS, agents shipped in the core product.
Pain points: LangChain writeup states Apollo "began using LangSmith after hitting the limits of its own observability tooling. The original multi-agent architecture made it nearly impossible to understand what was happening inside any given thread — which tools were called, in what order, with what latency, and where things went wrong." Also built a CLI specifically "to reduce the context-size costs of MCP-based interactions."
Challenges: Re-architecting a live production agent from a supervisor pattern to Deep Agents without breaking user trust; context bloat from MCP tool surface; homegrown observability that stopped scaling.
Must-haves: Per-thread trace visibility (tool call order, latency, failure point); control over context size per agent run.
Nice-to-haves: Structured eval framework layered on top of tracing (Apollo built an internal "AI Watchtower" 6-layer eval framework).
ICP confidence: High — exact title match, company squarely in 50–2,000 band, publicly documented agent-observability and context-cost pain in his own org.
Caveat: The observability/cost pain in the LangChain article is attributed to the Apollo team, not quoted from Gahlot personally. Do not attribute the quote to him directly in outreach.
Signal: 4 (web-research pivot — Chrome/LinkedIn NOT connected this run). Source: https://www.kapture.cx/press-releases/kapture-cx-appoints-himanshu-garg-as-chief-technology-officer/ ; https://www.bwpeople.in/article/kapture-cx-elevates-himanshu-garg-as-cto-503313 . Company size: ~434-594 employees (501-1000 band; Tracxn/PitchBook 2026); Pre-Series B; founded 2014, Bengaluru. Actively shipping AI agents: agentic enterprise CX / customer-support automation stack. Promoted from SVP Engineering to CTO (May 2023 join); prior CTO at Via.com, ex-Google engineering, YC W20 founder (Bizrise). Pain points (inferred from agentic-CX product + CTO role): scaling agentic support automation across enterprise clients while holding reliability; controlling LLM cost per resolution at enterprise volume. Challenges: moving CX agents from pilots to reliable production across many enterprise accounts. Must-haves: production reliability, cost-per-run visibility. Nice-to-haves: agent observability and guardrails. ICP confidence: High (CTO at a 500-1,000-employee company shipping agentic CX agents).
Himanshu Walia — Director, Integrations, Hippocratic AI
Signal: 4 (ICP-adjacent technical leader at validated target company Hippocratic AI; sourced via LinkedIn People search, run 2026-08-01). Source: https://www.linkedin.com/in/himanshu-walia-3327bb115/ . Company size: ~300-500 (Hippocratic AI — healthcare voice agents; validated in-range). Headline: "Building Teams, Platforms & Processes to Scale Healthcare Innovation." Pain points (role/company-contextual, not from an individual post this run): scaling healthcare-agent integrations and platforms into production systems; reliability across integrated agent workflows; observability/cost of agents spanning multiple systems. Challenges: platform/process scaling as agent count grows. Must-haves: reliability + cost observability across integrated agents. Nice-to-haves: standardized eval across integrations. ICP confidence: Medium (Director-level at in-range agent-native company, but integrations/platform-focused rather than core AI/model engineering).
Signal: 4 (ICP shipping agentic workflows in production). Source: https://supplychaindigital.com/digital-supply-chain/altana-achieves-unicorn-status ; https://www.cbinsights.com/company/altana-trade/people. Company size: ~270 employees (2026, up from 148 in 2023); Series C $200M, $1B valuation (unicorn); global trade / value-chain AI platform. Pain points (INFERRED): powering agentic trade and full brokerage workflows across a massive value-chain data platform; reliability of multi-step agents automating enterprise trade/compliance; cost/scale of agentic workflows on large supply-chain data. Challenges: reliable multi-step agent automation across enterprise workflows; observability + cost of agents running on large data. Must-haves: production reliability + per-run cost/behavior visibility for agentic workflows. Nice-to-haves: model/routing cost optimization. ICP confidence: High (CTO; Series C; ~270 emp in-band; agentic workflows in production).
Ian Chan — VP of Engineering, Postscript
Signal: 3 (ICP engaging with competitor/adjacent tooling — named customer testimonial for Freeplay, an agent eval/observability vendor). Found via web research — Claude-in-Chrome/LinkedIn NOT connected this run.
Source: https://freeplay.ai/ (named customer testimonial block)
Company size: ~286–309 employees (Tracxn: 286 as of 30 Jun 2026; LeadIQ: 309). Series C — $65M led by 01 Advisors (Twilio Ventures, Greylock, Accomplice, OpenView, Elephant participating); $117M total across 4 rounds. SMS/conversational commerce for Shopify brands; ships a customer-facing conversational agent ("Shopper", postscript.io/ai) that responds to shoppers in under a minute at high volume.
LinkedIn URL: https://www.linkedin.com/in/ianchan1/
Pain points: Verbatim — "Freeplay transformed what used to feel like black-box 'vibe-prompting' into a disciplined, testable workflow for our AI team. Today we ship and iterate on AI features with real confidence about how any change will impact hundreds of thousands of customers." No repeatable way to know whether a prompt/model change helps or hurts; large blast radius of AI changes; AI iteration bottlenecked on a narrow set of engineers.
Challenges: Regressions are revenue-affecting (conversational commerce), and per-conversation economics matter at tens of thousands of conversations/month.
Must-haves: Regression testing and experiments on prompt/model changes before release; production observability tied back to the eval suite; enough confidence to ship to a large live user base.
Nice-to-haves: Non-engineers (PM/support/ops) able to review and label AI interactions; model-swap comparison; dataset curation from production logs.
ICP confidence: Medium — clean title/size/stage fit (VP Eng, ~290 people, Series C, customer-facing agent in production), but his verified public signal is about eval discipline and release confidence rather than explicitly about cost or token waste. Cost angle is inferred, not quoted.
Signal: 4 (ICP technical co-founder shipping agents in production). Source: https://www.qventus.com/company/our-team/ ; https://tracxn.com/d/companies/qventus/. Company size: ~206–238 employees (Mar 2026); total funding ~$203M; Series D ($105M led by KKR, Jan 2025), ~$400M+ valuation. Qventus ships AI "operational assistants"/agents that automate hospital and health-system operations (capacity, discharge/patient flow, scheduling) using ML + generative/agentic AI. Pain points (inferred from domain): production reliability in a regulated clinical-ops setting where wrong actions affect patient care; cost/ROI justification as agents expand across hospital workflows; visibility/governance over what automated agents do. Challenges: scaling agentic automation from a few workflows to fleet-wide hospital operations while maintaining trust. Must-haves: reliability + auditability/governance of agent decisions. Nice-to-haves: per-run cost visibility, model routing. ICP confidence: Medium-High (named technical co-founder/CTO; ~238 emp squarely in band; agents in production. Caveat: Series D is one stage past the stated A–C band, but headcount and agent-native profile fit).
Ian Nowland — Head of Hardware Infrastructure, Baseten
Signal: 4 (infra leader at agent/inference-infra company). Discovery: company-scoped LinkedIn People search. Source: https://www.linkedin.com/in/inowland/ | Company: Baseten — AI inference platform for production & agentic workloads. Company size: ~312 employees (PitchBook 2026) — in band. Pain points (INFERRED from role): GPU/hardware cost efficiency for inference, reliability at scale, cost per run. Challenges: hardware cost governance for high-volume inference/agent workloads. Must-haves: cost/utilization visibility, reliability. Nice-to-haves: optimization. ICP confidence: Medium (Head of Hardware Infrastructure — narrower infra fit; in-band).
Signal: 1 (engineering leader publicly discussing agent/AI-feature reliability & evaluation in production). Source: https://www.unite.ai/iccha-sethi-vice-president-of-engineering-at-vanta-interview-series/ ; https://jam.dev/blog/how-vanta-builds-with-ai-with-vp-of-engineering-iccha-sethi/. Company size: ~700-900 (Vanta, Series C trust/GRC platform; ships the Vanta AI Agent in production). Pain points: ensuring AI feature/agent quality doesn't regress with code changes; needing a real-time signal on AI quality; augmentation-vs-automation reliability. Challenges: monitoring whether code changes impact AI performance at scale; grading LLM outputs against golden datasets curated by ex-auditors. Must-haves: LLM-as-judge evals integrated into CI/CD; real-time reliability signal for AI features. Nice-to-haves: broader observability across the full agent lifecycle. ICP confidence: High — VP/SVP Engineering, Series C ~700-900 employees, actively shipping AI agents in production, explicitly focused on reliability & evaluation. Prior: eng leadership at GitHub (Actions, Codespaces), InVision, Atlassian, Rackspace.
Idan Bassuk — Chief R&D and A.I. Officer, Aidoc
Signal: 4 (ICP speaking publicly about agents in production). Title verified live on LinkedIn people search 2026-09-02: "Chief R&D and A.I. Officer at Aidoc", Israel.
Source: https://langtalks.ai/conference — sat on the "Panel: Agents in Production" (May 2026, Tel Aviv).
Company size: 201-500 (Aidoc LinkedIn company page, read live 2026-09-02); Tracxn ~596 Jun 2026 and Revelio ~663 Mar 2026 — sources disagree but all sit inside the 50-2,000 band. Tel Aviv / New York. Clinical AI deployed at 130+ health systems.
Pain points: Chose to speak specifically on agents in production, not agents in general. Aidoc runs clinical AI where an unreliable or non-reproducible agent run is a patient-safety and regulatory event, not just a bad UX — so reliability, replay and auditability are hard requirements rather than nice-to-haves.
Challenges: determinism and reliability of agents inside regulated clinical workflows; scaling agent count without a proportional increase in ops and validation load; proving to hospital customers what each agent did and why.
Must-haves: per-run audit trail and replay; guardrails and failure containment; validation evidence for regulators.
Nice-to-haves: cost-per-run attribution by customer/tenant; SLOs per agent.
ICP confidence: Medium — Chief R&D/AI Officer (well above Director), 201-500/~596 employees (in band), explicit agents-in-production signal. Discount: Aidoc is later-stage (Series E) rather than Series A-C, and pain is inferred from panel topic plus regulatory context rather than a direct cost quote. Net-new name (Aidoc already in library).
Signal: 4 (ICP technical co-founder/CTO at AI-native company building & shipping agents). Source: https://www.prnewswire.com/news-releases/visitt-raises-22-million-series-b-funding-to-build-the-single-ai-interface-for-cre-property-operations-302669385.html ; https://visitt.io/blog/ai-agents-property-operations ; https://www.calcalistech.com/ctechnews/article/bychnr4uwg . Company size: Series B Israeli AI-native proptech; 150+ customers, 900%+ growth in managed sq ft in 2025 (headcount not published; Series B scale implies ~80-150). Stage: Series B ($22M led by Susquehanna Growth Equity, Jan 2026). Independent. Actively building/shipping agents: AI-native CRE property-operations platform "with AI built into the operational core from day one" — work-order intelligence, an AI Agent for certificate-of-insurance (COI) management in production, multilingual comms, and a stated roadmap of additional AI agents; June 2026 BGO Properties partnership rolling out across 46M sq ft / 300+ properties. Founders: CEO Itay Oren, CTO Idan Wender, CPO Jonathan Kroll. Pain points (INFERRED from positioning, NOT verbatim): reliably automating property-ops workflows with agents on top of messy legacy data; scaling from a first agent (COI) to a fleet of predictive agents; trust/accuracy of agent actions across many properties. Challenges: reliability and data-grounding of agents replacing legacy systems; scaling agent count. Must-haves: reliability, control, observability as agents multiply. Nice-to-haves: per-run cost visibility, predictive intelligence. ICP confidence: Medium (named technical co-founder & CTO, Series B AI-native, at least one agent in production + more in active development; exact headcount unconfirmed but Series B scale).
Signal: 3 (ICP engaging with competitor/adjacent tooling — named in a Confident AI customer case study; they evaluated and rejected LangSmith and MLflow). Found via web research — Claude-in-Chrome/LinkedIn NOT connected this run.
Source: https://www.confident-ai.com/case-study/finom (customer since January 2026)
Company size: 500–600 employees (stated in the case-study fact box; consistent with press describing Finom as serving 200,000+ SMBs). Series C — €115M led by AVP (AXA Venture Partners) with Headline Growth, General Catalyst, Northzone, Cogito Capital, June 2025; €300M+ total raised. Amsterdam; SME banking/financial platform.
LinkedIn URL: not verified
Pain points: "Without Confident AI, each active AI product needs dedicated engineering hours every week just for the prompt improvement cycle. With five to ten AI products inside Finom, that's over €250K in projected engineering costs." A 10-day iteration cycle dominated by queueing. No unified observability across multiple agents — "debugging meant digging through logs manually."
Challenges: "As agents become more complex, it's hard to understand what they do, how, and why." Agents execute real financial actions (issuing cards, setting limits, editing accounting records), so failures are financial harm not just bad UX. Sub-agent architecture with dedicated MCP servers per domain; multi-turn conversations that scripted tests could not capture. Colleague Ivo Dimitrov (Chief AI Officer) frames the goal as "making AI not just a copilot, but a decision-maker inside the process."
Must-haves: Evaluate the live end-to-end agent (tools, MCP servers, sub-agents) rather than isolated prompts; multi-turn simulation; unified tracing with latency and failure-pattern visibility; a UI that lets PMs/data scientists run evals without engineering tickets, plus an SDK that fits pytest/CI.
Nice-to-haves: Shared workspace across product and engineering; product-owned dataset management; CLI parity with the UI.
ICP confidence: High — owns the AI quality function at a 500–600 person Series C company running money-moving agents in production across 5–10 AI products, and one of very few named leaders publicly attaching a euro figure to the agent improvement loop.
Igor Ostrovsky — Co-Founder (technical), Augment Code
Signal: 4 (ICP technical co-founder building/shipping coding agents in production; found via 2026 web research on AI coding-agent companies — LinkedIn search blocked by login wall this run). Source: https://www.augmentcode.com/blog/augment-inc-raises-227-million ; https://www.crunchbase.com/person/igor-ostrovsky-4fdb ; https://www.linkedin.com/in/igoro/ . Company size: 176 employees (May 2026). Stage/funding: Series B ($227M+ raised; ~$1.4B valuation; Sutter Hill / Lightspeed / Evolution Equity). Actively building/shipping agents: Augment ships coding agents incl. cloud "Remote Agents" that run multiple dev tasks concurrently over large enterprise codebases. Background: ex-Chief Architect for FlashBlade at Pure Storage; ex-Microsoft software engineer. NOTE: co-founder Guy Gur-Ari is already in the brain (id ~160); Igor Ostrovsky was previously only mentioned in Guy's notes and is NOT yet a person record — this add is net-new. Pain points (role-inferred): running coding agents reliably over large enterprise codebases; concurrent cloud agents drive compute/cost scaling; context/retrieval management. Challenges: agent reliability across the SDLC; long-context over big repos; cost of concurrent agents. Must-haves: production reliability, robust context/retrieval, per-task cost visibility. Nice-to-haves: cost optimization across concurrent agents. ICP confidence: High (technical co-founder at 176-person Series B whose product is coding agents in production).
Ilan Chemla — Head of AI Innovation, Nimble (Nimble Way)
Signal: 4 (web-research proxy — LinkedIn/Chrome unavailable this run). Source: TheOrg org chart https://theorg.com/org/nimble-4 (leadership incl. Head of AI Innovation Ilan Chemla, VP R&D Alon Bar-Tzlil & Tsvika Naveh); company via https://techcrunch.com/2026/02/24/nimble-way-raises-47m-to-give-ai-agents-better-cleaner-data/ . Company size: ~136 employees; Series B. No confirmed LinkedIn URL captured (not fabricating).
Pain points (from company positioning): delivering AI/agent capability that turns messy real-time web data into reliable structured inputs; agent hallucination from bad data.
Challenges: productionizing AI features that agents depend on; accuracy/validation at scale.
Must-haves: reliable data + model outputs agents can act on.
Nice-to-haves: efficiency/cost of AI pipelines.
ICP confidence: High — "Head of AI" title (direct ICP) at a confirmed 50-2,000-emp agent-data company.
Inbal Budowski-Tal — VP of AI, Foundation AI
Signal: 4 (ICP AI leader at AI-native document-automation company; found via LinkedIn people search 'Director of AI ... agents production', page 2). Source: https://www.linkedin.com/in/inbal-budowski-tal/ ; headcount verified via PitchBook/Tracxn (https://pitchbook.com/profiles/company/234635-05). Company size: ~55-195 employees (PitchBook 195; other source 54) — in-range. Foundation AI (Irvine, CA; founded 2018) automates classification, profiling and routing of inbound documents/emails/attachments for legal and insurance — agentic document processing. Pain points (inferred): reliability of document-routing/classification agents in production; cost of LLM processing across high inbound-document volume; accuracy/SLA for legal/insurance customers. Challenges: keeping agent decisions accurate + auditable; per-document unit economics. Must-haves: production reliability, cost-per-run visibility, monitoring. Nice-to-haves: evals, cheaper-model routing for simple docs. ICP confidence: Medium (clear VP of AI target title; AI-native, agentic doc automation; in-range headcount).
Ioannis Antonoglou — Co-Founder & CTO, Reflection AI
Signal: 1/4 — ICP publicly writing/speaking about building agents at a qualifying company. Ioannis Antonoglou (co-creator of AlphaGo/AlphaZero/MuZero, ex-DeepMind Gemini post-training) is Co-Founder & CTO of Reflection AI, building the autonomous coding agent "Asimov." Public talks incl. "Building Superhuman Coding Agents."
Source: https://en.wikipedia.org/wiki/Reflection_AI ; https://www.crunchbase.com/person/ioannis-antonoglou
Company size: ~200 employees (confirmed 2026); Series B.
Pain points: long-horizon coding-agent reliability; cost of agentic tasks (industry cites ~$5–8 per agentic SWE task before retries/failures); whole-codebase context management / token waste; scaling agents.
Challenges: making autonomous coding agents reliable and economical at production scale.
Must-haves: reliability at long horizons, cost/token control, context efficiency.
Nice-to-haves: open-model flexibility.
ICP confidence: Medium-High — Co-Founder & CTO, ~200-employee Series B lab actively developing coding agents. (Note: some 2026 coverage describes product as not yet broadly shipped — "actively developing.")
Isaiah Granet — Co-founder & CEO (technical), Bland AI
Signal: 1 (surfaced via agent cost/reliability-in-production search; ICP building agents at scale). Source: https://fortune.com/2026/06/16/voice-ai-bland-50-million-after-being-rejected-by-180-investors/ , https://www.prnewswire.com/news-releases/bland-surpasses-100m-funding-with-new-series-c-... , https://www.crunchbase.com/organization/bland-ai . Company size: ~112-115 employees (Mar 2026), Series C ($50M led by Dell Technologies Capital, total $100M+). Handles 3.5M+ calls/week across healthcare, financial services and other regulated industries. Pain points: reliability of voice agents in complex, high-stakes conversations where mistakes carry real consequences; sub-second latency requirements; per-minute unit economics (pricing from ~$0.12/min) put direct pressure on LLM/inference cost per run. Challenges: scaling millions of concurrent agent calls reliably; controlling agent behavior in regulated/compliance settings; keeping cost-per-call viable at massive volume. Must-haves: reliability at scale, tight cost-per-run control, behavior control/guardrails for compliance. Nice-to-haves: granular per-call cost/observability, model routing. ICP confidence: High — technical CEO/co-founder of ~112-person Series C AI-native company running agents at very high production volume with acute cost + reliability pressure. (Note: profileUrl is company site.)
Isaiah Williams — Co-Founder & CTO, Casca (Cascading AI)
Signal: 4 (ICP CTO building/shipping lending agents; found via 2026 banking-agent sweep). Source: https://www.crunchbase.com/organization/cascading-ai , https://www.ycombinator.com/companies/casca . Company size: HEADCOUNT BORDERLINE — Tracxn ~53 (May 2026), PitchBook ~34, YC ~40 (confirm the 50-employee floor before outreach); SF; ~$44.6M raised (Series A); founded 2023 (YC S23). Casca = AI-native loan origination system used by FDIC-insured banks and fintechs (SBA/small-business lending), agents remove ~90% of manual origination effort. Co-founders met at Stanford; Williams did ML at EliseAI, Haffer core-banking at Avaloq. LinkedIn URL NOT confirmed — do not guess it. Pain points (INFERRED): lending agents in FDIC-regulated banks must be explainable/defensible to bank examiners and adverse-action reviewers; reliability of multi-step origination agents; cost control. Must-haves (INFERRED): per-action audit trail/determinism, regulatory defensibility, reliability. Nice-to-haves (INFERRED): per-run cost visibility. ICP confidence: MEDIUM (named technical co-founder/CTO and strong agent fit, but headcount sits right on the 50-employee floor with conflicting sources).
Signal: 4 (web-research pivot — Chrome/LinkedIn NOT connected this run). Source: https://techfundingnews.com/juicebox-lands-30m-to-revolutionise-hiring-with-ai-powered-recruitment-tool-peoplegpt/ ; https://www.unifygtm.com/insights-headcount/juicebox . Company size: ~110 employees (Unify, Jun 2026); Series B, $80M at $850M valuation (Mar 2026, DST Global/Sequoia/Coatue/YC). Actively shipping AI agents: PeopleGPT + 24/7 AI recruiting agents that source, evaluate and engage candidates across 800M+ profiles; 5,000+ customers, ARR tripled since Series A. Note: co-founder David Paffenholz (CEO) already in People Library; Ishan Gupta (CTO) is net-new. Pain points (inferred from product + CTO role): reliability of autonomous outreach/eval agents at scale; LLM cost per candidate search across 800M profiles; scaling from single search to always-on agent fleet. Challenges: keeping agent output accurate and cost-efficient as usage triples. Must-haves: cost-per-run visibility, reliability at scale. Nice-to-haves: agent observability. ICP confidence: High (Co-founder/CTO at a ~110-employee Series B company shipping recruiting agents).
Signal: 1 (ICP writing/speaking about agent reliability in production). Source: Qodo $70M Series B coverage (https://www.calcalistech.com/ctechnews/article/r1qdnboswx) + InfoQ "The Multi-Agent Approach: Building Reliable and Controllable Software Development Automation" (https://www.infoq.com/presentations/multi-agent-ai-architecture/). Company size: 115 (Israel/US/Europe), Series B, $120M raised. Pain points: AI-generated code causing production incidents (commissioned survey: 89% of eng orgs hit an AI-related production incident, 25% had an outage caused by AI code); trust/validation bottleneck for agent output. Challenges: making a multi-agent code-review swarm reliable and controllable at scale. Must-haves: agent reliability/verification and production-incident prevention. Nice-to-haves: cost control on agent runs. ICP confidence: High — 115-person Series B shipping a multi-agent system in production; CEO explicitly vocal on agent reliability.
Signal: 4 (ICP shipping/announcing multi-agent architecture; KubeCon EU Mar 2026). Source: https://cloudnativenow.com/kubecon-cloudnativecon-europe-2026/komodor-launches-extensible-multi-agent-architecture-for-ai-driven-site-reliability-engineering/. Company size: ~124 (Tracxn May 2026); company blog confirms 40 to 100+ hiring plan post-Series B. Stage: Series B ($42M Tiger Global; $67M total). Pain points (direct quote): "Most AI tools for operations focus on summarizing telemetry rather than resolving incidents, but complex outages require specialists from multiple domains working together to understand what's happening across the stack." Launch framing is explicitly that workflow agents dynamically invoke SME agents "retrieving precise context to avoid hallucinations and data overload" — context/token waste and reliability named as the design constraint. Challenges: orchestrating 50+ specialized agents plus customer-supplied agents via MCP/OpenAPI; Klaudia processes millions of Kubernetes events daily; sustaining claimed 95%+ root-cause accuracy across an open extensibility surface. Must-haves: per-agent context scoping to prevent data overload; multi-agent orchestration with detection/investigation/remediation staging; accuracy and hallucination controls; MCP extensibility. Nice-to-haves: per-agent and per-investigation cost attribution across a 50+ agent fleet; model portability (currently Bedrock/Claude). ICP confidence: High — textbook "hit a wall scaling from 1 to 5+ agents" except he is at 50+, and he names the context problem out loud.
Signal: 4 (technical co-founder of AI-native company shipping an agent platform). Source: https://en.wikipedia.org/wiki/Cohere ; https://betakit.com/cohere-pitches-security-and-productivity-with-general-release-of-north-enterprise-ai-platform/. Company size: ~350-500 (Cohere; North enterprise agent platform, sovereign AI focus). Pain points: reliable, secure, cost-efficient deployment of enterprise agents in production; on-prem/sovereign constraints. Challenges: production reliability and cost governance for customer-built agents; efficiency vs larger labs. Must-haves: reliability, security, cost efficiency/observability. Nice-to-haves: model efficiency gains. ICP confidence: Medium (technical co-founder at 50-2,000-emp AI-native company shipping agents; model-lab-adjacent so softer fit than a vertical agent deployer). LinkedIn URL not captured this run — do not fabricate.
Ivan Zilotti — Director, Engineering - Enterprise Conversational AI, SoundHound AI
Signal: 4 (senior technical leader at an agent-native company actively shipping AI agents). Source: https://www.linkedin.com/in/izilotti/ (found via LinkedIn people search "SoundHound AI director engineering agents"). Company size: ~750-950 employees (SoundHound AI, NASDAQ: SOUN; confirmed 954 as of Dec 2025). Company ships agents: YES — SoundHound named "Overall Agentic AI Company of the Year" 2026; ships Amelia agentic AI + Enterprise Conversational AI voice/chat agents in production. Pain points (inferred from role, not an observed individual quote): reliability of enterprise conversational agents in production, scaling voice/chat agents across many enterprise customers, LLM cost/latency per conversation. Challenges: keeping multi-tenant production agents reliable and cost-predictable. Must-haves: production agent reliability + cost visibility per run. Nice-to-haves: unified agent observability. ICP confidence: Medium (title + company confirmed; no individual post/comment observed — surfaced via company-scoped people search).
Iwona Bialynicka-Birula — Head of Applied Research, Cresta
Signal: 1 (ICP author writing publicly about building/shipping production-grade AI agents). Source: https://cresta.com/blog/building-and-deploying-production-grade-ai-agents-crestas-end-to-end-approach (Cresta Engineering blog, Nov 25 2025; she also authored a 3-part series on non-deterministic agent testing/evaluation). Company size: 500+ employees, Series D, AI-native contact-center agent platform (spun out of Stanford AI Lab 2017), ships autonomous AI Agents in production at Fortune 500 scale (100M+ conversations). Pain points: non-deterministic testing/evaluation of agents, catching agent failures before they hit production ("whack-a-mole in production"), aligning agent behavior with human preferences, agent drift after go-live, out-of-the-box evaluators being untrustworthy. Challenges: bridging the gap between a promising agent demo and production-grade reliability; multi-agent architecture and version-controlled agent config; regression/drift monitoring at scale. Must-haves: rigorous pre-production simulation + automated testing, per-version diffing/rollback, continuous production monitoring/telemetry for agents. Nice-to-haves: easier human-in-the-loop alignment workflows, unified human+AI observability. ICP confidence: High — unambiguously senior technical AI leader (Head of Applied Research) at a 50-2,000-emp AI-native company shipping agents in production; pain directly maps to Alpha's reliability/observability wedge. Note: found via web research + verification; Claude-in-Chrome/LinkedIn was not connected this run.
J.D. Martindale — Director, Machine Learning & AI, Cohere Health
Signal: 4 (ICP technical AI leader at an agent-shipping company; surfaced via LinkedIn people/company search this run, no specific post authored). Source: LinkedIn people search 'Cohere Health machine learning director' + profile in/jd-martindale. Company size: ~500-800 (Cohere Health, Series C, Boston; prior-authorization & clinical-intelligence AI/agents for health plans). Pain points (role/domain-inferred, not directly expressed this run): reliability of AI in regulated prior-auth decisioning; cost per authorization/run; scaling ML/agentic models across payers with governance/auditability. Challenges: reliable, auditable, cost-controlled AI at payer scale. Must-haves: reliability + per-run cost visibility under compliance. Nice-to-haves: eval/observability across models. ICP confidence: Medium (Director ML & AI at qualifying agent-native company; pain inferred this run).
Signal: 4 (ICP technical leader at a company shipping AI agents in production). Source: https://www.linkedin.com/in/jrjasperson/ ; https://theorg.com/org/filevine/org-chart/j-r-jasperson ; https://www.filevine.com/news/filevine-expands-to-the-czech-republic-hiring-dozens-of-engineers/ ; https://www.filevine.com/about/ ; https://tracxn.com/d/companies/filevine/. Company size: ~873-883 employees (Revelio May-2026 ~883; ZoomInfo Dec-2025 873). Legal practice-management SaaS; ships LOIS (Legal Operating Intelligence System) with AI document review, demand-letter generation, immigration application automation; acquired Pincites (AI contract redlining) Jan-2026. Pain points (inferred from public positioning, not verbatim): scaling AI-driven automation across a multi-product legal platform while standing up a new international R&D center - i.e. many teams shipping agent features against one cost/reliability envelope; integrating an acquired AI product (Pincites) into an existing agent stack. Challenges: agent reliability and accuracy on legal documents where errors carry malpractice risk; distributed engineering org (US + Czech Republic) building AI features in parallel. Must-haves: consistent reliability/guardrails across agents built by multiple teams; visibility into what each AI feature costs and does. Nice-to-haves: unified observability across acquired + native AI products. ICP confidence: High - CTO title, ~880 employees (in band), company is explicitly building its category around applied AI in production. Caveat: some sources still list co-founder Jim Blake as CTO; Jasperson is the current CTO per his LinkedIn, The Org and Filevine's own newsroom. NOTE: not yet contacted; research only.
JP Voltani (João Pedro Voltani) — Chief Technology Officer, TRACTIAN
Signal: 1 (ICP writing/quoted about agent cost + reliability at scale). Source: https://www.nvidia.com/en-us/case-studies/tractian/ (named quote), corroborated https://blogs.oracle.com/oracle-brasil/tractian-impulsiona-a-inovacao-ao-desenvolver-modelos-de-ia-ate-50-mais-rapidos-na-oracle-cloud, and LinkedIn profile read live 2026-08-31 at https://www.linkedin.com/in/jpvolt/. TITLE NOTE: theorg.com and older sources list him as VP of Engineering; his live LinkedIn headline on 2026-08-31 reads "CTO @ TRACTIAN | IoT, Software, AI/ML, Agents | Pragmatic, applied engineering" — treat CTO as current. Based in Atlanta, GA (TRACTIAN moved global HQ to Coda at Tech Square, 20k sqft office + 40k sqft AI Labs). Company size: ~400 employees (Dec-2024 Series C press release, "400 employees, with teams in the United States, Mexico and Brazil"). Stage: Series C, $120M led by Sapphire Ventures (Dec 2024). Pain points: NVIDIA case study states TRACTIAN runs 50 specialized AI agents in production serving 500M inference requests/day across 200,000+ machines on ~500 GPUs, and that the company needed faster training iterations and LOWER TOKEN COSTS for inference; Voltani is quoted by name saying inference costs dropped 15% while latency was cut 50%. Oracle Brazil post (Jan 2026) quotes him on a 40% cut in total infrastructure cost. Challenges: agent fleet is still expanding to more use cases, so cost and failure surface grow non-linearly; currently attacking cost via infrastructure/vendor deals (NVIDIA, OCI) rather than an agent operating layer; TRACTIAN's open AI Engineer req (LangChain/LangGraph) calls for "system reliability with fallback mechanisms for provider outages." Must-haves: per-agent-run cost attribution across a 50-agent fleet, provider failover, latency SLOs at 500M requests/day. Nice-to-haves: routing policies, token-waste detection, cost-per-successful-task metrics. ICP confidence: High (exact CTO title verified live on LinkedIn, Series C, ~400 employees, 50 agents in production and actively scaling, and a named public quote from him specifically about inference cost and latency). NOTE: net-new company for the People Library — no prior TRACTIAN record.
Signal: 1 — ICP publicly writing/speaking about building agents. Jacky Koh is Co-Founder & Co-CEO of Relevance AI ("the AI workforce" — platform to build/recruit teams of AI agents). Public talk: "Building the AI Workforce and Why Most AI Agents Are Just Workflows."
Source: https://relevanceai.com/blog/the-ai-workforce-revolution-24m-series-b-to-accelerate-our-mission ; https://dayone.fm/episode/jacky-koh-(relevance-ai)-on-building-the-ai-workforce-and-why-most-ai-agents-are-just-workflows/
Company size: ~80+ employees (grew from 19 in 2023; Sydney + SF), Series B ($24M/A$37M, May 2025).
Pain points: distinguishing true autonomous agents from brittle workflows; reliability/control of multi-step agents at scale; cost and visibility of running an "AI workforce" of many agents.
Challenges: helping teams scale from 1 agent to a reliable multi-agent workforce without losing control or blowing up cost.
Must-haves: control/reliability across many agents, cost visibility per agent/task.
Nice-to-haves: no-code agent building for ops teams.
ICP confidence: Medium — co-founder & co-CEO of ~80-person Series B agent-platform company. (Co-founder Daniel Vassilev already in library.)
Jacob Eckel — VP, Platform Division, Gong
Signal: 4 (ICP senior technical leader at a company shipping AI agents). Found via LinkedIn people search; Gong ships AI agents. Distinct from co-founder Eilon Reshef (already in brain).
Source: https://www.linkedin.com/search/results/people/?keywords=Gong%20VP%20Engineering%20AI%20agents
Company size: ~1,200-1,400 employees. Within ICP 50-2,000.
Pain points (inferred from role/company context, not a direct quote): owns the platform underpinning Gong's AI/agent features; infra cost and reliability of agents at scale; token/compute efficiency.
Challenges: platform-level cost control across agentic workloads; per-run cost/latency visibility.
Must-haves: cost/observability at the platform layer; reliability at scale.
Nice-to-haves: routing/caching, autoscaling cost optimization.
ICP confidence: High (VP owning the platform division at a mid-size company shipping AI agents).
Jacob Lauritzen — Chief Technology Officer, Legora
Signal: 1 (ICP CTO publicly speaking about agent token/cost efficiency at scale). Source: AI Engineer World's Fair 2026 (speaker/session data: https://www.ai.engineer/worldsfair/2026/speakers.json) — talk "How to Connect AI to Billions of Legal Documents" (connecting frontier LLMs to billions of legal documents to solve end-to-end legal workflows "without burning extra tokens"; retrieval architecture built with turbopuffer). Company size: ~517 employees (May 2026); note: recently Series D ($5.6B val) — beyond the A–C target band, which lowers confidence. Pain points: token/compute waste when running agents over huge document corpora; retrieval cost at scale. Challenges: efficient context/retrieval for long legal-workflow agents without token blowout. Must-haves: control over per-run token spend and context efficiency. Nice-to-haves: cost attribution per workflow. ICP confidence: Medium (ideal persona/size and explicit token-cost pain, but company is Series D, past the stated A–C stage band). LinkedIn not present in source dataset.
Jad Chamoun — CTO, Forethought
Signal: 4 (CTO at an in-band, AI-native company shipping a multi-agent platform; identified via Forethought LinkedIn company People page filtered to engineering). Source: https://www.linkedin.com/in/jad-chamoun . Company size: 51-200 employees (LinkedIn; ~98 per Revelio 2026); Forethought ships a multi-agent, omnichannel AI platform for customer experience (resolve/triage/assist agents) in production; Series C/D, $117M raised. Pain points (inferred from CTO role; no direct post/quote read this run): running multiple production agents means cost-per-resolution economics, token waste, and reliability are core to the P&L. Challenges: keeping multi-agent reliability high while cost-per-outcome stays viable at customer scale. Must-haves: per-run / per-agent cost visibility, guardrails, production observability. Nice-to-haves: automated model routing / prompt-cost optimization across the agent fleet. ICP confidence: High (CTO; in-band agent-native company; pains inferred not quoted).
Signal: Bucket 4 (ICP technical founder at an agent-active company shipping agents in production).
Source: https://www.law.com/legaltechnews/2026/02/05/ai-powered-legal-services-provider-lawhive-announces-60m-series-b-/ ; https://theorg.com/org/lawhive/org-chart/jaime-van-oers
Company size: ~200-450 staff (450 lawyers + eng post-acquisition), Series B $60M (Feb 2026, Mitch Rales/GV/Balderton), London/US. AI legal OS with 'Lawrence' AI paralegal (agentic: drafts, researches, manages cases, client intake, payments).
Pain points (INFERRED, not verbatim): running an agentic legal OS ('Lawrence') reliably across many live cases; cost per case / per agent run; accuracy in regulated legal work; scaling agents through US expansion.
Challenges: reliability & accuracy at case volume; per-run cost control; multi-step agent orchestration across intake→drafting→payments.
Must-haves: reliability, cost-per-case visibility, guardrails/auditability.
Nice-to-haves: replay/simulation, model routing.
ICP confidence: Medium-High — Co-Founder & CTO, Series B, right size, agent-active legal OS.
Jakob Freund — Co-founder & CEO (technical co-founder), Camunda
Signal: 1 (ICP writing about agent reliability/cost/control in production). Source: LinkedIn post surfaced via search — "At 500,000 conversations a month, even 99.9% reliability leaves 500 failures you can't trace," discussing monitoring and cost-tracking of AI spend on specific agent actions. Company size: ~567–694 employees (mid-market, Series B, ~$126M raised, Insight Partners/Highland Europe). Camunda positions itself as the enterprise platform for "agentic orchestration." Pain points: untraceable agent failures at scale, no per-action visibility into what agents cost/do, reliability at high conversation volume. Challenges: governing and controlling agent behavior across complex end-to-end processes; making agent runs auditable. Must-haves: per-action cost + behavior visibility, failure tracing, reliability at scale. Nice-to-haves: unified governance/control layer over agents, people, systems. ICP confidence: Medium — clear technical decision-maker at right-sized company shipping agent workflows; note Camunda's own "agentic orchestration" positioning partially overlaps the operating-layer category (adopter + adjacent).
Jakob Nederby Nielsen — Co-Founder & Chief Technology and Product Officer (CTPO), Dixa
Signal: 3 (ICP engaging with competitor content — Humanloop customer case study).
Source: https://humanloop.com/case-studies/dixa
Company size: ~158-161 employees (Crustdata / Tracxn, 2026). Series C, $105M raised. Copenhagen, Denmark.
Pain points (from the case study): the same cost-and-speed thesis as Alfarone — "Building AI products via traditional machine learning methods comes at an extremely high cost and slower development times" — plus company-level investment in compute-cost monitoring tied to pricing strategy.
Challenges: as CTPO he owns both the engineering cost line and the product pricing line, which is exactly where agent cost-per-run becomes a margin problem rather than an infra problem; GDPR/EU residency; keeping product/eng/ML aligned on prompt and model changes.
Must-haves: cost per AI interaction attributable to a customer/plan so pricing holds; model-swap regression safety; EU-compliant deployment.
Nice-to-haves: broader org-wide visibility so non-engineers can reason about AI unit economics.
ICP confidence: High — technical co-founder with the CTPO title, right headcount and stage, company demonstrably running agents in production and already measuring their compute cost.
PAIRED RECORD: Daniele Alfarone (Senior Director of Engineering, same company, ID above) is the practitioner entry point; Nielsen is the economic buyer. Approach Alfarone first.
Signal: 4 (ICP at a company shipping agents; company publicly writing about agent governance and auditability).
Source: name and title on GoodData's own management page https://www.gooddata.ai/company/ ; agent product evidence https://www.gooddata.ai/platform/ai-hub/ , https://www.gooddata.ai/platform/agentic-analytics/ , https://www.gooddata.ai/docs/cloud/experimental-features/mcp-server/ ; governance signal https://www.gooddata.ai/resources/gooddata-product-journey-mcp-and-governance-for-agentic-analytics/
Company size: ~280 (range 257–285 across trackers; 285 as of 31 May 2026, Revelio-class data; PitchBook bands 201–500 https://pitchbook.com/profiles/company/42226-12). NOTE: no first-party or filed figure found — aggregator-derived, treat as working estimate. Comfortably inside 50–2,000 under every source.
Pain points: No personal quote found from Franc. Company-level and directly on-message: GoodData positions the platform as letting teams "build, operationalize, embed, and monetize trusted analytics, workflows, and AI agents at scale — fully governed, auditable, and ready to scale." Their published product journey is explicitly about MCP plus governance for agentic analytics — i.e. they are publicly wrestling with how to govern and audit agents, which is the operating-layer problem.
Challenges: Agents are the product surface, not a side feature — Agent Builder, AI Assistant, AI Automation and an MCP Server all shipped. That means every customer deployment multiplies the number of agents they must keep governed, auditable and cost-predictable. Doing this at ~280 people across four continents.
Must-haves: Governance and auditability of agent actions at customer scale; a way to operationalize agents customers can trust; MCP-era tool-call visibility.
Nice-to-haves: Per-agent and per-customer cost attribution (not publicly stated); embedding/monetization economics for agents they resell.
ICP confidence: High — VP of Product Engineering is a clean Director+ match, headcount well inside band under every source, and agents are the core product with a public governance/auditability narrative. Downgraded slightly on the person: no LinkedIn URL found (GoodData's leadership page shows photos and titles only, no profile links) and no personal pain quote. Backup contact at the same company: Peter Fedoročko, "Field CTO" (same page) — ex-founder/CTO of Understand Labs, works with engineering and product on agentic analytics, but customer-facing rather than the platform owner.
Jakub Pavlik — Co-Founder & Head of Engineering (CTO-equivalent), Exaforce (exaforce.com)
Signal: 4 (ICP technical decision-maker building/shipping agents in production). Source: TechCrunch "Exaforce raises $125M Series B" (https://techcrunch.com/2026/05/12/exaforce-raises-125m-series-b...) and Exaforce/about + The Org (https://theorg.com/org/exaforce/org-chart/jakub-pavlik). LinkedIn URL not confirmed in this run — verifiable identifiers: GitHub github.com/pupapaik, X @JakubPav, based in Prague. Company size: 98 employees (confirmed, May 2026) — ideal ICP range. Exaforce is an agentic SOC platform; its multi-agent "Exabots" were already processing millions of security investigations for enterprise customers by Series B ($200M total raised, $725M valuation, founded 2023). Prior: co-founder/CTO tcp cloud (→Mirantis), Dir. Eng at Volterra (→F5). Pain points (inferred): running multi-agent investigation pipelines reliably at scale; cost of agents doing deep data analysis over millions of runs; observability/attribution of agent work. Challenges: reliability + accuracy of autonomous investigation; scaling agent count. Must-haves: reliable, cost-visible, controllable agents in production. ICP confidence: High (technical co-founder/eng head, AI-native agent co., 98 employees, agents in production).
James Allen — Director, Solutions Architecture EMEA, Writer
Signal: 4 (Director-level technical leader at ICP agent company; via LinkedIn People directory, Writer employees). Source: https://www.linkedin.com/in/james-m-allen. Company: Writer = enterprise agentic AI platform; 201-500 emp, Series C. Pain points (inferred): architecting/deploying enterprise agent solutions reliably; LLM cost at customer scale; production reliability. Challenges: scaling solution architecture; cost/reliability. Must-haves: cost observability + reliability guardrails. Nice-to-haves: model optimization. ICP confidence: Medium (Director-level & technical, customer-facing Solutions Architecture function). NOTE: pain points inferred from role/company, not verbatim quotes.
Signal: 4 (ICP senior technical leader at a company shipping AI agents). Found via LinkedIn people search targeting mid-size autonomous customer-service agent companies.
Source: https://www.linkedin.com/search/results/people/?keywords=DigitalGenius%20engineering%20AI%20agents
Company size: ~100-150 employees (ecommerce autonomous AI agents). Within ICP 50-2,000.
Pain points (inferred from role/company context, not a direct quote): running autonomous ecommerce support agents in production; scaling agent volume while controlling LLM spend; production reliability.
Challenges: unit economics of per-ticket agent runs; observability into cost/quality per deployment.
Must-haves: cost-per-run visibility; reliability/monitoring for agents in production.
Nice-to-haves: routing/caching optimization; anomaly alerts on runaway runs.
ICP confidence: High (VP Engineering at a mid-size autonomous-agent company).
James Fox — Co-Founder & CTO, Gamma
Signal: 4 — technical co-founder/CTO at an AI-native company scaling AI generation/agentic features to 70M users at ~$100M ARR.
Source: https://en.wikipedia.org/wiki/Gamma_(app) ; https://www.saastr.com/the-first-44-speakers-for-saastr-ai-annual-2026-the-founders-and-operators-actually-shipping-ai-at-scale/
Company size: ~360 employees (Series B)
Pain points: scaling AI features cost-effectively to consumer scale; reliability of generation at volume
Challenges: cost and reliability of AI at 70M-user scale
Must-haves: cost visibility and reliable generation
Nice-to-haves: model routing to cut spend
ICP confidence: Medium — technical C-suite, AI-native, right size/stage; caveat: primarily single-product AI generation, agent-fleet depth unclear; LinkedIn unverified/left blank
James Reggio — Chief Technology Officer, Brex
Signal: 1/4 (ICP CTO publicly describing a multi-agent fleet in production + the need for visibility/control). Source: https://venturebeat.com/orchestration/brex-bets-on-less-orchestration-as-it-builds-an-agent-mesh-for-autonomous ; https://www.latent.space/p/brexs-ai-hail-mary-with-cto-james ; https://www.startuphub.ai/ai-news/ai-video/2026/brexs-multi-agent-network-replaces-dashboards-with-executive-assistants/. Company size: ~1,100-1,900 employees (Brex; fintech). Fits 50-2,000. Actively shipping 5+ agents in production: "Agent Mesh" — network of narrow, role-specific agents (Brex Assistant orchestrator + Audit Agent, Procurement Agent, Reimbursement Agent, etc.) serving 40,000+ companies; customers hitting 99% automation. Reggio publicly argues traditional orchestration frameworks are becoming a constraint and emphasizes agents that "operate independently but with full visibility." Pain points: agents need FULL VISIBILITY while running autonomously; orchestration-framework limits; reliability of autonomous finance agents. Challenges: coordinating many role-specific agents reliably; visibility/control across the mesh; cost of running agents for 40k+ customers. Must-haves: agent observability/visibility across a fleet, behavior control, reliability in a regulated finance context. Nice-to-haves: per-agent cost attribution, model routing. ICP confidence: Medium (technical CTO/decision-maker running exactly the kind of multi-agent fleet Alpha targets; STRONG agent story). CAVEAT: Brex was acquired by Capital One (~$5.15B, April 2026) — no longer independent Series A-C; flag before outreach. LinkedIn URL not verified this run (left blank rather than fabricated).
Jamie Hall — Co-founder & CTO, Lorikeet
Signal: 4 (ICP technical co-founder building/shipping agents in production) + 1 (writes publicly on "AI humility" & agent reliability). Source: https://www.producttalk.org/building-lorikeet-how-ai-humility-and-a-dual-agent-architecture-are-redefining-customer-support/ ; https://www.lorikeetcx.ai/about ; https://tracxn.com/d/companies/lorikeet . Company size: 89 employees (Apr 2026); Series A, ~$49M raised over 3 rounds (latest Aug 2025). Sydney/US. AI-native customer support platform ("Universal Concierge") built on a dual-agent architecture; agents resolve complex, account-action support cases for financial-services/fintech customers. Hall is technical co-founder & CTO — ex-Google Brain research tech lead on factual grounding in LLMs (named author on LaMDA and Meena papers). Pain points: agents must know when NOT to act ("AI humility") — reliability/trust for high-stakes support actions; hallucination control; escalation logic. Challenges: making autonomous agents dependable and safe enough to take real account actions in production; scaling resolution rates (40-60% → 80%) without quality loss. Must-haves: reliability, control/guardrails, visibility into agent decisions. Nice-to-haves: per-run cost visibility, model routing. ICP confidence: High (technical co-founder/CTO at an 89-person AI-native agent company shipping in production; ex-Google Brain LLM researcher). Note: profileUrl is company LinkedIn; personal LinkedIn not verified this run.
Jamieson Fregeau — Co-Founder & President (technical co-founder), Quandri
Signal: 4 (technical co-founder shipping AI 'digital workers'/agents in production for insurance). Source: https://betakit.com/quandri-secures-16-5-million-cad-to-help-north-american-insurance-firms-automate-renewals-with-ai/ ; https://www.quandri.io/about/ . Company size: ~75 (Series A ~$16.5M CAD/$12M USD 2025; 100+ brokerages; 15x revenue growth; LinkedIn Top Startup Canada 2024). Pain points: scaling digital-worker agents reliably across 100+ brokerages; accurate automation of policy renewals/requoting; cost/efficiency at scale. Challenges: reliable, accurate automation of repetitive insurance workflows in production. Must-haves: reliability + accuracy at scale. Nice-to-haves: broader workflow coverage. ICP confidence: Medium (Series A AI-native shipping agents in production; computer-engineering technical co-founder; title is President).
Jan Hollez — Co-Founder & CTO, Deliverect
Signal: 4 (ICP CTO at a company shipping a named multi-agent fleet in production). Source: https://www.prnewswire.com/news-releases/deliverect-launches-ai-agents-and-smart-assistants-to-supercharge-restaurant-sales-protect-digital-revenue-and-maximise-efficiency-302737268.html ; https://restauranttechnologynews.com/2026/04/deliverect-launches-ai-agents-and-smart-assistants-to-boost-restaurant-sales-and-protect-digital-revenue/ ; headcount from Tracxn (458 employees as of 31 May 2026). LinkedIn: https://be.linkedin.com/in/janhollez. Company size: ~458 (Tracxn, 31 May 2026). Vertical: restaurant / food-delivery order-integration infrastructure (Ghent, Belgium; 11 offices, 5 continents; $230M+ raised, unicorn). Agent evidence: 'Deliverect AI' launched 9 Apr 2026 and explicitly described as 'a digital workforce of autonomous agents and smart assistants' that continuously rewrite digital menus, resolve technical issues and replace manual tasks; rolled out to selected launch markets first with AU/NZ/North America following 'in the coming weeks' — i.e. a STAGED multi-agent rollout, currently mid-rollout. CEO publicly predicting AI agents will transform the restaurant industry (Food On Demand, 29 Apr 2026) = executive-level commitment, budget exists. Pain points (INFERRED from public product/positioning, not verbatim quotes): agents run continuously against a very high-volume transactional order stream across 11 markets, so per-run inference cost is directly a COGS line, not an R&D line; 'resolve technical issues' agents are always-on exception handlers = unbounded run volume. Challenges (INFERRED): staged geographic rollout means the same agents must hold up on 3+ new market stacks in weeks; menu-rewriting agents touch live revenue, so a bad agent action is immediately monetary. Must-haves (INFERRED): per-agent and per-run cost attribution as the fleet goes from launch markets to global; deterministic replay / audit of what an agent changed on a live menu. Nice-to-haves (INFERRED): drift detection as menus and market data change; cross-market comparison of agent performance. ICP confidence: HIGH — technical co-founder, still CTO, company size squarely mid-range, dated 2026 multi-agent production launch, and he is an active public technical voice (ICT Personality of the Year; ran a Deliverect x AWS GenAI meetup).
Jan Schlie — VP of AI, Jimdo
Signal: 3 (ICP engaging with competitor content — speaking at LangChain's own conference about LangSmith) + 4 (ICP presenting on shipping agents).
Source: https://interrupt.langchain.com/london/agenda — Interrupt London, 13 Oct 2026. Talk: "From Trace to Outcome: Building an AI Companion for Small Businesses with LangSmith."
Company size: ~230–286 (Craft.co 286; Jimdo careers page states "230+ people"). Hamburg-based SMB website/business platform, shipping a customer-facing AI companion agent.
Pain points: Talk framing is explicitly about closing the gap between agent traces and business outcomes — i.e. having trace data but not being able to tie it to whether the agent actually helped the customer.
Challenges: Consumer/SMB-scale agent volume where per-interaction margin matters; proving agent ROI to the business, not just debugging it.
Must-haves: Linking trace-level agent telemetry to outcome metrics; per-run cost visibility given thin SMB unit economics.
Nice-to-haves: Outcome-weighted quality scoring rather than raw trace inspection.
ICP confidence: Medium-High — exact title (VP of AI), company clean in the 50–2,000 band, actively shipping a production agent, and already invested in the observability category. Geography: Germany (ICP is worldwide, so in scope). Pain is inferred from the talk abstract rather than a verbatim quote — confirm before outreach.
Janak Ramachandran — VP, Head of AI, Innovaccer
Signal: 4 (ICP senior technical leader at a 50-2,000-employee company actively shipping AI agents; surfaced via LinkedIn People search after Signal 1-3 content searches returned mostly non-ICP consultants/vendors). Source: https://www.linkedin.com/in/19051916/. Company size: ~1,780-1,800 (Innovaccer; healthcare data + AI, ships Sara AI and healthcare AI agents). Pain points (inferred, NOT verbatim): reliability/safety/accuracy of healthcare agents (high-stakes, regulated); cost of running documentation/insight agents at scale; auditability. Challenges: compliance + reliability in healthcare; scaling agent workloads across providers. Must-haves: reliability, audit trail, cost governance. Nice-to-haves: per-run cost attribution by workflow. ICP confidence: High (VP/Head of AI; ex-Netflix/Apple; PhD).
Signal: 4 (ICP speaking publicly about building and evaluating agents in production; spoke at LangChain Interrupt 2026 alongside other ICP-matching agent leaders). NOTE: LinkedIn/Claude-in-Chrome NOT connected this run — sourced via web research + primary-source verification.
Source: LangChain Interrupt 2026 Day 2 session "Generative AI for Clinical Conversations" (https://interrupt.langchain.com/event-agenda) and "Building Clinical AI Agents with LangGraph: Abridge's Eval Stack for High-Stakes Healthcare" (https://www.youtube.com/watch?v=mxweSHetuN8); Latent Space episode "AI-Native Healthcare: 100M Doctor Visits..." (https://www.latent.space/p/abridge); appointment announcement https://www.abridge.com/blog/janie-lee-vice-president-of-product.
Company size: Abridge ~539-635 employees (Revelio Labs Dec 2025: 539; Crustdata Jun 2026: 635, +65.8% YoY). Ships real-time clinical AI agents across ~250 health systems; 100M+ doctor-patient conversations processed.
Pain points: shipping agent changes fast without compromising patient safety in a regulated, high-stakes domain; real-time agents where failure is not recoverable; eval coverage for clinical decision support and prior-auth agents.
Challenges: an eval stack that has to hold up under regulatory and clinical scrutiny; scaling real-time agent workloads across 250 health systems; expanding from documentation agents into clinical decision support and prior auth without quality regressions.
Must-haves: trustworthy evals and production traceability for every agent decision; quality guardrails that let the team keep shipping at pace.
Nice-to-haves: cost/efficiency visibility per agent run as real-time agent volume grows.
ICP confidence: High — VP of Product on an explicitly AI/agent product line (ICP allows AI-focused VP of Product), company in the 50-2,000 band, multiple agents in production at scale.
Signal: 4 (ICP describing a live multi-model agent pipeline in production). Source: https://www.technologyreview.com/2025/09/22/1123873/medical-diagnosis-llm/ (MIT Technology Review) and ViVE 2026 talk https://www.youtube.com/watch?v=x7znEoDlujg. Company size: ~108-172 (LeadIQ Jul 2026 band 51-200). Stage: Series B — $60M led by Oak HC/FT, May 2025. Product: ScopeAI, AI-enabled care delivery (Los Angeles). Pain points (read): Goodner describes ScopeAI as "a set of large language models, each of which can perform a specific step in the visit" — a live multi-model/multi-agent pipeline (fine-tuned Llama + Anthropic Claude) in clinical production. Reliability gating is thin and manual: Akido "evaluates ScopeAI's performance by testing it on historical data and monitoring how often doctors correct its recommendations," requiring the correct diagnosis in the top three at least 92% of the time before deploying to a specialty; the article notes Akido "hasn't undertaken more rigorous testing." He says it "would be technically simpler to let patients converse with an AI agent on their own" but they keep a human in the loop as a reliability hedge. Challenges: proving safety of a chained multi-LLM system in a regulated clinical setting with only offline historical-data evals and human-correction rates as signal; scaling specialty by specialty. Must-haves: production trace-level observability per agent step; regression detection when a step in the chain changes; correction/override rate as a live monitored metric. Nice-to-haves: automated eval harness replacing manual historical-data testing; cost visibility across the fine-tuned-Llama vs frontier-model split. ICP confidence: High — exact-fit title, verified headcount/stage, documented multi-agent system in live production with a self-admitted immature eval regime. NOTE: cost/token pain inferred, not stated.
Jared Palmer — VP of Engineering, Cognition (Devin)
Signal: 4 (senior technical AI leader joining a company actively shipping agents — announced as new VP of Engineering at Cognition, makers of Devin, the autonomous AI software engineer; prev VP of AI at Vercel, VP Eng at Xbox).
Source: https://a16zbuild.substack.com/p/jacob-tsimerman-david-velez-and-35 (a16z Build talent-moves, Aug 4 2026)
Company size: est. ~200–500 employees (AI-native, Series C, well-funded; acquired Windsurf). Within 50–2,000.
Pain points: scaling autonomous coding agents in production; agent reliability at scale; cost per agent task (coding agents make 3–10x more LLM calls than chat, $5–8+/task in API fees).
Challenges: shipping reliable autonomous agents; controlling inference/token cost per run as usage scales.
Must-haves: production reliability + cost control/visibility for agents at scale.
Nice-to-haves: per-run cost observability and control tooling.
ICP confidence: High — VP of Engineering at a leading AI-native agent company that ships agents in production; role and company size squarely in ICP.
NOTE: Discovered via web/news fallback — Claude-in-Chrome/LinkedIn browsing was not connected during this scheduled run.
Signal: 4 (company-scoped LinkedIn people search of agent-native company Dialpad). Source: https://www.linkedin.com/search/results/people/?keywords=Dialpad%20AI%20director%20engineering | Company size: ~1,200 (est; <2,000). Pain points (INFERRED from role+company): production reliability + cost control for voice/CX AI agents; per-run cost visibility across agents. Challenges: reliable scaling of agent workloads. Must-haves: production reliability + governance; cost-per-run visibility. Nice-to-haves: token/context-waste reduction. ICP confidence: High (Director of Engineering at agent-native 50-2,000-emp company; confirmed via explicit '@ Dialpad' headline; net-new).
Jason Fang — Director of Engineering, AI Platforms, Aisera
Signal: 4 (company-scoped LinkedIn people search of agent-native company Aisera). Source: https://www.linkedin.com/search/results/people/?keywords=Aisera%20director%20engineering | Company size: ~300-500 (est; Series D). Pain points (INFERRED from role+company, not observed posting): cost + reliability of enterprise agentic-AI service agents (IT/HR/customer service) in production; per-run cost visibility across a multi-agent platform. Challenges: keeping large fleets of enterprise agents reliable and cost-controlled at scale. Must-haves: production reliability + governance; cost-per-run visibility. Nice-to-haves: token/context-waste reduction. ICP confidence: High (Director of Engineering, AI Platforms at agent-native 50-2,000-emp company; company confirmed via explicit '@ Aisera' headline; net-new vs 2 existing Aisera contacts in brain).
Jason Luce — CTO, Paperless Parts
Signal: 4 (ICP at a company shipping AI workflows in production in a non-obvious vertical). Source: https://www.paperlessparts.com/press/paperless-parts-appoints-jason-luce-as-chief-technology-officer-and-scott-sawyer-as-chief-scientist/ and https://www.paperlessparts.com/press/paperless-parts-cuts-quote-setup-time-by-90-with-new-ai-supported-workflow/ — title verified live on LinkedIn 2026-08-30 ("CTO at Paperless Parts", Boston). Company size: 139 employees (PitchBook ~May 2026; Tracxn 139, RocketReach 142); Boston MA; quoting/estimating platform for job shops and contract manufacturers; Series B $30M led by OpenView. Pain points: shipping AI in production — Wingman AI, an AI-supported quoting workflow that cut quote setup time ~90%, and features that surface critical requirements from CAD and RFQ documents. He is quoted describing AI as "an incredibly powerful unlock for manufacturing." Challenges: they operate on a CMMC-compliant secure foundation for defense manufacturing, which makes auditability and per-run traceability of every model call a hard compliance requirement rather than a nice-to-have; CAD/document parsing is token-expensive per quote. Must-haves: auditable per-run model call logs compatible with CMMC; predictable cost per quote. Nice-to-haves: routing cheaper models for the easy quote paths. ICP confidence: Medium-High — CTO, US, right size and stage, non-obvious vertical (discrete manufacturing) so likely un-worked by competitors; caveat is that their public framing is "AI workflow" rather than an explicit agent fleet, so agent count needs qualifying on the call.
Jason MacDonald — Senior Director, Engineering, Commure
Signal: Signal 4 (LinkedIn people search at named agent company; Senior Director Engineering. Sourced by role/company match)
Source: https://www.linkedin.com/search/results/people/?keywords=%22Commure%22%20director%20engineering%20AI
Profile: https://www.linkedin.com/in/macdonald-jason/
Company size: ~500-1,000 (Commure, healthcare AI platform incl. Athelas; ambient documentation & revenue-cycle AI agents)
Pain points: (inferred, not directly observed) Shipping healthcare AI agents reliably in a compliance-heavy setting; cost per encounter/run
Challenges: (inferred) Reliability + cost visibility across clinical/RCM agent workloads
Must-haves: (inferred) Per-run cost + reliability observability; guardrails
Nice-to-haves: (inferred) Token/context-waste reduction
ICP confidence: High (Senior Director Engineering at an in-band agent-shipping healthcare company)
Jason Poole — Director of Engineering, Zapier
Signal: 4 (ICP engineering leader at agent-shipping company; found via LinkedIn currentCompany people-search of Zapier engineering). Source: linkedin.com/in/jasonapoole (Zapier company id 2418251 people search, past-month). Company size: ~1,000-1,800 (Zapier; shipping AI agents via Zapier Agents/Central across millions of workflows). Pain points (INFERRED from role+company stage, not an individual quoted post): agent cost blowout at scale; lack of per-run/per-agent cost visibility; reliability of non-deterministic agent steps in production automations. Challenges: scaling from a few agents to many while keeping cost & reliability under control; token waste from retries/loops. Must-haves: per-run cost attribution + ceilings; production reliability guardrails/observability. Nice-to-haves: model routing; per-agent cost dashboards. ICP confidence: High (Director of Engineering, technical, in-range agent-shipping company).
Jason Rosenfeld — Head of Engineering, Candid Health
Signal: 4 (ICP leader at a company actively shipping AI agents). Found via LinkedIn company People-tab sweep of Candid Health.
Source: https://www.linkedin.com/company/candid-health/people/?keywords=engineering
Company size: 51-200 employees (per Candid Health LinkedIn company page, SF/NY). Candid Health = AI-native revenue-cycle / medical-billing automation; agents handle claim scrubbing, denials and payer follow-up. Series B/C.
Pain points: NONE DIRECTLY OBSERVED THIS RUN — no post or comment found. Inferred only: high-volume, high-stakes agent runs where a wrong output is a denied claim, so both cost-per-claim and correctness matter simultaneously.
Challenges (inferred): unit economics — agent cost per claim has to stay well under the fee they charge; correctness/audit trail for payer disputes.
Must-haves (inferred): cost per agent run attributable to a claim; reliability/eval harness with replayable traces.
Nice-to-haves (inferred): shadow testing new prompts/models against historical claims before rollout.
ICP confidence: Medium-High. Clean role fit (Head of Engineering) and a genuinely agent-native company at the right size. Pain is inferred, not expressed.
Note: engineering org here is notably FDE-heavy — see also Steve Yazicioglu, Head of Forward Deployed Engineering, added this same run.
Jason St. Pierre — Co-founder & Chief Product Officer, Liberate
Signal: 4 (VP/Head of Product, AI-focused, at an agent-native company; found via press primary sources, not LinkedIn — Chrome/LinkedIn unavailable this run). Source: https://techcrunch.com/2025/10/15/liberate-bags-50m-at-300m-valuation-to-bring-ai-deeper-into-insurance-back-offices/ (names him as Liberate's CPO; previously product roles at Twitter, Google, and Verily/Alphabet). Company size: ESTIMATED ~60-150, NOT independently confirmed — inferred from Series B ($50M Oct 2025, $300M valuation) and 60+ enterprise customers with voice agents in production. DEDUPE NOTE: distinct from "Ryan St Pierre" already in the People Library (different first name/person). Pain points: productizing reliable voice AI agents for regulated insurance; compliance/audit and human-in-the-loop; workflow automation across sales, servicing, and claims. Challenges: shipping compliant, reliable voice agents that meet carrier requirements. Must-haves: reliability, auditability, compliance. Nice-to-haves: cost efficiency per interaction. ICP confidence: Medium (CPO = VP of Product AI-focused; company size estimated, not confirmed).
Jason Turpin — VP of Product, Observe.AI, Observe.AI
Signal: 4 (ICP senior product leader, AI-focused, at a company shipping AI agents). Source: https://www.linkedin.com/in/jason-turpin-22672b1 (found via Observe.AI People directory, filter 'VP'). Company size: 201-500 employees (LinkedIn company page verified); Observe.AI ships AI agents for contact centers. Pain points (inferred from VP of Product remit at an agent company): production agent reliability/quality, cost per interaction, expanding agent coverage across use cases. Challenges: shipping reliable agents customers trust while managing LLM cost. Must-haves: visibility into agent cost + reliability per run; eval/QA on agent output. Nice-to-haves: cost optimization and routing. ICP confidence: Medium-High (VP of Product at a verified 201-500 agent company; ICP explicitly includes AI-focused VP of Product).
Jason Warner — Co-founder & Co-CEO (ex-GitHub CTO), Poolside
Signal: 4 (technical founder building/shipping coding agents; talks about background agents in 2026). Source: https://en.wikipedia.org/wiki/Poolside_AI and https://www.focal.vc/transcripts/5yf-episode-20-poolside-ceo-jason-warner. Company size: 150 employees (Dec 2025); Series B ($500M, Bain/DST/eBay), $3B→~$12B valuation after Nvidia investment Oct 2025. Building foundation models + background coding agents ("everyone is talking about skills and prompts and background agents"). Pain points: reliability and control of background/autonomous coding agents, compute/cost of running agents at scale. Challenges: making coding agents dependable enough for real production software work. Must-haves: agent observability + cost control across many concurrent agents. Nice-to-haves: skills/prompt governance. ICP confidence: High (technical decision-maker, AI-native company shipping agents, 150 employees). Note: co-founder Eiso Kant (CTO) already in brain; Jason is a distinct new contact.
Signal: 4 (co-founder at a company shipping voice AI agents in production). NOTE ON METHOD: LinkedIn browsing unavailable this run (Chrome extension not connected); found/verified via web research (YC + Crunchbase). Pain points INFERRED from HappyRobot's domain, not verbatim quotes.
Source: https://www.ycombinator.com/companies/happyrobot ; https://www.crunchbase.com/organization/happyrobot ; https://startupintros.com/orgs/happyrobot
Company size: ~80 (Series B, ~$44M raised; voice AI agents automating logistics/supply-chain operations in production). Co-founders: Luis Paarup (CEO), Pablo Palafox, Javier Palafox.
CAVEAT: Confirmed Javier is a co-founder; his specific function (technical vs. commercial) is NOT confirmed from public sources — verify he is a technical/engineering leader before treating as a full ICP match (co-founder brother Pablo Palafox is the ML/technical co-founder).
Pain points (inferred): reliability of voice agents handling real logistics transactions; cost per call across high volume; scaling agent coverage across carriers/brokers.
Challenges (inferred): keeping automated voice agents reliable on messy real-world calls; unit economics of agent runs.
Must-haves (inferred): production reliability + cost visibility for voice agents.
Nice-to-haves (inferred): per-run cost attribution.
ICP confidence: Medium — right company profile (80 emp, Series B, agents in production) and a co-founder, but individual's technical scope unconfirmed; confirm role before prioritizing.
Jay Hack — Head of AI, ClickUp
Signal: 4 (ICP building/shipping agents at scale; found via 2026 coverage of ClickUp's agent deployment). Source: https://fortune.com/2026/05/18/ai-agent-to-human-ratio-clickup/ ; https://www.crunchbase.com/person/jay-hack ; https://clickup.com/brain/agents. Company size: ~1,010 employees (1,300 base minus 22% May-2026 layoff); Series C ($400M, ~$4B val). Actively shipping agents: ClickUp runs ~3,000 internal AI agents across eng/marketing/support and ships 'Super Agents' AI teammates; Jay Hack (ex-founder/CEO of coding-agent startup Codegen, acquired by ClickUp late 2025) now leads AI. Pain points (inferred from public statements/product): running thousands of agents reliably in production; per-run cost and observability as agent-to-human ratio hits 3:1; controlling agent behavior at scale. Challenges: reliability + cost control across a very large internal+product agent fleet. Must-haves: reliability, per-run cost visibility, observability/control at scale. Nice-to-haves: compounding intelligence across agent runs. ICP confidence: High (Head of AI, ~1,010 emp Series C, thousands of agents in production; exact-fit decision-maker and role).
Jay Tomasello — Chief Technology Officer, Transflo
Signal: 4 (ICP CTO at a company shipping specialised exception-handling agents in production). Source: https://www.businesswire.com/news/home/20260122099350/en/Transflo-Unveils-Workflow-AI-for-LTL-The-Freight-Audit-and-Invoice-Resolution-Engine-for-Brokers-and-Carriers ; https://www.transflo.com/news/transflo-unveils-workflow-ai-for-ltl/ ; https://www.transflo.com/news/transflo-appoints-jay-tomasello-as-cto/ ; https://www.freightwaves.com/news/transflo-names-jay-tomasello-as-cto-to-advance-ai-digital-strategy ; headcount ~314 (LeadIQ, Feb 2026), 321 (PitchBook), ~306 (third source). LinkedIn: https://www.linkedin.com/in/jaytomasello/ (profile still lists Forward Air — appointed at Transflo Oct 2025; verify before outreach). Company size: ~306-321. Vertical: freight/transportation back-office automation (Tampa FL; carriers, brokers, 3PLs, freight factors, shippers; digitises 800M+ documents/year). Agent evidence: Transflo Workflow AI for LTL launched 22 Jan 2026, described as deploying 'specialized AI agents, each one purpose-built to handle a specific, high-impact exception type with expert-level reasoning in real time' — that is an explicit per-exception-type AGENT FLEET, not one assistant. A second product, Workflow AI for Factors, launched subsequently. Appointed CTO 29 Oct 2025 to lead 'AI-driven freight optimization and workflow automation'; prior CIO at Forward Air and CIO/VP IT at FedEx Supply Chain. Pain points (INFERRED from public product/positioning, not verbatim quotes): agents run against a document stream of 800M+ documents/year, so per-document inference cost is a direct and very large COGS line; freight audit is a cents-per-invoice business, so an agent run that costs more than the invoice it resolves destroys the product's economics. Challenges (INFERRED): the fleet grows by one agent per new exception type, so the 1->N agent scaling wall is built into the roadmap; a large-enterprise-IT CTO now owning probabilistic systems where he previously owned deterministic ones. Must-haves (INFERRED): cost per agent run benchmarked against the value of the transaction it is resolving; reliability guarantees on money-touching invoice decisions. Nice-to-haves (INFERRED): a way to decide which exception types justify a new agent at all. ICP confidence: MEDIUM-HIGH — role, company size and agent evidence are all confirmed and dated; downgraded from HIGH only because no verbatim pain statement from him was found and his background is enterprise IT leadership rather than a public technical voice, so there is no warm content surface. PE/strategic-owned rather than Series A-C, which is a stage deviation from the stated ICP but well inside the headcount band.
Signal: 4 (ICP building/shipping agents in production — found via 2026 research on voice/chat AI-agent companies). Source: https://yellow.ai/author/kishore/ , Crunchbase, Wikipedia. Company size: ~875 employees (Feb 2026, after ~30% workforce cuts across two 2025 layoff rounds); Series C ($78M round, ~$102M total). Actively building AI agents: conversational + Voice AI platform (54%+ of customer base on Voice Stack) explicitly pivoting to autonomous AI agents for 2026. Pain points (inferred, not verbatim): cost/margin pressure (evidenced by late-2025 layoffs) while pivoting to autonomous agents; scaling voice agents reliably at enterprise volume. Challenges: doing more with a leaner org; reliability + unit economics of autonomous agents. Must-haves: cost efficiency, reliable autonomous agents at scale. Nice-to-haves: per-run / per-conversation cost visibility. ICP confidence: High (CTO & co-founder, ~875-person Series C shipping voice/chat agents in production; visible cost-pressure signal).
Jayanth Madheswaran — Founder & CEO (technical; ex-Rubrik founding engineer, ex-Facebook), Eve (eve.legal)
Signal: Targeted ICP people-search — technical founder at an agent-native legal-AI company not yet represented by this individual in the brain (brain already has David Zeng at Eve; Jayanth is a distinct, more senior contact). Source: https://www.linkedin.com/in/jayanth1/ (LinkedIn people search 'Jay Madheswaran Eve'). Company size: ~80-150 (Series B legal-AI, 2025). Pain points (ICP-inferred; no direct post captured this run): accuracy & reliability of legal AI agents on high-stakes litigation work; hallucination/verifiability risk; cost per matter as agents run long document + multi-step reasoning chains. Challenges: trust/auditability of agent outputs for law firms; latency vs thoroughness tradeoff; scaling agentic workflows across many matters. Must-haves: agent reliability, auditability, predictable cost. Nice-to-haves: per-run token/cost visibility, model routing. ICP confidence: Medium-High (technical Founder/CEO, agent-native, 50-2000 emp).
Jean-Olivier Racine — Chief Technology Officer, Rad AI
Signal: 4 (ICP CTO at a company shipping AI agents in production). Source: https://www.radai.com/about ; https://www.radai.com/news/rad-ai-expands-executive-team ; Contrary Research company report ; PitchBook/LeadIQ headcount. Company size: ~189-200 (PitchBook 189; LeadIQ ~195; Series C, $60M raised, >$140M total). Actively shipping agents: Rad AI runs generative-AI radiology-reporting agents (RadMate / continuous reporting + impressions) used daily by thousands of radiologists; works with 40%+ of US health systems and 9 of the 10 largest US radiology practices (~50% of US medical imaging). Racine joined as CTO Aug 2024 (ex-CTO Outset Medical; ex-AWS Health AI/Alexa/Fire TV). Pain points (inferred from product + regulated healthcare domain, NOT verbatim quotes): accuracy/reliability of clinical agents where errors carry patient-safety and liability risk; per-report/per-study inference cost at ~50%-of-US-imaging scale; auditability of agent outputs for clinical governance. Challenges: scaling many report-generation agents across health systems while holding reliability high. Must-haves: reliability/accuracy guardrails, audit trails, cost-per-run visibility at high volume. Nice-to-haves: agents that compound/improve from radiologist feedback. ICP confidence: High (clean CTO title; in-range headcount; Series C; production healthcare agents).
Jean-Philippe Joyal — Senior Director of Product Management, Ada
Signal: 4 (ICP senior product leader at agent-shipping company). Source: https://www.linkedin.com/company/ada-cx/people/?keywords=director. Company size: 201–500 employees (per Ada/ada CX LinkedIn). Company context: Ada ships customer-service AI agents in production for enterprises. Pain points (inferred from role/ICP): owning product for AI agents — accountable for agent resolution quality, reliability in production, and the unit economics (cost per resolution/run) as customers scale agent usage. Challenges: proving ROI and controlling cost of agents while scaling automation. Must-haves: visibility into agent cost-per-run and reliability metrics customers can trust. Nice-to-haves: levers to tune cost vs. quality per deployment. ICP confidence: Medium-High — Senior Director, product for AI agents at an ICP agent company (product/AI-focused leadership at Director+ level).
Signal: 4 (ICP senior AI/platform leader at validated target company Dataiku; sourced via LinkedIn People search, run 2026-08-01). Source: https://www.linkedin.com/in/jediv/ . Company size: ~1,000-1,200 (Dataiku — enterprise AI/ML platform now shipping agentic AI: LLM Mesh, Dataiku Answers, and AI Agents; AI-native, in-range on headcount). Pain points (role/company-contextual, not from an individual post this run): governing and controlling cost of many LLM/agent workloads across a multi-model enterprise platform; giving customers visibility into per-agent/per-model spend; reliability and guardrails for agents running in production at enterprise scale. Challenges: standardizing observability and cost attribution across heterogeneous models/agents; balancing capability vs. spend for enterprise buyers. Must-haves: per-agent cost & token visibility; production reliability/guardrails. Nice-to-haves: automated model routing/eval, regression testing. ICP confidence: High (SVP-level AI & Platform leader at 50-2,000-emp AI-native company actively shipping agentic products).
Jeegar Shah — Head of Applied AI & Platform Engineering, Atomicwork
Signal: 4 (senior technical AI leader at AI-native agent company; via people search 'Head of AI agents production platform').
Source: https://www.linkedin.com/in/jeegar-tilak-shah/ ; Atomicwork $25M Series A (Khosla Ventures, Z47), $40.3M total.
Company size: 50+ employees (US/Singapore/India); Series A.
Pain points: building and running a "crew" of built-in AI agents that reason, plan and execute enterprise service-management workflows autonomously; platform reliability at scale.
Challenges: agent reliability in autonomous enterprise workflows; grounding on an enterprise knowledge graph.
Must-haves: reliable agent orchestration/platform, observability.
Nice-to-haves: cost control per run.
ICP confidence: Medium-High — Head of Applied AI & Platform Eng at ~50-100-emp Series A AI-native agent company; head-level technical leader, right size/stage. Minor caveat: company sits near the 50-employee floor.
Jeff Barg — Head of AI, Clay
Signal: 3 (ICP engaging with competitor/ecosystem content — spoke at LangChain's Interrupt conference; team uses LangChain/LangSmith ecosystem tooling). Source: https://www.youtube.com/watch?v=LmQtSORYPfw ("How Clay runs 350 million GTM agents a month | Interrupt 26") ; https://www.linkedin.com/posts/langchain_clays-claygent-runs-350-million-gtm-agents-activity-7475536976725630976-MB7F . Company size: ~1,000–1,167 employees (Tracxn Feb 2026), $3.1B Series C — in range. Clay/Claygent runs GTM research agents. Pain points: runs 350M GTM agents/month; treats "infrastructure, throughput, cost, and quality as four discrete engineering disciplines"; built TCP-congestion-style back-pressure (4–10x throughput), caps retries because uncapped agents spin, used Anthropic prompt caching to cut costs "up to 70%." Challenges: controlling cost and reliability at extreme agent volume. Must-haves: token/cost control, retry/backpressure control, throughput at scale. Nice-to-haves: per-run cost visibility and quality eval. ICP confidence: High (public and specific on agent cost control, reliability, and scaling at production volume).
Signal: 4 (Director-level eng leader at agent-building company; same Traba search). Source: https://www.linkedin.com/search/results/people/?keywords=Traba%20co-founder%20CTO (profile https://www.linkedin.com/in/jeff-zhifan-chen/). Company size: ~173 (confirmed) — Traba, Series A ~$49M; building agentic platform for industrial workforce ops (see CTO Akshay Buddiga record — multi-threaded account). Pain points [INFERRED from verified role+company, NOT verbatim]: shipping reliable agent systems in production; agent reliability/eval; cost + scaling 1->many. Challenges: production reliability + observability for a 0->1 agents team. Must-haves: reliability + observability. Nice-to-haves: cost-per-run visibility. ICP confidence: Medium-High (Director of Engineering, agent-native ~173-emp Series A).
Signal: Bucket 4 (ICP founder at a company shipping agents — LinkedIn headline "AI Agents for the Patient Journey"). Source: https://www.linkedin.com/in/jefferyliu300/ ; company confirmation https://www.fiercehealthcare.com/ai-and-machine-learning/assort-health-scores-120m-series-c-scale-voice-ai-agent-platform-healthcare . Company size: ~100–150 and growing (15 in 2025 → ~250 by end 2026); Series C $120M. Co-founder & co-CEO of Assort Health, which ships specialty-specific voice AI agents in production for healthcare (patient calls, scheduling, insurance verification). NOTE: distinct from existing library contact Jonathan (Jon) Wang (co-founder) — net-new person, same company. Pain points: reliability of patient-facing agents in a regulated setting; cost per call at scale; visibility into agent cost/quality across specialties. Challenges: scaling reliable agents across many provider groups while managing cost. Must-haves: reliability + cost-per-run visibility across the fleet. Nice-to-haves: operating layer for eval/observability as the company scales. ICP confidence: Medium (co-founder and co-CEO at a qualifying Series C agent company; role is co-CEO rather than an explicitly technical title, hence Medium not High).
Jeffrey Davis — Senior Director of Technology Engineering, SoundHound AI
Signal: 4 (senior engineering leader at agent-native company). Source: https://www.linkedin.com/in/penland365/ (LinkedIn people search "SoundHound AI director engineering agents"). Company size: ~750-950 (SoundHound AI). Ships agents: YES (agentic/voice AI in production). Pain points (inferred from role): platform reliability + cost of production AI/agent workloads. Challenges: scaling agent infra. Must-haves: reliability + cost visibility. Nice-to-haves: unified observability. ICP confidence: Medium (title + company confirmed; found via company-scoped search).
Signal: 4 (ICP technical founder/CTO shipping AI agents in production). Source: https://www.anomalo.com/blog/anomalos-data-quality-imperative-with-founder-cto-jeremy-stanley/ ; https://www.anomalo.com/about/ ; Tracxn/CBInsights headcount. Company size: ~88 (as of Apr 2026); Series B; ~$72M raised over 3 rounds (SignalFire, Databricks Ventures). Founded 2018 (Stanley ex-Instacart SVP Data Science/Eng; co-founder w/ CEO Elliot Shmukler). Actively shipping agents: Anomalo ships AIDA (Anomalo Intelligent Data Analyst) + AI-powered data-quality monitoring agents that continuously validate structured/semi/unstructured data, plus an "AI Guardian" trust layer to safely deploy agentic AI at scale. Pain points (inferred from product/positioning, NOT verbatim quotes): trust/reliability of agents acting on enterprise data; visibility/control over what agentic AI does with data at scale; cost of continuously-running monitoring agents across large data estates. Challenges: giving enterprises control to deploy agents safely. Must-haves: agent observability/control, reliability guardrails, cost visibility. Nice-to-haves: compounding quality from feedback. ICP confidence: High (founder & CTO; in-range headcount; Series B; production data + agentic-trust product).
Jeremy Suriel — Co-Founder & Chief Technology Officer, Kustomer
Signal: 4 (ICP technical co-founder/CTO at a qualifying agent company). Source: targeted ICP search of CX platforms shipping AI agents; profile https://www.linkedin.com/in/jeremys/ (headline 'CTO, Kustomer'). Company size: ~350-500 (Kustomer, CX CRM now shipping AI agents; independent again post-Meta divestiture). Pain points [INFERRED, not a verbatim quote]: scaling AI agents across the CRM reliably in production; cost of LLM calls across very high message volumes; observability of agent performance per conversation. Challenges: retrofitting agentic AI onto an established CX platform; multi-tenant reliability. Must-haves: reliability + per-run cost visibility for agents. Nice-to-haves: agent eval tooling. ICP confidence: Medium-High (co-founder CTO at a ~400-emp company actively shipping AI agents).
Jeroen Van Hautte — Co-Founder & CTO, TechWolf
Signal: 1 (ICP writing about agent/LLM costs). Source: https://www.linkedin.com/posts/jeroenvanhautte_how-much-should-engineers-spend-on-ai-tokens-activity-7434153041135562752-CEVu — post "How much should engineers spend on AI tokens?". Company size: ~137 employees (Mar 2026), Series B ($42.75M led by Felix Capital, June 2024), SAP/Workday/ServiceNow Ventures backers; skills-intelligence AI infra, leads AI agents for the workforce. Pain points: engineer/team-level token budgeting, deciding how much AI spend is justified per engineer. Challenges: attributing and governing token spend as agents scale across the org. Must-haves: visibility into per-engineer/per-agent token cost, spend guardrails. Nice-to-haves: benchmarks for reasonable agent spend. ICP confidence: High (technical co-founder/CTO, 50-2000 emp, AI-native company actively building agents, publicly writing about token cost/control).
Jerry Zhou — Co-founder & CEO (technical, ex-Microsoft engineer), Supio
Signal: 4 (ICP building/shipping agents in production). Source: https://www.geekwire.com/2026/the-rise-of-vertical-ai-agents-and-the-startups-racing-to-build-them/ | Company size: ~100–200 employees (doubling from ~100 in Apr 2025), Series B ($60M), Seattle legal AI platform (personal injury / legal analysis). Technical co-founder (ex-Microsoft Office365 engineer). Pain points: describes vertical AI as "a shift from tools to agents"; agents must "operate within real workflows and take action," not just generate insights; reliability/trust bar — turning complex data (e.g. medical records) into "verified, structured outputs attorneys can rely on without second-guessing." Challenges: production reliability/accuracy high enough for regulated legal use; verifiable outputs. Must-haves: agent reliability and output verification; workflow-embedded action. Nice-to-haves: structured/traceable outputs. ICP confidence: High (technical co-founder & CEO at a 50–2,000-employee AI-native company shipping agents in production).
Jesse Zhang — Co-founder & CEO (technical), Decagon
Signal: 4 (ICP technical leader building/shipping agents in production). Source: https://www.linkedin.com/in/thejessezhang/ , https://research.contrary.com/company/decagon | Company size: 434 employees (as of May 2026). AI-native conversational-AI platform running autonomous CX agents for enterprises. Zhang is technical co-founder & CEO (CS, Harvard; ex-Google, ex-Citadel; prior founder of Lowkey, acquired by Niantic). Pain points: scaling many enterprise agents in production while keeping resolution quality high; usage-based billing means LLM/inference cost directly shapes gross margin; needs visibility into what each agent run costs and does. Challenges: reliability at enterprise scale; controlling agent behavior across many customer deployments; cost per run. Must-haves: production reliability, per-run cost visibility, observability. Nice-to-haves: behavior control/guardrails. ICP confidence: High (technical co-founder/CEO at a 50–2,000-employee AI-native company shipping agents in production). Note: same company as Ashwin Sreenivas (id 9) — distinct decision-maker.
Signal: 4 (ICP Head-of-AI persona at agent-native company shipping agents in production). Source: https://www.linkedin.com/in/criver ; https://www.moveworks.com/us/en/company/jiang-chen. Company size: ~1,000–1,400 (Moveworks; agentic RAG platform, 350+ enterprises, 5M+ workers). NOTE: Moveworks acquired by ServiceNow (Dec 2025). Oversees ALL machine learning / the agentic RAG platform (ex-Google/Airbnb/Yahoo; PhD Yale). Pain points (inferred): keeping agentic RAG accurate and reliable in production across many enterprise tenants; ML/agent quality at scale; cost per agent run. Challenges: seamless ML integration into products; reliability/eval at scale. Must-haves: reliability, eval/observability, cost efficiency. Nice-to-haves: agents that improve over time. ICP confidence: Medium-High (owns the AI/agent org — perfect Head-of-AI persona; softened by ServiceNow acquisition).
Signal: 4 (ICP shipping agents). NOTE: LinkedIn/Chrome NOT connected this run — found via web research. LinkedIn URL NOT captured (not fabricating); profile_url is the exec-listing source where the title was verified. Source: https://www.clay.com/dossier/sendbird-executives ; https://sendbird.com/news ; engineering blog https://sendbird.com/blog/category/engineering ; https://leadiq.com/c/sendbird/5a1d9d942300005e008d7074/employee-directory. Company size: ~302 employees as of July 2026 (LeadIQ). Agent platform confirmed: Sendbird describes itself as the omnichannel AI agent platform used by global enterprises — autonomous support and sales conversations, human-in-the-loop escalation for complex inquiries, proactive re-engagement messaging; 2026 platform updates push toward autonomous operation. Cost/observability signal: Sendbird engineering has publicly written about re-architecting global routing with a unified edge proxy specifically for cost efficiency, control, and observability at scale — a strong adjacent indicator that infra cost visibility is an active internal priority. Pain points (INFERRED from company-published material, NOT personal quotes): autonomous conversational agents at enterprise messaging volume; cost and observability at scale explicitly named as engineering priorities; human-in-the-loop handoff adds orchestration complexity. Challenges: moving from assisted to autonomous operation while keeping per-conversation economics viable; multi-tenant enterprise deployments. Must-haves: per-conversation/per-run cost visibility; observability across an omnichannel agent fleet. Nice-to-haves: routing/model-tiering to reduce spend on simple conversations. ICP confidence: High — CTO title, ~302 employees (in band), agent platform is the core product. UNVERIFIED: LinkedIn URL, personal pain statements; the cost/observability engineering post is company-authored, not attributed to Ku personally.
Signal: 4 (Director-level ICP at an agent company). Second contact at the new Talkdesk account. Found via LinkedIn people search + verification. Source: https://www.linkedin.com/in/jing-jin-a5b19317/ | Company: Talkdesk — AI/agent org (Voice AI Agents, AI Agent Platform). Company size: ~1,300. NOTE: later-stage than Series A-C target but within size band and shipping agents. Pain points (inferred from role/product, not verbatim quotes): production agent reliability, cost per run, observability across the agent platform. Challenges: scaling + cost control of the agent platform. Must-haves: reliability, cost visibility. Nice-to-haves: benchmarking, model routing. ICP confidence: Medium — Director-of-AI ICP title; size + agents confirmed; later-stage company noted.
Signal: 4 (ICP building/shipping agents in production — found via 2026 research on contact-center AI-agent companies). Source: https://www.observe.ai/about-us , https://ai-techpark.com/aitech-interview-with-swapnil-jain/ ; leadership confirmed via TheOrg (CEO Swapnil Jain, CTO & co-founder Jithendra Vepa). NOTE: no verified personal LinkedIn /in/ handle found — profile_url points to TheOrg org chart entry. Company size: 404 employees (Jan 2026); ~$214M raised (Series C-stage). Actively building AI agents: contact-center AI agent platform automating customer interactions with "predictable outcomes." Pain points (inferred, not verbatim): reliability of automated customer interactions; delivering predictable outcomes at enterprise scale; agent testing before deployment. Challenges: guaranteeing predictable agent behavior across large enterprise deployments. Must-haves: production reliability, conversation intelligence / observability. Nice-to-haves: cost-per-interaction visibility. ICP confidence: High (CTO & co-founder, 404-person contact-center company shipping agents in production).
Jithin George — Co-Founder & Head of Engineering, Lyzr AI
Signal: 4 (ICP technical co-founder / Head of Engineering at enterprise agent-platform company). Source: web search (Forbes Tech Council profile of CEO Siva Surendira; LinkedIn) + SiliconANGLE (Lyzr app builder for moving agents to production in volume) + Gravity funding tracker. Company size: ~120-190 employees (Latka 73→PitchBook/Tracxn ~190; doubled to 120 in Feb 2026) — within range. Lyzr = enterprise agent production-infrastructure platform for building/deploying agents across support, procurement, HR, sales; $100-125M Series B (Nexus Venture Partners, July 2026). Pain points: moving agents from pilot to production at volume; enterprise reliability, governance, and secure/sovereign AI infra for large agent fleets. Challenges: production-grade reliability + control across many business functions. Must-haves: deployment infra, observability, governance. Nice-to-haves: per-agent cost visibility, agents that improve over time. ICP confidence: High (Co-Founder & Head of Engineering at 120-190-person agent-native Series B company).
Signal: 1/4 (CTO at an enterprise production-agent platform with explicit governance/reliability focus; found via company/press primary sources, not LinkedIn — Chrome/LinkedIn unavailable this run). Source: https://techfundingnews.com/lyzr-ai-14-5m-accenture-250m-valuation/ and Lyzr leadership listings (Tracxn/LinkedIn). Company size: ~190 (confirmed, PitchBook June 2026). Stage: Series B ($100M, July 2026; ~$250M valuation); enterprise/on-prem production AI agents. Role note: Jithin Jimmy is Lyzr's CTO (ex-Ericsson and Cisco; previously founded an ad-tech startup). IMPORTANT DEDUPE NOTE: this is a DISTINCT person from Lyzr co-founder & Head of Engineering "Jithin George," who is already in the People Library — verify before any merge. Pain points: production AI-agent governance, reliability, and enterprise/on-prem deployment; token/cost optimization for enterprise agent workloads. Challenges: reliability and governance at enterprise scale; on-prem/air-gapped constraints. Must-haves: governance, reliability, observability. Nice-to-haves: token/cost optimization. ICP confidence: Medium (clean CTO role at a confirmed mid-size Series B agent company; flagged name-similarity dedupe risk with Jithin George).
Joao Moura — Founder & CEO, CrewAI
Signal: 4 (ICP senior technical leader at a company actively shipping AI agents; found via LinkedIn people-search, not an observed post). Source: https://www.linkedin.com/in/joaomdmoura/ | Company size: ~50-150 (Series A; multi-agent orchestration platform, enterprise 'agentic systems in production'). Pain points (INFERRED from role + company product focus — no observed quote): scaling agents reliably in production; controlling per-run LLM/agent cost as volume grows; visibility into what agents cost and how they behave across customers. Challenges: keeping agent accuracy/guardrails consistent at scale; token/context efficiency; cost blowout as they move from pilots to broad production. Must-haves: production reliability + observability into agent cost and behavior. Nice-to-haves: per-run cost attribution, context/token optimization, spend guardrails. ICP confidence: High (Technical founder/engineer who created the CrewAI multi-agent framework; core to agent tooling ICP).
Joe Chang — Chief Technology Officer, Suki
Signal: 4 (ICP technical decision-maker at company shipping AI into production and expanding to agents). Source: https://www.fiercehealthcare.com/ai-and-machine-learning/suki-adds-cto-cmo-cco-slate-executive-hires ; https://www.suki.ai/press-releases/suki-expands-executive-leadership-team-to-accelerate-the-future-of-ambient-ai-in-healthcare/ ; https://tracxn.com/d/companies/suki ; https://theorg.com/org/suki/org-chart/joe-chang . Company size: 410 employees (Apr 2026); well-funded ambient-AI healthcare company (Series D, $70M Oct 2024). Ambient clinical intelligence platform trusted by 400+ health systems, deep EHR integration; expanding beyond documentation into agentic workflows — orders, billing, and patient communication ("powering every interaction"), positioning as an integration/agent layer across healthcare workflows. Chang joined as CTO (July 2025) to scale engineering; ex-Uber, Google, Primer.ai, Marin Software. Pain points: reliability/accuracy for clinical-grade AI in a regulated domain; scaling agentic workflows (orders/billing/comms) safely; trust with clinicians. Challenges: production reliability across many health systems; controlling agent behavior in high-stakes clinical settings; cost at scale across 400+ systems. Must-haves: reliability, control/governance, auditability. Nice-to-haves: per-run cost visibility, model routing. ICP confidence: High (CTO / technical decision-maker at a 410-person AI-native healthcare company moving from ambient AI into production agents). Note: Suki's agent expansion is recent — reliability/control wedge strongest as they scale agentic actions.
Signal: 4 (ICP building/shipping agents in production). Source: https://www.geekwire.com/2026/the-rise-of-vertical-ai-agents-and-the-startups-racing-to-build-them/ | Company size: ~130 employees (2026), Seattle infrastructure-as-code company; ships production AI agent "Neo" that automates cloud infra tasks (cost optimization, compliance, deployments). Duffy is a deeply technical founder (ex-Microsoft, creator of Pulumi). Pain points: agent trust/quality — "You need to first build trust and build quality into the systems you're building"; controlling autonomy for high-stakes actions (references Karpathy's "autonomy slider," human oversight for prod infra changes); orchestrating "swarms of agents" collaborating; notes the hard part "is not just LLM tokens — it's much more complex," i.e. the agent harness (orchestrate tasks, find context, verify outputs). Challenges: reliability/verification of agent actions on complex systems; giving every engineer a team of agents ("100x developers"). Must-haves: control over agent behavior/autonomy; output verification; context management. Nice-to-haves: multi-agent orchestration. ICP confidence: High (technical founder/CEO at a 50–2,000-employee company running agents in production).
Joe Durante — Senior Vice President of Artificial Intelligence, Optimal Dynamics
Signal: 4 (ICP owning the agentic automation layer at a company shipping autonomous freight agents). Source: https://www.optimaldynamics.com/our-team ; product: https://www.truckinginfo.com/news/optimal-dynamics-launches-ai-system-to-help-carriers-choose-better-freight
Company size: ~82 employees across 3 continents (Apr 2026, Crunchbase) — in range. Series C, $40M led by Koch Disruptive Technologies (Apr 2025). Logistics/freight decision automation.
Pain points: Isolated, shallow agents that do not compose. Company framing: "Even traditional agentic AI tools... respond to individual requests or operate in isolation," and CEO Daniel Powell: "Carriers don't just need automation; they need a system that understands their network and proactively works to improve it." Their "Scale" product runs autonomous agents that search, negotiate, and procure freight across multiple channels — a genuine 1→many agent orchestration problem.
Challenges: Coordinating multiple autonomous agents (search, negotiate, procure) against a shared network state; keeping agent decisions consistent with the network optimizer; operating agents that transact real money on carriers' behalf.
Must-haves: Cross-agent state/context sharing; control over what agents are allowed to commit to; visibility into why an agent took a given action.
Nice-to-haves: Cost-per-load-booked attribution; per-agent performance benchmarking.
ICP confidence: Medium — SVP of AI is a clean VP-AI/ML ICP match, headcount and Series C stage fit well, agents demonstrably in production; pain-point quote is CEO-attributed rather than personal to Durante.
Signal: 4 (adapted — recent-hire web search + LinkedIn verification; content/post search low-yield this run). Source: https://rogo.ai/news/joe-xavier and https://www.linkedin.com/in/joexavier/. Company size: ~150-300 (estimate; Series C finance-AI, rapidly scaling). NEW CTO (joined ~July 2026; ex-Grammarly CTO, prior Amazon/Twitter/Microsoft). Rogo builds AI research/analysis agents for investment banks, PE & asset managers. Pain points (INFERRED, no quotes): accuracy/reliability of agents on financial data; cost per research run; security/compliance in a regulated vertical; scaling agents across many banking clients. Challenges: making long-running research agents trustworthy + cost-controlled. Must-haves: per-run cost visibility, reliability/eval guardrails. Nice-to-haves: model routing, governance. ICP confidence: High (C-level technical, in-band 50-2000, agent-native, direct cost/reliability exposure).
Joel Hellermark — Co-Founder & CEO (technical founder), Sana (Sana Labs)
Signal: 4 (ICP building/shipping agents). Source: https://sanalabs.com/products/sana/ai-agents ; https://www.nea.com/blog/sana-ceo-joel-hellermark-interview ; https://www.linkedin.com/in/joel-hellermark/
Company size: ~495 employees (Stockholm/London/NY); Series C ($55M Nov 2024, led by NEA; $130M+ total). Independent, eyeing US IPO.
Company context: Sana builds enterprise "expert AI agents" and an agent platform for knowledge work ("superintelligence for work"). Joel is a technical founder (started coding/ML young; founded Sana 2016).
Pain points (inferred from product/scale): deploying expert AI agents across a large enterprise reliably; controlling agent behavior and grounding on enterprise knowledge; cost of running agents at org scale.
Challenges: reliability and trust of enterprise-facing agents; scaling many agents across teams; visibility into agent behavior/cost.
Must-haves: reliable, grounded enterprise agents; control and governance at scale.
Nice-to-haves: cost-per-run visibility; agents that compound/improve over time.
ICP confidence: Medium-High — technical co-founder & CEO (decision-maker) at a well-funded 50–2,000 company actively shipping an AI agent platform. Role is CEO rather than CTO, but qualifies as technical co-founder.
Signal: 4 (Head-of-AI-equivalent decision-maker steering AI strategy around Cohere's agent platform North; surfaced via same Cohere agent-platform search).
Source: https://techcrunch.com/2025/08/14/cohere-hires-long-time-meta-research-head-joelle-pineau-as-its-chief-ai-officer/ ; https://cohere.com/blog/north-ga
Company size: Cohere ~800+ employees early 2026. Fits 50–2,000.
Background: Ex-Meta VP of AI Research; joined Cohere Aug 2025 as first Chief AI Officer, overseeing AI strategy across research, product & policy. Publicly focused on gearing research around North — developing agents in private/secure settings and building benchmarks to evaluate them.
Pain points (inferred from role/company focus): evaluating and controlling agent reliability; secure agent development; measuring agent performance rigorously.
Challenges: creating benchmarks/evals for agent systems; aligning research with production agent needs.
Must-haves: agent evaluation/benchmarking, reliability, security.
Nice-to-haves: compounding/self-improving agent research tooling.
ICP confidence: Medium (senior AI decision-maker at qualifying-size AI-native company building agents; buying-intent caveat as with model-lab peers).
Signal: 4 (ICP leader at a company actively shipping AI agents). Found via LinkedIn company People-tab sweep of Parloa, not via a post — LinkedIn content search was unusable this run (see run summary entry).
Source: https://www.linkedin.com/company/parloa/people/?keywords=VP
Company size: 201-500 employees (per Parloa LinkedIn company page, Berlin). Parloa = "AI agent management platform" for customer service; Series C, agents running in production at enterprise contact-centre volume.
Pain points: NONE DIRECTLY OBSERVED THIS RUN — no post or comment by him was found. Inferred profile only, based on role + company: voice/chat agents at contact-centre volume mean per-conversation inference cost is a direct COGS line; multi-agent orchestration across enterprise tenants.
Challenges (inferred): per-tenant cost attribution across many concurrent live agents; latency vs. model-quality tradeoff in real-time voice.
Must-haves (inferred): per-run and per-tenant cost visibility; reliability guardrails for live customer conversations.
Nice-to-haves (inferred): routing/model-swap experimentation, shadow testing before rollout.
ICP confidence: High on role + company fit (VP Engineering, agent-native company, right headcount, Series C). LOW on demonstrated pain signal — no expressed pain captured. Worth an activity re-check next run before outreach.
Note: Parloa colleague Masashi Beheim (VP Engineering) is already in the brain (ids 433 and 587 — pre-existing duplicate worth merging).
Johannes Hagemann — Co-Founder & Chief Technology Officer, Prime Intellect
Signal: Bucket 4 — ICP building/shipping agents at a qualifying company. Found via web research (funding news); LinkedIn/Chrome unavailable this run.
Source: https://techcrunch.com/2026/07/08/prime-intellect-raises-130m-series-a-to-help-enterprises-build-their-own-ai-agents/ (titles corroborated at https://mlq.ai/news/prime-intellect-raises-130m-series-a-at-1b-valuation-for-enterprise-ai-agent-training/)
Company size: 56 employees as of 2026-06-30 (Tracxn: https://tracxn.com/d/companies/prime-intellect/__c00KIKAVH9b1POsp5fnIynw_8joG2ie8qsqmCTAB6Bc). $130M Series A at $1B valuation, 2026-07-08. $100M+ ARR, ~6,000 customers.
Agent evidence: full stack for enterprises to train and deploy their own agents — compute, RL frameworks, sandboxes, evaluation tooling. Named customers building production agents include Ramp and Zapier.
Pain points: enterprises can't get from agent demo to production-grade without infra expertise; dependency and data-control risk from closed frontier labs; cost/performance of running agents in production (article cites a Ramp agent beating frontier models at lower cost).
Challenges: scaling compute + RL infrastructure reliably across ~6,000 customers; inferred — as CTO of the infra layer he owns the same cost/reliability problems internally that Alpha sells to solve.
Must-haves: cost visibility and control across model + compute spend; reliability tooling for continuously running agents; enterprise audit/control (inferred from company positioning).
Nice-to-haves: integration with existing RL/eval stack; multi-model routing (inferred).
ICP confidence: Medium-High — clean CTO/co-founder title, headcount 56 (just above floor), Series A, company is literally infrastructure for shipping agents. Downgraded from High because Hagemann was not personally quoted (CEO Vincent Weisser was); pain is company-level.
John Capobianco — Head of AI & Developer Relations, Itential
Signal: Bucket 1 (ICP publishing about building/running agents in production) + Bucket 4 (author of production-agent content).
Source: https://www.itential.com/author/john-capobianco/ ; https://www.itential.com/blog/company/ai-networking/from-intent-to-impact-building-production-network-agents-in-minutes ; https://futr.tv/thefeed/2026/7/30/the-end-of-handwritten-network-configs-with-itentials-john-capobianco
Company size: ~200 employees (Craft/Growjo). Atlanta, GA. ~$25.5M raised (Tech Square Ventures, Elsewhere Partners) — venture-backed, roughly Series B/C scale.
Agent activity (verified): Itential publishes "From Intent to Impact: Building Production Network Agents in Minutes." Capobianco personally built NetClaw, an open-source CCIE-level AI agent that manages network infrastructure via Slack and WhatsApp — analyzes packet captures, configures routers, runs compliance tests, writes its own documentation (300 GitHub stars in two weeks). Previously Head of DevRel at Selector AI (AI-driven network observability) and Cisco AI Technical Leader.
Pain points: autonomous agents taking write actions on production network infrastructure — reliability and blast radius are the dominant concern; also multi-agent orchestration across network + IT + business systems.
Challenges: moving network teams from scripted automation to agentic operations without losing determinism/auditability; agent behavior verification before a config push.
Must-haves: guardrails and approval/observability layer for agents that mutate infrastructure; visibility into what each agent run actually did and cost.
Nice-to-haves: cost-per-run attribution for agent workflows; reusable agent evaluation for compliance tests.
ICP confidence: High on title/role fit (Head of AI, explicitly in ICP) and on agent-in-production activity; Medium on company profile (200 emp is in range; total raised $25.5M is modest for Series A–C framing). Very high-signal publicly — active podcaster, blogger, open-source agent author, runs the 400+ member VibeOps Forum. Best warm-inbound target of this run.
John Cottongim — Co-Founder & CTO, Roots (Roots Automation)
Signal: 4 (ICP technical co-founder/CTO shipping agents in production). Source: https://www.roots.ai/company/leadership/john-cottongim ; https://www.crunchbase.com/person/john-cottongim ; pulse2.com Chaz Perera interview ; RocketReach/PitchBook headcount. Company size: ~123-147 (RocketReach 123; PitchBook 147; founded 2018; ~$57.4M raised; ~$16M revenue). Actively shipping agents: Roots ships an insurance "Digital Coworker"/agentic AI platform automating underwriting, claims and policy servicing for carriers/MGAs. Cottongim = technical co-founder & CTO (ex-AIG, ex-Mars Enterprise Automation Hub). Pain points (inferred from product + regulated insurance domain, NOT verbatim quotes): reliability/accuracy of agents making claims & underwriting decisions; auditability/compliance of agent actions on policyholder data; cost-per-transaction as document/claim agents scale. Challenges: proving trust to carrier compliance teams while scaling many agents. Must-haves: audit trails, human-in-the-loop control, per-run cost/accuracy visibility. Nice-to-haves: agents that learn from adjuster feedback. ICP confidence: High (technical co-founder + CTO; in-range headcount; agentic AI in production in a regulated vertical).
John Edstrom — Engineering Director, Augment Code
Signal: 4 (ICP eng leader at an agent-native company; LinkedIn people-search 'Augment Code engineering director'). Source: https://www.linkedin.com/in/john-edstrom-9625408/ . Company size: ~150-300 (Augment Code; Series B, ~$250M raised; agentic SDLC / autonomous coding agents for software teams). Pain points (INFERRED from role + company; no verbatim quote observed this run): reliability and token/context efficiency of coding agents on large codebases; cost-per-run control as agent usage scales; production observability for agent workflows. Challenges: making agentic SDLC reliable and cost-efficient at enterprise scale. Must-haves: reliability guardrails + per-run cost/token visibility. Nice-to-haves: context/token-waste reduction. ICP confidence: High (Engineering Director at a 50-2,000-emp company actively shipping coding agents). Note: brain already has Augment's CTO Dion Almaer, Chief Scientist Guy Gur-Ari, and VP Eng Vinay Perneti — Edstrom is net-new.
Signal: 4 (ICP at a company mid-pivot to agents in production). Source: https://www.skedulo.com/leadership/, https://www.skedulo.com/platform/ai-scheduling-operations/, https://www.skedulo.com/news/skedulo-raises-75m-in-series-c-surpassing-35-million-appointments-scheduled/ — title verified live on LinkedIn 2026-08-30 ("CTO", Skedulo, Brisbane). Company size: 256 employees (2026); Australian-founded (Brisbane), SF HQ, offices in Australia, Vietnam and UK; Series C $75M led by SoftBank Vision Fund 2, $255M total. Pain points: Skedulo's 2026 platform repositioning is an explicit shift from scheduling tool to autonomous operations system — "AI Agents — every operational blocker, removed" — running against 35M+ scheduled appointments across 150 enterprise customers. Deskless-workforce agents firing per-shift and per-job produce very high run counts against relatively low revenue per run. Challenges: this is a classic 1-to-5+ agent scaling moment layered onto an existing non-AI SaaS cost base, where agent spend has to be justified against per-appointment unit economics; multi-region deployment across three continents. Must-haves: per-run and per-customer cost visibility so the agent pivot does not erode gross margin. Nice-to-haves: reliability SLAs they can put in enterprise contracts. ICP confidence: Medium — title, size, stage all fit and the agent pivot is real and recent; caveats are that legal HQ is San Francisco rather than APAC, and I found no first-person statement from McKim on agent cost or reliability, so pain is inferred from the company's product direction.
John Nay — Founder & CEO (technical founder — AI+law researcher, PhD), Norm Ai
Signal: 3 (building the verification/control layer for AI agents — directly adjacent to agent control, reliability & observability). Source: https://www.prnewswire.com/news-releases/norm-ai-raises-120-million-at-a-1-2-billion-valuation-led-by-khosla-ventures-to-deliver-the-full-stack-model-for-legal-ai-302819152.html ; https://siliconangle.com/2026/07/07/norm-ai-nabs-120m-1-2b-valuation-bring-ai-agents-law/. Company size: ~100-150 (Norm Ai, Series C — $120M at $1.2B valuation, July 2026; AI-native legal/compliance). Pain points: verifying & governing AI agents operating in regulated environments; agents supervising other legal AI agents; reliability/compliance oversight at scale (clients >$30T AUM). Challenges: production reliability & auditability of legal AI agents; controlling agent behavior under law. Must-haves: verification layer, governance/compliance controls, reliable agent oversight. Nice-to-haves: cost visibility across multi-agent verification workloads. ICP confidence: High — technical founder (PhD, ex-Stanford AI+law researcher) at a Series C agent-native company, 50-2000 employees. Note: CEO title, but deeply technical decision-maker.
Signal: 4 (ICP technical co-founder/CTO at agent-native legal company). Found via web research + verification — LinkedIn content/people search UNAVAILABLE this run (session not authenticated / login wall). Source: https://www.forbes.com/sites/rashishrivastava/2026/03/31/ (Crosby AI law firm) ; https://www.artificiallawyer.com/2026/06/18/the-crosby-story-with-co-founder-ryan-daniels/ ; https://baincapitalventures.com/insight/crosby-is-redefining-legal-work-with-ai-powered-contract-automation/. Company size: ~50 (engineering team scaled 0->25+ in under a year plus ~two dozen lawyers). SIZE CAVEAT: headcount is right at the ~50-employee lower bound. Series B April 2026, ~$400M valuation; grew negotiated contract volume $30M->$1B; ~100 clients incl. Cursor, Clay, Cognition. Actively shipping agents: Crosby is an "AI law firm/neofirm" where AI agents + lawyers review and negotiate contracts in production (billed by the page). John Sarihan is Co-Founder & CTO (ex-Ramp engineering manager). Pain points (INFERRED): reliability/accuracy of contract-review agents; keeping agent output auditable alongside human lawyers; per-run cost as contract volume scales. Challenges: production-grade accuracy for legal agents at speed. Must-haves: reliability, accuracy/auditability. Nice-to-haves: per-run cost visibility. ICP confidence: Medium (technical co-founder/CTO at agent-native company; headcount right at the ~50-emp floor). LinkedIn URL not confirmed this run — left blank to avoid fabrication.
John Stecher — Chief Technology Officer, Norm Ai
Signal: 3 & 4 — Norm Ai builds legal/regulatory-compliance AI agents deployed for in-house legal/compliance teams; product is increasingly used to SUPERVISE other AI agents operating in regulated environments (directly adjacent to an agent operating/control layer — strong strategic fit). Founder/CEO: John Nay (AI & Law researcher, ex-Stanford). Source: https://siliconangle.com/2026/07/07/norm-ai-nabs-120m-1-2b-valuation-bring-ai-agents-law/ ; https://www.norm.ai/company. Company size: estimate ~100-200 (NOT officially disclosed; Series C $120M at $1.2B valuation, July 2026; $260M+ raised; NYC/WTC HQ). Clients represent $30T+ AUM. Pain points: reliability/controllability of legal agents in regulated settings; supervising and monitoring fleets of agents; auditability of agent actions. Challenges: ensuring agent outputs meet legal/regulatory bar; visibility into agent decisioning. Must-haves: agent supervision, reliability, audit trail. Nice-to-haves: cost visibility per run. ICP confidence: Medium (CTO = clean ICP decision-maker; Series C; size inferred, verify before outreach).
John Wang — Co-founder & CTO, Assembled
Signal: 4 (ICP building/shipping agents in production — found via 2026 research on AI support-agent companies). Source: https://aiproem.substack.com/p/founder-ep-assembled-co-founder-on (podcast: building an AI-native support system for enterprises), https://www.assembled.com/blog/why-i-code-as-a-cto . Company size: 145 employees (May 2026); Series B ($70.7M total). Actively building AI agents: stood up an AI-native support-agent product as a "startup within a startup"; agents resolve ~30-60% of incoming issues depending on knowledge coverage. Pain points (inferred, not verbatim): reliability + coverage gap of support agents (resolution rate swings with knowledge quality); standing up a new agent product alongside the core WFM business. Challenges: raising and stabilizing agent resolution rates; production reliability. Must-haves: production reliability, agent quality/coverage measurement. Nice-to-haves: cost-per-resolution visibility. ICP confidence: High (CTO & co-founder, 145-person Series B shipping support agents in production).
Jon Noronha — Co-Founder & Chief Product Officer, Gamma
Signal: 4 — AI-focused product leader / co-founder at an AI-native company (Harvard-trained engineer; ex-VP Product Optimizely; ML on Bing). Speaking at SaaStr AI Annual 2026 on shipping AI at scale.
Source: https://theorg.com/org/gamma-app/org-chart/jon-noronha ; https://www.saastr.com/the-first-44-speakers-for-saastr-ai-annual-2026-the-founders-and-operators-actually-shipping-ai-at-scale/
Company size: ~360 employees (Series B, $100M ARR, 70M users)
Pain points: shipping reliable AI product features at scale; balancing generation quality vs cost
Challenges: productizing AI at consumer scale reliably
Must-haves: reliability and quality visibility
Nice-to-haves: cost-per-generation insight
ICP confidence: Medium — VP of Product (AI-focused) analog at C-level, AI-native, right size/stage
Jon Perl — Co-Founder & CEO (technical; ex-CTO Zipdrug), QA Wolf
Signal: 4 (ICP technical co-founder shipping a multi-agent system) + 1 (agent reliability/cost at scale). Source: https://theorg.com/org/qa-wolf/org-chart/jon-perl ; https://www.citybiz.co/article/580809/qa-wolf-raises-36m-series-b/ ; https://www.qawolf.com/blog/qa-wolf-raises-36-million-series-b-and-opens-mobile-app-waitlist . Company size: 248 employees (Apr-2026). Series B ($36M, Scale Venture Partners; $56M total). Agentic automated-testing platform: multi-agent QA Wolf AI for test creation, retry bots for flake reduction, 24-hr AI failure investigation (AI + human oversight). Technical co-founder/CEO (prev CTO Zipdrug, head of tech at Dispatch). Pain points (INFERRED): reliability/flakiness of AI-generated tests; cost of running multi-agent test-generation at scale; visibility into agent behavior & failures. Challenges: guaranteeing flake-free output while controlling agent run cost. Must-haves: reliability/flake reduction; per-run cost control; failure observability. Nice-to-haves: compounding test-quality improvement. ICP confidence: MEDIUM-HIGH (technical co-founder/CEO at 248-emp Series B company shipping a multi-agent system in production). Note: exact LinkedIn handle not verified; profile_url points to The Org profile.
Jon Wang — Co-Founder & Co-CEO, Assort Health
Signal: 4 (technical co-founder shipping voice agents at scale; regulated vertical). Source: https://www.assorthealth.com/blog/assort-health-raises-120-million-series-c-to-scale-largest-deployment-of-ai-agents-for-the-patient-journey ; https://www.fiercehealthcare.com/ai-and-machine-learning/assort-health-scores-120m-series-c-scale-voice-ai-agent-platform-healthcare . Company size: ~172 employees (Tracxn, May 2026; noted rapid ~10x YoY growth); Series C $120M led by Menlo Ventures at $1.2B valuation, ~$222M total. Co-founders Jon Wang (Stanford ML researcher, 7 AI+healthcare papers) and Jeff Liu (ex-Facebook SWE); Jon is the strongest technical decision-maker. Pain points (inferred from public product/positioning; no first-party quotes captured): operates the most widely-used voice AI agent platform for the patient journey (scheduling, intake, referrals, forms, document processing, refills, payments) built on 190M+ patient interactions — reliability + accuracy are critical in a regulated healthcare setting; real-time latency for voice; scaling agents across specialties. Challenges: reliability/accuracy of regulated healthcare voice interactions at scale; latency; auditability. Must-haves: reliability, auditability/governance, real-time performance. Nice-to-haves: cost-per-call visibility, continuous model improvement. ICP confidence: Medium-High (technical co-founder, Series C, agents at scale in production, regulated vertical; headcount ~172 confirmed in range; title is Co-CEO not CTO).
Jonas Diezun — Co-founder & CEO (technical), Beam AI
Signal: 4 (ICP building/shipping agents in production; writing about production agent reliability). Source: https://beam.ai/about and https://beam.ai/agentic-insights/top-5-ai-agents-in-2026-the-ones-that-actually-work-in-production. Company size: NOT verified this run (Tracxn funding figure looked incomplete; likely small-to-mid, needs confirmation). Product: agentic automation platform running governed agents on an organizational ontology; positions around solving the "maintenance trap" — agents that break when an SOP/process is updated. Pain points: agents breaking when business processes/SOPs change; keeping governed agents reliable as the org changes. Challenges: durable, self-updating agents in production without constant rework. Must-haves: governance + adaptability of agents to changing processes, reliability in back-office automation. Nice-to-haves: continuous learning from interactions. ICP confidence: Medium-Low — clear technical founder building/shipping agents, but company headcount unverified this run (confirm 50–2,000 before outreach).
Signal: 4 (ICP CTO publicly building and shipping agents at a qualifying company). Source: https://www.artificiallawyer.com/2026/04/08/clio-rolls-out-agents-for-work-and-vincent/ ; https://www.lawnext.com/2025/12/on-lawnext-inside-clios-ai-driven-transformation-cpo-john-foreman-and-cto-jonathan-watson.html ; https://legaltech.ca/2026/05/20/clio-cto-headlines-legal-tech-fireside/ ; https://www.clio.com/about/team/. Company size: ~1,200-1,600 (legal practice-management SaaS; late-stage private, post-vLex acquisition — above the Series A-C band but inside the 50-2,000 headcount band). Pain points: running an AI-driven transformation of a large existing SaaS platform while shipping multiple agent products (Clio Duo, Agents for Work, Vincent) into a risk-averse, regulated professional-services customer base. Challenges: multi-agent rollout across a merged Clio + vLex platform; accuracy and trust bar for legal work; integrating vLex legal intelligence into agentic workflows; scaling from a single assistant to a fleet of agents. Must-haves: reliability and defensibility of agent output for legal customers; a way to operate several agent products on one platform. Nice-to-haves: unified cost/usage reporting across agent products. ICP confidence: High on role and agent activity (CTO, company shipping 3+ named agent products, headcount in band); Medium on funding-stage fit (late-stage, not Series A-C). Public pain-point language on cost/token specifically was NOT found — treat cost pain as unverified. NOTE: The brain also lists an 'Angel' and a 'David Thompson' as Clio VPs of Engineering per public reporting; not added this run pending confirmation of full name / profile URL.
Signal: 4 (ICP shipping an AI decision agent, quoted in the launch). Found via web research — LinkedIn/Chrome unavailable this run.
Source: https://www.worktruckonline.com/news/fleetio-launches-ai-powered-service-advisor (11 Mar 2026)
Company size: ~460 employees as of May 2026 across 3 continents (LeadIQ company profile). Fleet maintenance & logistics operations SaaS. Series D — stage is the ICP stretch.
Agent evidence: launched Service Advisor into open beta at Work Truck Week 2026 — AI that assesses repair orders, assigns issue priorities, flags exceptions and summarizes service/approval history inside Fleetio's Maintenance Shop Network. Early results: assets spending 16% fewer hours in the shop. Also shipped Smart Uploads (AI extraction of invoices/receipts into structured maintenance data; customers report 20 min per multi-page invoice saved).
Verbatim: "We built Service Advisor to strengthen human judgment as fleet operations scale across assets, shops, and locations. Because Fleetio sits at the center of the maintenance workflow, from issue to repair to outcome, we're able to bring the right context forward, apply standards consistently, and automate much of the coordination work that traditionally slows fleets down. Service Advisor is a major step toward our broader vision for Fleetio to become the operational brain of the fleet."
Pain points: approval decisions vary "depending on who's reviewing and how much context they have in the moment"; issue backlogs across assets and locations causing downtime; manual invoice data entry.
Challenges: surfacing the right context from a decade of fleet service records at decision time; applying consistent standards whether a seasoned expert or a stand-in reviews the work; expanding from point AI features to an end-to-end automated maintenance workflow.
Must-haves: grounded guidance inside the existing approval workflow (not a separate tool); consistent policy application + exception flagging; human-in-the-loop framing; measurable downtime/cost outcomes.
Nice-to-haves: document-to-structured-data extraction; smart prioritization across multi-location fleets; narrative summarization of service history for fast verification.
ICP confidence: Medium — CTO, ~460 employees, actively shipping AI decision agents in 2026; Service Advisor is currently advisory/assistive rather than fully autonomous and Fleetio is Series D. LinkedIn URL not confirmed from a source, left blank rather than guessed.
Joschka Braun — Head of ML (headline: "Leading ML @ Tennr"), Tennr
Signal: 4 (ML leader at an agent-native company; surfaced via company-scoped LinkedIn people search, activity reviewed).
Source: https://www.linkedin.com/in/joschkabraun/recent-activity/all/ — recent activity is largely reposts of Tennr domain content (e.g. a colleague's post on the CMS DME enrolment moratorium effective Feb 2026 and what it means for referral volume), i.e. he amplifies the operational-load problem Tennr's agents absorb rather than posting about agent infrastructure directly. Company page: https://www.linkedin.com/company/tennrai/about/
Company size: 201-500 employees (LinkedIn company page, verified this run). Tennr (YC S23) runs document/referral-processing agents across US healthcare providers — high-volume, per-document agent runs.
Pain points: Not directly expressed in the posts reviewed. Domain context he amplifies: referral volume keeps rising while supply consolidates, which pushes more work through the agent pipeline.
Challenges: Leading ML for a document-agent pipeline where volume is growing and each run has a real unit cost — the per-run economics question is structurally live at Tennr, but he has not stated it publicly.
Must-haves: Unknown / not stated publicly.
Nice-to-haves: Unknown / not stated publicly.
ICP confidence: Medium — he leads ML (functionally head-of-function, so at/above the Director bar, though the headline is informal), and the company is 201-500 and demonstrably agent-native. Downgraded from High because no cost/reliability pain is stated in his own words; verify seniority before outreach.
Signal: 4 (ICP AI leader at agent-native finance company). Found via web research + verification — LinkedIn content/people search was UNAVAILABLE this run because the logged-in LinkedIn session was not authenticated (login wall); pivoted to web verification as prior runs did. Source: https://rogo.ai/news/welcoming-our-new-head-of-applied-ai-joseph-kim ; https://openai.com/index/rogo/ ; https://www.linkedin.com/in/joseph-kim-gradients/. Company size: ~50-100 employees (Rogo, NYC; Series B/C-stage AI finance research platform serving 5,000+ bankers; multi-model o1/GPT-4o architecture). Actively shipping agents: Rogo automates financial research, analysis and workflows for investment banks/investors via AI agents in production. Joseph Kim leads Applied AI (ex-Google Gemini RLHF researcher; ex-NASA research scientist; MIT Media Lab). Pain points (INFERRED from role/company context, not a verbatim quote): reliability/accuracy of finance-research agents; multi-model routing and token/cost efficiency; eval of agent outputs on high-stakes financial analysis. Challenges: keeping agent answers accurate and auditable at scale for regulated finance users. Must-haves: production reliability, per-run cost/token visibility, robust evals. Nice-to-haves: automated model routing / cost optimization. ICP confidence: High (Head of AI, technical, at 50-100-emp agent-native finance company in target band).
Signal: 4 (technical co-founder at net-new agent-native company; LinkedIn people search + web verification). Source: https://www.linkedin.com/in/joshalbrecht/ ; verified via theorg.com/Crunchbase/imbue.com. Company size: ~83 employees (Tracxn, Mar 2026). Company context: Series C (~$232M total; $50M Series C led by Tiger Global). Imbue builds AI systems/agents that reason and code, shipping coding/reasoning agent products. Pain points (INFERRED from verified role+company, NOT verbatim): reliability of long-running reasoning/coding agents in production; compute/token cost of reasoning-heavy multi-step agent loops; context efficiency. Challenges: scaling reliable coding agents; controlling cost of reasoning-heavy runs. Must-haves: per-run visibility & control over agent cost + reliability. Nice-to-haves: per-step token profiling, eval tooling. ICP confidence: HIGH — technical co-founder/CTO, 83 emp (in 50-2000), Series C, agent-native, NET-NEW company. Caveat: pain points inferred, not quoted.
Signal: 4 (ICP leader at a company actively shipping AI agents). Found via LinkedIn company People-tab sweep of Clarifai.
Source: https://www.linkedin.com/company/clarifai/people/?keywords=engineering
Company size: 51-200 employees (per Clarifai LinkedIn company page, NYC). Clarifai = AI platform that has repositioned toward compute orchestration / reasoning + agent runtime; Series C.
Pain points: NONE DIRECTLY OBSERVED THIS RUN. Inferred only: as a platform selling inference and agent orchestration, their own gross margin is set by how efficiently agent runs consume compute.
Challenges (inferred): GPU/inference utilisation per agent run; giving customers cost visibility they can trust.
Must-haves (inferred): granular per-run cost telemetry.
Nice-to-haves (inferred): routing and caching levers exposed to customers.
ICP confidence: Medium. Role fits (Director of Engineering) and headcount fits. CAVEAT: Clarifai sits partly in the agent-infrastructure/orchestration space themselves, so they may read as a peer or competitor rather than a buyer — qualify positioning before any outreach.
José Caldeira — Head of Production Engineering, Poolside
Signal: 4 (senior technical leader at a company building/shipping AI coding agents). Source: https://theorg.com/org/poolside-1/org-chart/jose-caldeira | company: https://poolside.ai/. Company size: estimated ~200–350 employees (NOT exactly verified this run — estimate based on Poolside's ~$500M Series B raise in 2024, ~$3B valuation); Series B — within Series A–C. Poolside builds frontier code models and coding agents ("operational intelligence"). Role since May 2024. Pain points (inferred from Head of Production Engineering at a coding-agent company): reliability of agents in production, cost of large-scale inference/token consumption for agentic coding workloads, observability into per-run cost and failure. Challenges: scaling agent inference infra economically; keeping long-running agent sessions reliable and cost-bounded. Must-haves: cost-per-run/inference visibility, production reliability, guardrails. Nice-to-haves: model routing, context/token optimization. ICP confidence: Medium (title = Head-of/Director+; company in ICP on stage and agent focus; exact headcount is an ESTIMATE, not verified; pain inferred from role + company). METHOD NOTE: Chrome/LinkedIn NOT connected this run; verified title/company via web research (org chart). Company headcount not confirmed to a source — flagged as estimate. LinkedIn URL not directly verified.
Signal: 2 (non-ICP author's LinkedIn post surfacing an ICP company shipping agents in production). Source: LinkedIn post by Raj P. (AI @ RapidClaims), which states RapidClaims has 'AI already live in production across US health systems' with 'in-house LLM infra, RL, fine-tuning, agents, evals, retrieval — with a relentless focus on latency, reliability, and cost' (company: https://www.linkedin.com/company/rapidclaims-ai). Company size: ~59-89 employees (Tracxn 89 / PitchBook 59, mid-2026); Series A, $11.1M raised (Accel, Together Fund), founded 2023, AI-native healthcare revenue-cycle management. Pain points: unpredictable LLM/token cost of agents at production scale; latency; reliability of multi-step medical-coding/denial agents; eval + retrieval quality for high-stakes healthcare RCM. Challenges: holding reliability & low latency while controlling LLM spend across live health-system deployments. Must-haves: production reliability, cost control per run, evals/observability. Nice-to-haves: retrieval tuning, model-cost optimization. ICP confidence: High (technical co-founder / CPTO = valid C-suite technical target; AI-native company actively shipping agents in production; headcount in 50-2,000 range). Note: person identified via RapidClaims leadership (Crunchbase/theOrg/press); pain points inferred from the company's publicly documented production-agent focus, not a personal quote.
Jove Zhong — Head of Forward Deployed Engineering (FDE), Cresta
Signal: 4 (engineering leader deploying agents into customer production; found via LinkedIn people search of Cresta). Note: Cresta's CTO/founders (Tim Shi, Daniel Hoske, Ping Wu) are already in Alpha Brain; Jove Zhong is net-new.
Source: https://www.linkedin.com/search/results/people/?keywords=Cresta%20AI%20head%20of%20engineering%20director%20agents
Company size: ~500-700 employees; Cresta, Series D ($270M+ raised, $1.6B val), real-time AI agents for contact centers. In-band on size (50-2,000); stage past Series C noted.
Pain points (inferred from role): standing up and running customer agents in production reliably; cost/observability across many deployed agents.
Challenges: production reliability and cost visibility across a fleet of customer agent deployments.
Must-haves: reliability, cost-per-run visibility, monitoring.
Nice-to-haves: fleet-level health dashboards.
ICP confidence: Medium — Head of FDE (senior eng leader, customer-facing) at agent-native company, in-band size.
LinkedIn: https://www.linkedin.com/in/joaomdmoura/. Signal: 4 (ICP building/shipping agents; vocal author/speaker on multi-agent reliability & cost). Source: https://tracxn.com/d/companies/crewai/ ; https://shomik.substack.com/p/the-future-of-ai-agents-joao-moura ; https://blog.crewai.com/author/joao/. Company size: ~74 employees (May 2026; earlier reports ~29 — fast-growing; $18M raised, Insight Partners/Boldstart). Pain points: multi-agent orchestration reliability; unbounded tool loops causing runaway token/API costs; agents that demo well but turn non-deterministic and expensive under real load. Challenges: making collaborative multi-agent systems reliable and cost-bounded in production across fintech compliance, healthcare intake, cloud support triage, e-commerce returns. Must-haves: cost ceilings/budgets+alerts, retries, tight observability for multi-agent runs. Nice-to-haves: cross-enterprise agent governance. ICP confidence: High (technical founder/CEO of ~74-employee agent-platform company with agents in production and public focus on cost/reliability).
João Freitas — Chief AI Officer (previously VP of Engineering, AI & Automation), PagerDuty
Signal: 4 (ICP writing/speaking publicly about shipping agents in production) + 3 (company is an Arize + LangSmith user for agent observability).
Source: https://leaddev.com/ai/production-ai-agents-the-gap-between-promise-and-reality (LDX3 London 2026 talk "Production AI agents: The gap between promise and reality", slides at https://leaddev.com/wp-content/uploads/2026/06/Production-AI-agents_-The-gap-between-promise-and-reality-Joao-Freitas.pdf); title confirmed https://www.pagerduty.com/leadership/ and https://www.techjournal.uk/p/pagerdutys-ai-chief-says-outage-response (10 Jul 2026); company agent-observability stack at https://www.pagerduty.com/eng/pagerduty-arize-building-end-to-end-observability-for-ai-agents-in-production/
Company size: 1,155 employees as of 31 Jan 2026 (S&P Global / FY2026 10-K via stockanalysis.com/stocks/pd/employees). Headcount declining YoY (-7%).
Pain points: Non-determinism in production agents; error propagation across agent-to-agent chains — "If you have an agent that is talking to another agent and this agent makes a mistake, then it propagates to the other agent." Reliability/always-on guarantees: "In order to have a reliable system, you want the system to operate and be always on. If you're developing something yourself, you need to maintain it and guarantee the reliability." Talk abstract names reliability, observability and evaluation as what "separate successful deployments from failures."
Challenges: PagerDuty ships three named production agents (Insights Agent, Shift Agent, SRE Agent) into customer incident-response workflows where a wrong action is high-blast-radius; they run golden-dataset regression tests, LLM-as-judge, relevance/groundedness/tool-selection evals and page on-call engineers via a custom "Agent Output Incident" type. Doing this while headcount shrinks 7% YoY.
Must-haves: Reliability and evaluation guarantees for multi-agent chains; observability that catches cascading failure between agents, not just single-call traces; deterministic workflow fallbacks around non-deterministic agents.
Nice-to-haves: Cost/token spend visibility — notably absent from his public commentary; verification pass found NO quote from him on token economics. Cost is likely a latent rather than stated pain.
ICP confidence: High — Director+ (C-level, owns AI engineering), company 1,155 employees, three agents shipped in production, and a dated self-authored public talk naming exactly the reliability/observability/evaluation problem. Note: promoted from VP Eng AI & Automation to Chief AI Officer between Jun and Aug 2026, so scraped data may carry the stale title.
Signal: 4 (ICP writing/speaking about multi-agent production reliability, control & cost). Source: https://pod.wave.co/podcast/venture-with-grace/joao-moura-crewai-ceo-on-scaling-multi-agent-ai-systems + https://www.crewai.com/blog/the-state-of-agentic-ai-in-2026 + https://www.insightpartners.com/ideas/crewai-scaleup-ai-story/. Company size: ~50-120 (est.; Insight Partners-backed 'scaleup', Series A). Pain points: agents fail in production not from lack of autonomy but 'autonomy without structure is impossible to trust'; enterprises need to go prototype->production with governance + reliability; operational cost reduction. Challenges: making multi-agent workflows trustworthy/controllable at scale; security & governance ranked highest enterprise concern (34%), integration (30%), reliability/performance (24%). Must-haves: structure/governance around autonomy, reliability at scale, prototype-to-production path. Nice-to-haves: workflow automation breadth, cost/time savings measurement. ICP confidence: Medium (technical co-founder/CEO at agent-orchestration scaleup; strong public reliability+control voice; company size at lower bound of ICP — flagged as estimate). Note: personal LinkedIn URL not confirmed in search; profile_url points to authoritative author page.
Signal: 4 (ICP technical leader publicly describing shipping agents in production). Source: https://www.computerweekly.com/blog/CW-Developer-Network/Astronomer-Otto-A-data-engineering-agent-built-for-Apache-Airflow (May 2026 interview on Otto, Astronomer's production data-engineering agent). Company size: ~322-386 (Unify ~322 2026; PitchBook 386). Stage: Series C ($213M, Insight Partners). Pain points: context drift / "memory rot" — asked how he prevents accumulated context drifting into outdated or conflicting states, says Otto memories are stored in Git deliberately, "It's not a black box accumulating silently in the background"; conflict-review and curation memory API still "coming soon". Model/provider reliability — he hand-rolled an LLM Gateway: "we whitelist particular models... We also use multiple providers for each model, so if one provider happens to go down, there's a fallback, making it more durable." Governance had to be made "consistent across surfaces by design" (CLI, IDE, API). Challenges: running an agent across 25+ warehouse connectors and hundreds of enterprise Airflow fleets without a separate data plane; keeping a compatibility knowledge base current as Airflow 3.x providers churn; per-customer memory curation at scale. Must-haves: model-agnostic routing with provider-level failover; permission inheritance (agent sees only what the user sees); transparent versionable agent memory; org-level kill switch. Nice-to-haves: automated memory conflict detection; MCP exposure of proprietary context; per-model quality benchmarking before whitelisting. ICP confidence: High — CTO of a Series C data platform with a shipped production agent who has ALREADY built a gateway/routing/fallback layer in-house, i.e. exactly the adjacent buy. NOTE: cost-blowout pain not stated first-person; reliability + context pain is verbatim.
Signal: 4 (ICP publicly describing a multi-agent orchestration layer in production). Source: https://www.helpnetsecurity.com/2026/06/09/filigran-xtm-one/ and https://www.lastwatchdog.com/my-take-black-hat-2026-part-3-agentic-ai-can-do-the-work-but-somebody-has-to-prove-it/ (interviewed in person at Black Hat USA 2026). Company size: ~200 stated by the company Oct 2025 ("we've grown to almost 200 teammates"); Tracxn shows 232 as of 31 Mar 2026. Stage: Series C — $58M Oct 2025 led by Eurazeo with Insight, Accel, Deutsche Telekom; >$100M total. Products: OpenCTI / OpenAEV / XTM One (Paris, France). Pain points (read): "The volume of CVEs, threat actors, and attack campaigns has reached a scale no human team can process manually." Positions XTM One as "not AI as a feature. It is AI as the operating system for threat management" — an orchestration layer "where agents coordinate across products, not just assist within them", i.e. an explicit 1-agent-to-many-coordinated-agents architecture problem. Ships BYOLLM + on-prem, implying model-cost and model-control pressure. Challenges: orchestrating prepackaged agents across multiple products and customer-supplied LLMs, on-prem and in cloud, with no control over which model a customer plugs in. Must-haves: cross-product agent orchestration observability; model-agnostic (BYOLLM) instrumentation; on-prem/self-hosted deployment of any control layer. Nice-to-haves: per-agent and per-model cost comparison to guide customer BYOLLM choices; reliability SLAs on coordinated multi-agent runs. ICP confidence: High — exact title, verified headcount/stage, agents shipping in production as of Jun 2026, and he is personally the public voice of the orchestration layer. NOTE: per-run cost visibility and token waste are inferred; he never mentions cost or tokens in the sources read.
Signal: 1 (ICP speaking/publishing on agent token spend and ROI). Found via web research; title VERIFIED live on LinkedIn people search 2026-08-31 ("Deputy CTO at DX", United States, 12K followers).
Source: https://ai.engineer/worldsfair/schedule (AIEWF 2026 talks "AI-Assisted Engineering: 5 Trends We're Seeing From 500+ Organizations" and "The state of AI in software development: Insights across 400+ organizations") ; https://www.infoq.com/presentations/ai-assisted-engineering/ ; https://rootly.com/humans-of-reliability/justin-reock
Company size: ~169 (LeadIQ). NOTE: one source returned ~401 which is likely the unrelated "DX Group" logistics company — headcount is an estimate with conflicting sources, but both candidate figures sit inside the 50-2,000 band.
Pain points: ROI of agent token spend. His July 2026 research framing is "adoption is above 90% — is ROI positive given token spend?", citing AI spend up 28x in one year in tech while the innovation ratio stayed roughly flat (57% to 58%).
Challenges: Getting engineering leadership to own measurement and guardrails rather than treating adoption as the win condition; he sees this across 400-500+ organizations, so he is a strong pattern-level voice.
Must-haves: Defensible ROI measurement for agent/AI spend; guardrails owned by leadership.
Nice-to-haves: Benchmarks against peer organizations.
ICP confidence: Medium-High — Deputy CTO, in band, and exactly the right pain vocabulary with unusually broad cross-org visibility. Flagged: DX itself sells engineering-intelligence and AI-spend measurement, so treat as partially competitive/adjacent — likely better as a design partner or channel voice than a straight buyer.
Signal: 4 (ICP technical decision-maker at a company shipping a fleet of agents in production). Source: https://www.notablehealth.com/blog/notable-raises-series-b-to-expand-intelligent-automation-in-healthcare ; https://tracxn.com/d/companies/notable . Company size: ~258 employees (Tracxn, Mar 2026); Series B $100M led by ICONIQ Growth (Greylock, F-Prime, Oak HC/FT), ~$116M total; founded 2017 (Justin White, Pranay Kapadia, Adam Ting). Pain points (inferred from public product/positioning; no first-party quotes captured): runs a collection of agents (Registration/Intake, Scheduling & Referrals, Authorizations, Care Gap Closure, HCC Chart Review, billing compliance, patient follow-up) automating 1M+ workflows/day across 10,000+ sites of care — reliability + auditability are critical because errors carry clinical, billing and regulatory consequence; scaling many distinct agent types. Challenges: governance/auditability of autonomous actions in a HIPAA/payer-regulated environment; consistent reliability at very high volume. Must-haves: production reliability, action-level auditability/governance, visibility into agent behavior per run. Nice-to-haves: cost-per-workflow visibility, agents that improve over time. ICP confidence: High (technical decision-maker, agents at scale in production, regulated vertical, Series B, ~258 emp).
Kai Bakker — Founder & CTO, DataSnipper
Signal: 4 (+3 competitor-adjacent audit/finance automation). Source: https://www.datasnipper.com/resources/datasnipper-and-microsoft-launch-ai-agents-for-audit-and-finance ; https://siliconcanals.com/datasnipper-teams-up-with-microsoft/. Company size: ~130 employees; Series B EUR92M (Index Ventures), $1B valuation (unicorn); Amsterdam; 600k+ audit/finance users incl all Big Four (Deloitte, EY, KPMG, PwC). Pain points (INFERRED): launched AI Agents (Disclosure Agents, Excel Agents, Oct 2025 w/ Microsoft) into regulated audit — accuracy/trust required; scaling agents across Big Four workflows; token/context cost of document-heavy agents. Challenges: reliability + auditability of agents in regulated audit/finance; multi-agent orchestration over large workbooks; cost control at 600k-user scale. Must-haves: reliable, auditable agent outputs; per-run cost + behavior visibility. Nice-to-haves: model/routing cost optimization. ICP confidence: High (Founder & CTO; Series B/unicorn; ~130 emp in-band; agents shipping with Microsoft).
Kaja Bargiel — Head of AI & Data, Abridge
Signal: 4 (ICP-title AI leader at an agent-active company; surfaced via LinkedIn/web leadership search this run). Source: https://www.linkedin.com/in/kaja-bargiel and Abridge leadership. Company size: ~300-400 (Abridge, ambient clinical AI documentation + clinical agents in healthcare). STAGE CAVEAT: late-stage (Series D/E, multi-billion valuation as of 2026) — past the strict Series A-C guideline, but firmly in the 50-2000-emp band and clearly agent-active; retained with stage noted (same treatment as PolyAI/Cresta/Seismic in prior runs). Pain points [INFERRED from role+company, not a verbatim quote]: reliability/accuracy and auditability of clinical AI in a regulated setting; cost of long ambient sessions; determinism/consistency of outputs. Challenges: making agents trustworthy enough to reduce human review in clinical workflows; controlling per-encounter cost at scale. Must-haves: reliability + auditability + per-run cost visibility. Nice-to-haves: replay/simulation for regression testing. ICP confidence: Medium (right title + right size + agent-active; late stage is the only knock).
Signal: 4 (ICP technical leader at an agent-native company absent from the brain). Source: gap-company research after brain dedup (EvenUp not in the current list); title/URL verified on LinkedIn ('Kallol Das — CTO at EvenUp', San Francisco). EvenUp is an AI-native legal (personal-injury) unicorn shipping AI agents for case workup / demand generation / pre-litigation workflows in production; ~853 employees (2026), Engineering ~31.6% of staff; grew from 216 (2023). Kallol Das is CTO (promoted Jul 2025). Pain points: reliability/accuracy of legal agents on high-stakes output; cost per run at scale; observability + evals; scaling agents across a large customer base. ICP confidence: High (CTO; agent-native vertical AI; headcount in range). Note: pains inferred from role/company profile, not a personal quote.
Kamer Ali Y. — Head of Agentic AI / Distinguished Scientist, aiXplain
Signal: 1 (ICP writing about agents). Source: https://www.linkedin.com/in/kyuksel/
Company size: ~50-150 (aiXplain, agentic-AI platform / AgenticOS; Series A; Munich + US)
Pain points (EXPRESSED in his posts): agent evaluation & reliability at scale; black-box model routing & ensembling (cost/quality tradeoff); simulation-based evaluation and evolutionary self-improvement of agents; multi-agent enterprise systems; robustness to distributional/regime shift.
Challenges: identifying and solving "fundamental challenges in Agentic AI" - routing, eval, self-improvement - across large-enterprise & government deployments.
Must-haves: robust agent evaluation, model routing across providers, agent observability.
Nice-to-haves: interpretable/auditable agent behavior.
ICP confidence: High (Head of Agentic AI at agent-native platform company, in range, genuine agent-engineering signal). New company for Brain.
Signal: 4 (ICP shipping agents; posts on production GenAI: AI agents, RAG, realtime voice, evals). Source: https://www.linkedin.com/in/kangkanboro/ — BorderPlus named WEF Technology Pioneer 2026; shipping Nurse AI Companion (conversational agent), CarePilot voice-to-documentation, clinical knowledge Q&A. Company size: ~80-200 (BorderPlus, AI-native healthcare workforce-mobility, offices Bangalore and Mumbai, ops-heavy). Pain points: production voice/conversational agents (code-switching ASR, cocktail-party problem, off-the-shelf Whisper limits), evals, reliability in real clinical settings. Challenges: shipping fast, model architecture beyond APIs, guardrails/compliance. Must-haves: reliable production voice agents, evals, guardrails. Nice-to-haves: agent unit-economics/cost visibility. ICP confidence: Medium (agent-native plus Director-level role strong; exact headcount and 5+ agent count not fully confirmed).
Kanjun Qiu — Co-Founder & CEO, Imbue
Signal: 4 (ICP senior technical leader at a company actively shipping AI agents; found via LinkedIn people-search, not an observed post). Source: https://www.linkedin.com/in/kanjun/ | Company size: ~40-70 (agent/reasoning-systems startup) — SMALL END, near lower ICP bound. Pain points (INFERRED from role + company product focus — no observed quote): scaling agents reliably in production; controlling per-run LLM/agent cost as volume grows; visibility into what agents cost and how they behave across customers. Challenges: keeping agent accuracy/guardrails consistent at scale; token/context efficiency; cost blowout as they move from pilots to broad production. Must-haves: production reliability + observability into agent cost and behavior. Nice-to-haves: per-run cost attribution, context/token optimization, spend guardrails. ICP confidence: Medium (Technical co-founder building AI agents/reasoning systems; company already curated as ICP in brain, but headcount is near the 50-employee floor — flagged).
Kapil Garg — Technical Director, DevRev
Signal: 4 (ICP Director-level engineering leader at validated target company DevRev; sourced via LinkedIn People search, run 2026-08-01). Source: https://www.linkedin.com/in/kapilgarg12/ . Company size: ~700-1,000 (DevRev — AI-native support/CRM platform shipping AI agents via AgentOS; in-range). Pain points (role/company-contextual): reliability of support agents in production; cost-per-resolution / per-run economics of LLM-driven agents; observability into agent behavior and spend. Challenges: scaling agents across many customer tenants reliably; controlling token/compute cost at support volume. Must-haves: per-agent cost & reliability visibility; guardrails. Nice-to-haves: model routing, automated eval/regression. ICP confidence: Medium-High (Technical Director at 50-2,000-emp AI-native agent-shipping company).
Signal: 3/4 (technical founder; voice-agent platform + competitor-adjacent infra). Source: https://fortune.com/2025/03/11/exclusive-cartesia-voice-ai-startup-raises-64-million-series-a/ ; https://www.cartesia.ai/agents ; https://tracxn.com/d/companies/cartesia/. Company size: ~116–122 employees (May 2026); Series A $64M, ~$191M total raised (Kleiner Perkins, Index, Lightspeed). Cartesia (out of Stanford AI Lab, SSM/Mamba lineage; co-founders Karan Goel, Albert Gu, Arjun Desai, Brandon Yang) ships Sonic (TTS), Ink (STT) and Line — a platform for building and shipping production enterprise voice agents; heavy focus on latency (~40–45ms) and linear-scaling compute cost. Pain points (product-stated): real-time production reliability + latency for voice agents; compute/cost efficiency at scale (explicit SSM cost angle); helping customers ship reliable voice agents. Challenges: production reliability + cost/latency as voice-agent volume scales. Must-haves: low-latency reliability, cost efficiency. Nice-to-haves: observability/cost-per-run for deployed voice agents. ICP confidence: Medium (named technical founder-CEO; ~122 emp Series A in band; ships a voice-agent platform. Caveat: infra/model-adjacent — partly a competitor to an agent operating layer; validate positioning before outreach).
Kareem Amin — Co-founder & CEO (technical), Clay
Signal: 4 (ICP technical leader at AI-native company building agents into product). Source: https://www.linkedin.com/in/kareemamin , https://getlatka.com/companies/clay , Tracxn Clay profile | Company size: ~1,000–1,200 employees (2026; Latka ~1K, Tracxn ~1,167 — within 50–2,000). GTM data/enrichment platform; ships AI research agents ("Claygent") that autonomously research accounts/contacts and enrich data at high volume across customer workflows. Amin is technical co-founder & CEO (engineering background; prior founder). Pain points: running research/enrichment agents at massive volume where token/inference cost compounds fast; keeping agent outputs accurate/reliable at scale; visibility into what each agent run costs per customer/credit. Challenges: cost blowout as agent usage scales across a large customer base; reliability of autonomous research output; per-run cost attribution to pricing/credits. Must-haves: cost-per-run visibility, reliability at scale. Nice-to-haves: observability, behavior control. ICP confidence: Medium (technical CEO at a 50–2,000-employee AI-native company embedding agents in product; primary product is data/enrichment rather than pure agent platform, so agent-ops centrality is slightly lower).
Signal: 4 (CTO at a qualifying company publicly moving to an agentic architecture). LinkedIn verified live 2026-09-02: headline reads exactly "Chief Technology Officer at Seedtag", New York.
Source: https://events.seedtag.com/Seedtag-Innovation-Summit (Seedtag Innovation Summit session listing, "Seedtag 2.0 Agentic AI")
Company size: 751 employees (Tracxn, 2026); Series C (2022, $250M from Advent International). Seedtag = contextual advertising platform, historically ML-heavy, now publicly repositioning around "Seedtag 2.0 Agentic AI" for content and intent analysis.
Pain points: HONEST CAVEAT — the sourced signal is a summit session abstract (marketing copy about a company-wide shift to agentic AI), not a personal pain statement from Goksel. Do not quote him as having complained about cost or reliability.
Challenges (inferred from the transition, NOT quoted): converting a single-purpose, high-throughput ML platform into a multi-agent architecture while ad-serving economics stay razor-thin; agent inference cost per impression/analysis at ad-tech volume is a genuine unit-economics problem.
Must-haves (inferred): agent-level cost and latency visibility across very high-volume workloads; reliability parity with the deterministic ML system being replaced.
Nice-to-haves (inferred): model routing to keep cost-per-analysis flat as agent coverage expands.
ICP confidence: Medium — title, headcount and Series C stage are a clean ICP match and the agentic shift is publicly stated by the company, but the pain evidence is company-level positioning rather than a first-person signal. Best approached with a cost-per-analysis / ad-tech-unit-economics hypothesis rather than a "you said X" reference.
Signal: 1 (ICP writing publicly about agent architecture, observability and cost visibility). Found via web research — LinkedIn/Chrome unavailable this run.
Source: https://medium.com/life-at-moengage/building-ai-agents-at-moengage-architecture-and-key-lessons-c24c05c711fa (6 Jan 2026)
Company size: ~757–900 employees in 2026 (PitchBook ~800; RocketReach 902; Revelio 1,013 Dec 2025 trending to ~757 Jul 2026). Bengaluru + San Francisco. ~79% of workforce in India. Late-stage (Series E) — stage is the only ICP stretch.
Agent evidence: named production agents — campaign/HTML generator, Flow Builder, Segmentation, Support Agent, internal OnCall Agent grounded in runbooks. Merlin AI Custom Agents shipped Jun 2026 with guardrails + activity-log visibility + open MCP server. Acquired Aampe Jun 2026 for per-user decisioning agents. Stack: Google ADK (migrated off LangChain), FastMCP, Arize Phoenix for prompt mgmt + OTEL agent observability, LiteLLM as model gateway for retries/fallbacks/guardrails, AWS OpenSearch RAG, BFF layer streaming agent JSON into UI widgets.
Verbatim (his three stated lessons): check agent ROI before building; decentralization is fast but "leads to chaos quickly"; "Invest in evaluation & observability upfront." Also describes non-determinism, latency and evaluation as "battles we fight daily."
Pain points: "fat agents" that are hard to debug and require full redeployment for a business-logic change; agents hammering internal APIs causing unpredictable upstream traffic spikes; prompts coupled to code so PMs/business users cannot change them without a deploy; NO VISIBILITY into which LLMs teams were using or what they cost.
Challenges: no single eval strategy works — HTML-generator agents need Playwright-based visual/accessibility evaluation while OnCall/Support/Flow Builder agents need entirely different approaches; evaluating in production against real customer inputs not just static suites; MCP tool-poisoning risk requiring a gateway with strict auth; standardization vs. team autonomy across many product teams.
Must-haves: agent observability with OTEL tracing across agent↔LLM and agent↔MCP hops plus token and context-window tracking; a single model gateway with retries, fallbacks and I/O guardrails; prompt management decoupled from code with versioning and non-engineer editing; production (online) evaluation; MCP gateway with enforced auth.
Nice-to-haves: reusable tool/MCP layer shared across product teams; caching and traffic control at the MCP proxy; real-time streaming of stateful agent interactions to the UI; per-team LLM choice while retaining central cost visibility.
ICP confidence: Medium — headcount, geography and production-agent depth are all strong and the technical content is unusually good; title is VP-level (VP Architect) rather than a line VP Eng, and MoEngage is Series E rather than A–C. HIGH-PRIORITY for outreach: his stated #1 gap is central cost/model visibility.
Signal: 1 (senior technical leader at a startup actively building production AI agents). Source: https://www.linkedin.com/in/karthikjayanthi/ (surfaced via WebSearch + LinkedIn; colleague Kangkan Boro's profile confirms BorderPlus is "Production GenAI: AI agents, RAG, realtime voice, evals"). Company size: ~92 employees (BorderPlus; $7M seed led by Owl Ventures, 2026). Ships agents: YES — building production GenAI incl. realtime voice AI agents for healthcare talent-mobility workflows. Pain points (inferred from role at an early team shipping voice agents): reliability + cost of realtime voice agents in production, scaling from first agents to more. Challenges: small team operating production LLM/voice agents cost-effectively. Must-haves: reliable, cost-controlled production agents. Nice-to-haves: eval + observability tooling. ICP confidence: Medium (VP Eng title + company confirmed 92 emp building agents; individual signal inferred).
Signal: 1 (ICP authoring directly about agent token cost / control). Source: LinkedIn post 1w before 2026-09-03, "Tokens Are the New Headcount Nobody's Budgeting For" (48 reactions, 6 reposts) — https://www.linkedin.com/in/karthikkannan001/recent-activity/all/ ; corroborated by his own 2mo post on Blueprints learnings. Company size: 115 (already verified in this brain via Crustdata/Tracxn on the Deb Banerjee record, 30 Jun 2026). Stage: Series C ($45M led by Evolution Equity Partners, $85M total). Agents in production: Blueprints orchestrator agent + discrete SOC agents (federated search, data management, detection, triage, investigation), GA'd ~1mo ago and now deployed across the customer base. Pain points (his words): "are we really just spending money for low value tasks by using LLMs, and is it causing new spend areas that were previously un-budgeted"; "you're going to be wasting a lot of money in token burn for limited return in value"; "off-the-shelf products as well as home-grown tooling over-interrogating LLMs leading to not only token overage ($$$) but also rendering highly inaccurate or at best, vague, results." Challenges: pricing/architecting agentic SOC workflows so frontier-model calls are the exception not the default; proving ROI per automated workflow to enterprise buyers who now ask about token line items; keeping accuracy at 98%+ while cutting LLM calls. Must-haves: per-workflow / per-agent token attribution and budgeting; ability to show cost-per-automated-outcome to customers; controls that stop over-interrogation of frontier models. Nice-to-haves: routing between local/small models and frontier models by task; cost telemetry exposed to their own customers as part of the platform. ICP confidence: High — exact-fit pain expressed verbatim and unprompted, verified company size and stage, multiple agents shipped to production. NOTE: Deb Banerjee (Anvilogic CTO) is already in the People Library — this is a second, distinct buyer-side contact at the same account, and the cost angle is stronger here than on Deb's record.
Karthik Paulramachandran — VP of Engineering, DataBahn
Signal: 4 (ICP in the exact buyer seat at a company shipping agents; no public voice yet — unclaimed conversation).
Source: title verified directly on https://www.databahn.ai/leadership. Company context: https://www.databahn.ai/blog/the-pipeline-was-just-the-beginning (30 Jul 2026) and Series B PR https://www.prnewswire.com/news-releases/databahn-launches-federated-search-and-orchestration-...-302842675.html
Company size: ~112 employees as of 30 Jun 2026 per Tracxn (search snippet; Tracxn page JS-gated so medium confidence). Careers page publishes no number. Series B, $40M Jul 2026 led by Insight Partners. Estimate 100–130 — inside band.
Pain points: No public pain quote from him personally. Company-level pain stated by CEO Nanda Santhana (verbatim): "Organizations consume more storage, more compute, and more AI tokens while delaying the moment when intelligence can be extracted." DataBahn's own commercial pitch is explicitly "lower AI token cost."
Challenges: Runs engineering for an agentic data control plane where token consumption is the headline cost line; per his own DataBahn bio he "personally spearheaded the company's transition to an Agentic AI Platform" at CommerceIQ — direct prior experience hitting the 1→many agent scaling wall.
Must-haves: Token/compute cost attribution across agent pipelines; visibility into which agents and stages burn spend.
Nice-to-haves: Cost controls that survive federated search/orchestration workloads.
ICP confidence: Medium-High — exact ICP title verified on a company-owned page, headcount inside band, Series B, company actively ships agents. Downgraded from High because he has zero public commentary, so the pain is inferred from company positioning and prior role rather than stated.
Watch-out: DataBahn markets token-cost reduction itself — partial competitive overlap. Also do NOT confuse with databahn.com (unrelated sales-intelligence firm, CTO Bill Gosse).
Karthik Rajan — Co-founder & CTO, Suki AI
Signal: 4 (ICP technical leader / CTO at a qualifying agent company; found via web verification of Suki leadership). Source: https://theorg.com/org/suki ; https://www.suki.ai/about/. Company size: ~413 employees (June 2026); healthcare AI — ambient clinical documentation + AI Assistant with agentic features (e.g., ambient orders staging); ~$168M raised. Co-founded Suki (2017) with Punit Soni; confirmed CTO, co-founder & board member in 2026. Pain points (role-inferred): production reliability and clinical safety of agents at scale across many clinicians; per-encounter LLM/inference cost as usage scales; latency requirements in clinical workflows. Challenges: scaling agentic features reliably in a regulated clinical setting; controlling inference spend per clinician/encounter. Must-haves: production reliability + observability; cost-per-run visibility. Nice-to-haves: inference cost optimization / model routing. ICP confidence: High (technical co-founder/CTO at an agent-native 50-2,000-emp company). NOTE: pain points inferred from role/company context, not a verbatim quote.
Signal: 4 (senior technical leader at a company confirmed to be actively building AI agents). Source: LinkedIn profile above — found via people search scoped to Cresta ("Cresta AI director of engineering"). Company: Cresta (cresta.ai), unified platform for human + AI agents for CX / contact centers; raised $125M Series D. Company size: 501-1,000 employees (confirmed via LinkedIn company page, Sunnyvale). Pain points (inferred from role + company, NOT a verbatim quote): deploying/customizing production CX agents for enterprise customers (healthcare & life sciences focus); agent reliability and cost efficiency at contact-center scale; demonstrating ROI / cost-per-interaction. Challenges: field/deployment-side reliability, latency and cost of live voice+chat agents. Must-haves: reliability at scale, cost-per-run visibility, guardrails. Nice-to-haves: observability across deployed agents. ICP confidence: Medium (Field CTO is a customer-facing/solutions CTO role rather than core internal AI/eng leadership, but title-level and company clearly ICP).
Signal: 4 (ICP technical co-founder shipping agents in production). Source: https://techcrunch.com/2026/04/15/hightouch-reaches-100m-arr-fueled-by-marketing-tools-powered-by-ai/ ; https://hightouch.com/blog/hightouch-funding-series-d ; https://www.ycombinator.com/companies/hightouch. Company size: ~380 employees (Apr 2026); $100M ARR; Series D $2.75B valuation (prior Series C Feb 2025 at $1.2B). Technical co-founder (YC alum). Hightouch Agents platform runs always-on marketing agents (audience research, creative generation, cross-channel campaign execution) — 5+ agents in production. Pain points (inferred): reliably orchestrating many autonomous marketing agents; controlling agent behavior/brand safety at scale; agent run cost as adoption scales. Challenges: production reliability + governance of agents acting in live systems. Must-haves: agent control/observability, production reliability. Nice-to-haves: per-run cost visibility. ICP confidence: High (technical co-founder/co-CEO at a 380-person agentic-AI company shipping agents; second decision-maker at same account as Tejas Manohar).
Signal: 1 (ICP quoted publicly on token/context waste). Found via web research; title VERIFIED live on LinkedIn people search 2026-08-31 ("CTO - Investment Products @ INDmoney | Ex Goibibo, MakeMyTrip, Intel", Bengaluru).
Source: https://inc42.com/features/the-enterprise-fight-against-runaway-ai-costs/
Company size: ~543-696 (Revelio Labs Dec 2025 = 696; mid-2026 aggregator = 543). In 50-2,000 band.
Pain points: Context/token waste as the dominant cost driver. On record: much of an AI agent's cost comes from the amount of information it sends to the model — "A task that takes 40 steps can therefore consume millions of tokens, even if the original request contained only a few thousand." Estimates caching alone cuts effective cost by ~80%.
Challenges: Multi-step agent runs where context is re-sent on every hop; cost scales with step count rather than with the size of the original request.
Must-haves: Context/token accounting per agent run; caching and context-reuse controls.
Nice-to-haves: Prediction of run cost before execution; per-step token attribution.
ICP confidence: High on pain (cleanest token-waste articulation found this run), Medium on company fit — INDmoney is a late-stage fintech (~$143-158M raised), not Series A-C SaaS, though headcount is squarely in band and agents are in production.
Kaushik Chandrashekar — VP of Engineering, interface.ai
Signal: Bucket 4 (ICP engineering leader at an agent-native company shipping agents in production).
Source: https://theorg.com/org/interface-ai/org-chart/kaushik-chandrashekar ; https://natlawreview.com/press-releases/interfaceai-unveils-industry-first-agentic-bankgpt-platform-moves-cx
Company size: ~204 emp; agentic BankGPT banking-agent platform. VP Eng focused on LLM/GenAI since Apr 2025 (prev Dir Eng @ Mindtickle, Navi; EM @ Hotstar).
Pain points (INFERRED, not verbatim): engineering reliability & cost of banking agents in production; scaling the agent platform; LLM token/cost control per interaction.
Challenges: reliability at regulated-banking scale; cost-per-run visibility; multi-agent orchestration.
Must-haves: production reliability, cost/token visibility, guardrails.
Nice-to-haves: model comparison, replay/simulation.
ICP confidence: High — VP of Engineering, ~204 emp, agent-native banking (net-new; company not previously in brain).
Signal: 4 (web-research proxy — LinkedIn/Chrome unavailable this run; found via Mar 2026 launch/funding announcement + company profile). Source: https://www.businesswire.com/news/home/20260311880123/en/ ; title/headcount confirmed via https://tracxn.com/d/companies/axiamatic and ZoomInfo. Company size: ~80 employees (Tracxn/TheOrg, Jun 2026); Series B, $54M from Greylock + Bessemer; founded 2021. Prior: co-founded Skyhigh Networks (CASB category, acq. by McAfee) with CEO Rajiv Gupta.
Pain points (from company positioning): autonomous agents must monitor a live "digital twin" of enterprise programs 24/7 across 250+ systems and reliably detect risk, translation loss, and workstream drift — i.e. agent reliability + context loss at scale.
Challenges: keeping many production agents accurate as they ingest signals from hundreds of enterprise systems; avoiding drift/hallucination; giving humans visibility into what agents flag and why.
Must-haves: production-grade agent reliability, observability into agent decisions, control across a large agent fleet.
Nice-to-haves: cost visibility per agent/run as fleet scales.
ICP confidence: High — clear CTO title, technical serial founder, confirmed 50-2,000 headcount, actively shipping enterprise agents.
Kaushik Vatsa — Director / VP of AI Engineering, Mantra (Mikshi AI)
Signal: 1 (ICP writing directly about agent cost, reliability and control). Source: https://www.linkedin.com/in/kaushik-vatsa-4582536/ (multiple public posts, past ~3 months). Company size: ~114 employees (Mantra / Mantra Software Pvt Ltd, Bengaluru; Mikshi AI is its production agentic video-intelligence VLM platform + internal HRMS RAG). Verified 50-2,000 band and actively shipping agents in production ("shipping current AI agent (Mikshi) to production"). Pain points (his own words/paraphrase): multi-agent architectures where "every hop is a new context window, a new failure mode, a new bill"; bounded loops needed so agents don't keep "spinning and burning cost"; production false positives (flagged an event that never happened); detailed per-scale vector-DB cost curves (pgvector vs Milvus). Challenges: making agentic systems debuggable and deterministic ("you own the control flow" vs model-routed non-deterministic multi-agent); latency (4.55s p50 vs 11.82s on bigger model); grounding/hallucination control at 3AM production workflows. Must-haves: grounding guards, confidence gates on verifiable signals (grounding + logprob), bounded loops with convergence checks, evals-as-foundation (golden dataset of 170 videos / 1000 QA pairs before tuning). Nice-to-haves: cheaper reversible infra bets, model-commodity-proof moats. ICP confidence: High (senior technical AI leader shipping agents in production at an in-range company, with explicit, repeated agent-cost + reliability + control pain — a textbook Signal-1 author). profile_url may not persist via API, so LinkedIn URL is captured here in Source.
Signal: 4 (ICP technical co-founder/CTO at AI-native super-agent company). Source: https://tracxn.com/d/companies/genspark/ ; https://www.linkedin.com/in/kay-zhu-522b275 ; https://ainativegtm.substack.com/p/genspark-from-zero-to-36m-arr-45. Company size: ~143 employees (Mar 2026); AI-native, unicorn valuation, ~$36M+ ARR within 45 days of launch. Founders: Eric Jing (CEO, ex-Bing), Kay Zhu (CTO, ex-Google deep neural ranking), Wen Sang (COO). Actively shipping agents: Genspark 'Super Agent' — multi-agent workspace executing autonomous multi-step tasks at high volume. Pain points (inferred from product + domain, not verbatim quotes): per-task cost of running many model calls across a general super-agent; latency and reliability at rapid-growth scale; controlling autonomous multi-agent behavior. Challenges: keeping a general-purpose agent reliable + cost-efficient as usage explodes. Must-haves: per-run cost control, reliability, orchestration/observability across many agents. Nice-to-haves: compounding intelligence. ICP confidence: High (technical co-founder & CTO of AI-native agent company, ~143 emp, agents in production at scale).
Keita Morikawa — Co-Head of Agent Development (Japan), Sierra
Signal: 4 (senior leader owning agent development at an agent-native company). Source: LinkedIn people search (https://www.linkedin.com/search/results/people/?keywords=Sierra%20head%20of%20engineering%20agents). Company size: ~855 employees (Sierra; agent-native CX). Pain points (INFERRED): shipping reliable enterprise agents in a new region; scaling agent development while keeping quality/cost in check. Challenges: reliability/quality in production agent deployments; localization at scale. Must-haves: reliable, cost-visible agent runs. Nice-to-haves: evals/observability tooling. ICP confidence: Medium (senior, owns regional agent dev; also a venture partner so focus is split — noted; stage past Series C).
Keith Morrison — VP of AI Data Platform Engineering, Cohere Health
Signal: 4 (VP-level AI engineering leader at company shipping agentic AI in production). Source: https://www.coherehealth.com/news/cohere-unify-agentic-ai-redesigning-health-plan-operations ; Cohere Health careers/press (VP, AI Engineering & Agent Platforms role) ; searches 2026. Company size estimate: ~400-800 employees (healthcare prior-authorization / clinical-intelligence SaaS; ~$106M+ raised through Series C). Product: 'Cohere Unify' — agentic AI redesigning health-plan operations (prior auth, utilization management); company is actively building an Agent Platforms function (open VP, AI Engineering & Agent Platforms role reporting to Chief Digital & Technology Officer). Role: Keith Morrison, VP of AI Data Platform Engineering. Pain points (inferred from role + domain, NOT verbatim quotes): reliability and auditability of agents acting on health-plan/clinical data; data-platform + agent infra to support scaling agents; cost/observability at high transaction volume in a regulated setting. Challenges: building reliable, governable agent platforms on regulated healthcare data. Must-haves: reliability, governance/auditability, data-platform robustness. Nice-to-haves: per-run cost visibility/observability. ICP confidence: Medium (named VP of AI (data platform) engineering; company in band actively building agent platforms; caveat — role is data-platform-focused rather than pure agent ownership, and company-size estimate not precisely confirmed). LinkedIn URL not captured — do not fabricate.
Ken Koch — Director of Engineering, EliseAI
Signal: 4 (ICP technical leader at validated target company EliseAI; sourced via LinkedIn People search, run 2026-08-01). Source: https://www.linkedin.com/in/kenneth-koch-iv/ . Company size: ~250-500 (EliseAI — Series D, property-management AI agents; validated in-range). Pain points (role/company-contextual, not from an individual post this run): production reliability of housing AI agents; token/compute cost per conversation at scale; visibility into what agents cost per run. Challenges: scaling multiple production agents while controlling spend and quality. Must-haves: per-run cost observability, reliability/eval tooling. Nice-to-haves: automated regression on agent changes. ICP confidence: High (Director of Engineering, agent-native, in-range).
Signal: LinkedIn People search (agent-native company + agents), closest to Bucket 4. Source: https://www.linkedin.com/in/ketanpatel33/ (found via people search 'Gupshup VP engineering AI agents'). Company size: ~1,000-1,500 employees (Gupshup, conversational-AI/CPaaS leader, in-band). Gupshup is agent-native: Conversation Cloud + Autonomous AI Agents / ACE LLM for enterprise CX. Pain points: deploying & scaling autonomous AI agents reliably for enterprise customers ('Leading AI-First Transformation'), production reliability/cost of customer-facing agents. Challenges: building future-ready teams to ship agent deployments at enterprise scale; reliability/cost of agents in production. Must-haves: reliability, deployability at scale. Nice-to-haves: per-run cost visibility, observability. ICP confidence: Medium. Note: role is 'VP - Customer Engineering' (VP-level, engineering-adjacent/solutions function at an agent-native in-band company).
Kevin Gao — Head of Infrastructure & Security, Sierra
Signal: 4 (senior technical leader at an agent-native company). Source: LinkedIn people search (https://www.linkedin.com/search/results/people/?keywords=Sierra%20head%20of%20engineering%20agents). Company size: ~855 employees (Bret Taylor's AI CX-agent company; $15.8B val 2026 — late-stage but firmly in 50-2,000 band). Pain points (INFERRED): scaling agent infrastructure reliably and securely; compute/token spend visibility across large production agent fleets; governance of in-path model calls. Challenges: infra reliability, security/governance, cost control at scale. Must-haves: control + observability over agent infra spend and reliability. Nice-to-haves: neutral in-path enforcement/governance. ICP confidence: Medium-High (agent-native, right-size; infra/security head is directly cost- and reliability-relevant; stage past Series C — noted).
Signal: 4 (ICP eng leader at a company actively building AI agents). Source: https://theorg.com/org/ottimate/org-chart/kevin-leduc ; company AI posts https://www.linkedin.com/posts/ottimate_ai-in-ap-touchless-processing-profitability-activity-7358909479305859073-JlMM ; hiring Director of AI Engineering + VP Engineering. Company size: Ottimate (formerly Plate IQ) ~200-400 employees; Series C AP-automation SaaS. Pain points: scaling AI-driven touchless invoice processing reliably; moving from AI features to production agent workflows over ERP data. Challenges: production reliability of AI over complex invoice/ERP data; scaling the AI eng org. Must-haves: reliable agent execution over financial data with audit trails. Nice-to-haves: cost visibility per automated invoice/workflow. ICP confidence: Medium (Director of Engineering = target role; company actively building agents; personal signal is company-level rather than a direct quote).
Signal: 1 (ICP publicly writing/speaking about agent architecture and where LLMs should NOT be trusted) + 4.
Source: https://www.lucanet.com/en/press-releases/lucanet-ai-agents-autonomous-cfo-platform-30-06-2026/ ; https://www.linkedin.com/posts/fintech-finance_reimagining-the-finance-stack-lucanet-cto-activity-7358033346938970112-tIUy ; https://www.linkedin.com/posts/kevinsmi_cfosolutionplatform-ai-lucanetinnovation-activity-7382504963093995522-BELB
LinkedIn: https://www.linkedin.com/in/kevinsmi/
Company size: 900+ employees worldwide (company About page). Berlin; CFO/financial-close/consolidation platform; PE-backed (Hg).
Agent evidence: family of specialized workflow agents launched 30 Jun 2026 across planning, closing, reporting and ESG/tax — shipping into the platform "over the coming months", i.e. actively mid-rollout right now.
Pain points: QUOTED — "AI is probabilistic in nature and so not suitable to perform deterministic calculations, but it is exceptionally good at reasoning, interpreting, and explaining." He has architected an explicit split: deterministic core does the math, AI layer reasons/explains. That is a man who has already been burned by non-determinism and is engineering around it.
Challenges: keeping a probabilistic layer auditable inside statutory financial close and tax — regulated, wrong answers are material; running a multi-agent rollout across four product pillars at once.
Must-haves: deterministic replay / provable boundary between AI reasoning and financial calculation; audit evidence for regulators.
Nice-to-haves: cost visibility per agent as agents multiply across pillars; drift detection between releases.
ICP confidence: HIGH — CTO, ~900 employees, agent fleet actively shipping, and he is already publicly reasoning about the control problem in our vocabulary. Active LinkedIn poster = warm Signal-1 surface.
Kevin Wang — SVP of Engineering, Abnormal AI
Signal: Signal 4 (LinkedIn people search at named agent company; SVP Engineering. Sourced by role/company match)
Source: https://www.linkedin.com/search/results/people/?keywords=%22Abnormal%20AI%22%20head%20of%20engineering
Profile: https://www.linkedin.com/in/kevin-wang-7443584/
Company size: ~1,200-1,800 (Abnormal AI; autonomous AI security agents; Series D)
Pain points: (inferred, not directly observed) Scaling an engineering org shipping many production AI agents; reliability + cost control at fleet scale
Challenges: (inferred) Visibility into what agents cost per run and where they fail
Must-haves: (inferred) Cost/reliability observability and governance across agents
Nice-to-haves: (inferred) Context/token-waste reduction
ICP confidence: High (SVP Engineering; in-band agent-native company)
Kiyo Kunii — Co-Head of Agent Development (Japan); former Co-Founder & CTO, OPERA, Sierra
Signal: 4 (ICP agent-engineering leader at an agent-native company; found via LinkedIn people search "Sierra AI director of engineering agent platform" this run).
Source: https://www.linkedin.com/search/results/people/?keywords=Sierra%20AI%20director%20of%20engineering%20agent%20platform — profile summary: "Co-Head of Agent Development at Sierra for the last 5 months, building agents after co-founding OPERA TECH and serving as CTO." Tokyo, ~3K followers.
Company size: 201–500 employees (LinkedIn company page for Sierra, San Francisco, verified this run — note there is a decoy "SIERRA AI" page in Frisco TX with 2–10 employees; this is NOT that company). Sierra = AI-native conversational agent platform, Series C, agents deployed in production across large enterprise customers.
Pain points: NO public pain post captured this run — do not treat as expressed. Qualification is role + company based.
Challenges (inferred from role scope, flagged as inference): "Agent Development" at Sierra is the function that stands up and operates customer agents; scaling that across many enterprise deployments is exactly where per-agent cost/run visibility and reliability regressions bite.
Must-haves: unverified this run.
Nice-to-haves: unverified this run.
ICP confidence: Medium — Head-level (Co-Head of Agent Development) so at the ICP floor, at a clearly qualifying agent-native company of the right size; downgraded from High because the role is regional/deployment-facing rather than owning central engineering or AI platform budget.
Signal: Bucket 3/4 (competitor/agent space) — LinkedIn people-search proxy. Source: https://www.linkedin.com/in/klauskrogmann (Karlsruhe, Germany). Company size: ~300 (Cognigy, conversational + agentic AI for contact centers / voice + chat agents). NOTE: Cognigy acquired by NiCE in 2025; now operates as a contact-center agentic AI unit inside a >2000-emp parent. Pain points (INFERRED from role): reliability + cost of voice/chat agents in production, scaling multi-agent CX deployments, cost-per-run visibility. Challenges: enterprise-grade reliability + governance for agents. Must-haves: production cost + reliability observability. Nice-to-haves: unified agent analytics. ICP confidence: Medium — VP Engineering at agent-shipping company; parent (NiCE) exceeds 2000 emp so size is borderline. Caveat: title search; pains inferred (no fabricated quote).
Signal: 2 — quoted/featured in an AllTech Magazine LinkedIn post on why enterprise agentic AI stalls at the infrastructure layer (found via content search: "Director of AI" agents production scaling cost). Source: https://www.linkedin.com/search/results/content/?keywords=%22Director%20of%20AI%22%20agents%20production%20scaling%20cost . Company size: ~150-300 (est; Saritasa is a custom software dev firm — a SERVICE PROVIDER building agentic AI for clients, not agent-native SaaS). Pain points: proof-of-concept trap (pilots stall in production); legacy infra can't support agents (real-time data, clean APIs, microservices, elastic compute); AI ROI plateau. Challenges: moving agent pilots into reliable production. Must-haves: production-ready infra for agents. Nice-to-haves: modernization roadmap. ICP confidence: Medium — Director-level, in-range size, but services/agency provider.
Konstantin Zhandov — VP of Engineering — Agentic & MCP Enterprise Platforms (founding team), Workato
Signal: 4 (VP Engineering explicitly self-describing agentic + MCP platform ownership at a company shipping a fleet of production agents).
Source: LinkedIn people search — headline verbatim: "VP of Engineering | Agentic & MCP Enterprise Platforms | Founding Team @ Workato". Company agent activity verified via workato.com/the-connector/introducing-workato-genies/, docs.workato.com/agentic/workato-genies.html, workato.com/agentic/hr.
Company size: 1,517 employees worldwide as of March 2026 (Revelio) — grown 56.5% from 969 in 2023.
Pain points: running a genuine agent fleet, not one agent — Workato Genies are packaged per business function (sales, marketing, IT, HR & recruiting, License Genie), which is exactly the 1→5+ agents scaling wall in the ICP definition; each Genie carries its own skills, knowledge bases and app integrations, multiplying orchestration surface and per-run cost.
Challenges: MCP/tool sprawl across many agents; cost and reliability attribution per Genie rather than per platform; governing autonomous agents that "execute approved changes" against real systems.
Must-haves: per-agent cost and reliability visibility across a multi-agent fleet; consistent control plane spanning agentic + MCP surfaces.
Nice-to-haves: chargeback/showback by business function, since Genies are already packaged that way.
ICP confidence: High — VP Engineering, 1,517-person company, agentic platform is his explicit remit, and the company is shipping 5+ distinct production agents. Strongest fleet-scale fit found this run.
Note: Adam Seligman (Workato) is already in the People Library; this is a different, more technical entry point.
Signal: 1/4 (senior technical leader at an agent-active company; surfaced via LinkedIn people/leadership search after content-search buckets ran low). Source: https://www.linkedin.com/in/kefland/ and https://deepgram.com/company/leadership. Company size: ~250-300 (Deepgram, voice AI platform; Series B ~$86M raised; ships Voice Agent API and powers 200k+ developers building production voice agents). Pain points [INFERRED from role+company, not a verbatim quote]: cost and latency per voice-agent session at scale; reliability/accuracy of speech-to-action agents in production; proving cost-per-successful-interaction. Challenges: keeping real-time voice agents reliable and cheap as call volume scales; visibility into what each agent run costs. Must-haves: per-run/per-session cost + latency visibility; reliability guardrails in production. Nice-to-haves: model/route comparison on cost-per-completed-task. ICP confidence: Medium-High (VP Eng, ex-AWS SageMaker/Personalize & Lyft autonomous AI; company is agent-active and in the 50-2000 band; Deepgram is voice-AI infra so agent-shipping is via its Voice Agent product).
Signal: 1 & 4 (ICP writing about agent reliability/control in production AND building agents). Source: https://www.linkedin.com/in/kuldeepksc/ ; talk "Architecting Reliable Agentic Systems for Production" (https://www.startupgrantsindia.com/events/inside-emergent-ep1-architecting-frontier-agents). Company size: ~310 employees (75 in Jan 2026 -> 310 by Jun 2026); Series C unicorn ($1.5B, $130M Series C), AI-native vibe-coding company shipping autonomous coding agents, 5M+ users, $100M+ ARR. Pain points: agent reliability in production (prototypes that work in demo but break at scale); scaling autonomous coding agents to millions of users; cost of running agents at 200k+ paying-customer scale. Challenges: making multi-step agentic systems dependable and reliable at scale. Must-haves: production reliability, visibility/observability into agent behavior. Nice-to-haves: cost-per-run visibility, agents that compound/improve over time. ICP confidence: High (Head of AI at Series C AI-native firm actively shipping agents; publicly focused on reliable agentic systems at scale).
Kumar Ashutosh — AVP - Data Engineering & Product Management, Easyrewardz Software Services
Signal: Bucket 3 — ICP engaging with competitor/agent-framework content. He commented on LangChain's post announcing the new MCP-on-FastMCP agent API: "This will change the way how teams are integrating the MCP..."
Source: https://www.linkedin.com/feed/update/urn:li:activity:7499243729774387200/ ; profile detail from https://www.linkedin.com/in/shutupashu/details/experience/
Company size: 201-500 employees (LinkedIn company page, Easyrewardz — cloud CRM, loyalty and conversational commerce SaaS, Gurgaon).
Pain points (from his own experience entry): leading the data and AI product function at Easyrewardz with a 25-person cross-functional team (data engineers, AI engineers, PMs, analysts); responsible for "the transition of our platform stack toward GenAI-powered, agentic architectures"; architecting a production voice AI agent system to replace inbound call-centre operations, built on Exotel + Deepgram + Claude API + ElevenLabs, targeting 70%+ automation of inbound volume.
Challenges: moving a multi-tenant enterprise-retail CRM/loyalty platform from conventional stack to agentic architecture; integrating MCP/tooling consistently across agent stacks; hitting a hard automation-rate target on a multi-vendor voice agent pipeline (4 external model/infra vendors in one path).
Must-haves: reliable multi-vendor tool/model integration for a production voice agent; measurable automation rate; agent behaviour that holds up across many enterprise-retail clients.
Nice-to-haves: standardised MCP integration patterns so tool wiring is not re-done per stack.
ICP confidence: Medium-High — AVP (VP-level) owning the data + AI product function at a 201-500-employee SaaS with a named production agent programme. Caveat: since Apr 2026 he also holds an AVP - Engineering role at OneConsent.ai (small); the Easyrewardz role (Sep 2023-present) is the ICP-relevant anchor and both are listed as current.
Signal: 1 (ICP writing about context waste and agent reliability in production). NOTE: LinkedIn/Chrome not connected this run; found via first-party engineering blog. NOTE: Unit21's co-founder/CTO Clarence Chio (id 857) is already in the People Library — this is a second, distinct contact at the same account.
Source: https://www.unit21.ai/blog/4-ways-weve-engineered-around-the-ai-hallucination-problem-in-financial-crime-compliance
Company size: 120 employees (Tracxn, May 2026). Series C, $92M total. Fraud/AML detection, US.
What he said (paraphrased): too much information confuses the model — context engineering is the primary constraint on production agents.
Pain points: context-window overload degrading accuracy; hallucination/reliability risk across 213,000+ alerts per month; model drift over time; cost/quality tradeoff across 11-12 different models.
Challenges: making agent output auditable and regulator-ready; routing each task to the right model by benchmark rather than by default.
Must-haves: eval-set benchmarking, deterministic guardrails around LLM calls, continuous LLM-as-judge monitoring, human-in-the-loop escalation.
Nice-to-haves: automatic model swapping as new models beat the benchmark.
ICP confidence: High — CPO of an AI-focused product at a Series C, 120-person company; the blog is a near-verbatim statement of our context-waste thesis.
Signal: 4 (ICP technical leader at a company shipping AI agents). Method: web/company-leadership research. Source: https://www.gupshup.ai/ ('Autonomous AI Agents for Sales, Marketing & Customer Support'), https://www.linkedin.com/in/kunalpatke/. Company size: ~1,000-1,100 employees (2026; down from ~1,250 in 2025 after layoffs) - within 50-2,000 band. Series F. Agent activity: autonomous AI agents on Conversation Cloud across messaging channels (confirmed). Pain points (INFERRED, not verbatim): operating autonomous agents at massive messaging scale for many enterprise customers; LLM cost efficiency at high volume (post-layoff cost discipline heightens this); reliability across channels/languages. Challenges: cost per agent run at scale; margins. Must-haves: visibility + control over what each agent run costs. Nice-to-haves: right-size model routing per task. ICP confidence: Medium-High (SVP Eng = VP-level; ~1,000 emp, agent-active; recent downsizing noted).
Signal: 4 (ICP shipping agents). NOTE: LinkedIn/Chrome NOT connected this run — found via web research, not the prescribed LinkedIn searches. Source: https://theorg.com/org/appzen/teams/leadership-team ; https://www.crunchbase.com/person/kunal-verma ; authored company technical post https://www.appzen.com/blog/ai-journey-revolutionary-transformers-finance-ai-kunal-verma ; funding https://siliconangle.com/2025/09/22/appzen-raises-180m-automate-enterprise-finance-operations/. Company size: 373 employees as of June 2026 (Tracxn). Series D, $283M raised total incl. $180M round Sept 2025 (Lightspeed, Redpoint, 500 Global). Agentic AI confirmed: AppZen positions itself as an agentic AI platform for finance teams — AP, corporate card, T&E agents; company blog "How authoritative AI Agents are accelerating finance ROI". Verma personally owns product vision and oversees R&D and data science teams. Pain points (INFERRED from company-published material, NOT personal quotes): high-volume autonomous approval agents where a wrong call has direct financial consequence; audit-grade traceability per agent decision; per-invoice/per-transaction inference cost at enterprise document volumes. Challenges: multiple agents across AP/expense/card lines sharing infrastructure; proving ROI per agent to CFO buyers who scrutinize spend. Must-haves: cost-per-run visibility at transaction granularity; deterministic reliability and auditability. Nice-to-haves: model routing to cheaper models for low-risk documents. ICP confidence: High — CTO (technical co-founder), 373 employees, multiple agents in production. UNVERIFIED: no personal quote on cost/reliability captured; pains are inferred.
Signal: 4 (ICP building/shipping agents in production). Source: https://www.geekwire.com/2024/ex-microsoft-engineers-raise-25m-for-legal-tech-startup-that-uses-ai-to-help-lawyers-analyze-data/ and https://www.supio.com/about (LinkedIn URL not captured — do not fabricate). Company size: ~100–200 employees (doubling from ~100 in Apr 2025), Series B ($60M), Seattle legal AI platform. Lam is the technical co-founder & CTO (ex-Avalara/Microsoft; CS, University of Washington). Pain points (company-level, expressed via Supio): shift "from tools to agents," agents that take action inside real legal workflows; high reliability bar to produce verified, structured outputs attorneys trust without second-guessing. Challenges: production reliability/accuracy for regulated legal outputs; scaling agents as headcount doubles. Must-haves: agent reliability, output verification, observability. Nice-to-haves: cost/performance visibility per agent run. ICP confidence: High (CTO / technical decision-maker at a 50–2,000-employee AI-native company shipping agents in production). Note: added on company + role confirmation; no personal quote captured this run.
Laksh Krishnamurthy — Chief Technology Officer, Autonomize AI
Signal: 1+4 (ICP CTO at a company shipping healthcare agents + a no-code agent builder; found via 2026 dated agent-launch sweep). Source: https://www.globenewswire.com/news-release/2026/07/15/3327948/0/en/Autonomize-AI-Launches-Genie-AI-Autonomous-Agent-Transforming-Every-Healthcare-Expert-into-an-AI-Builder.html , https://www.linkedin.com/in/lkrishnamurthy/ . Company size: HEADCOUNT CONFLICT — PitchBook ~147, GetLatka ~45 (verify before outreach; likely in-band but confirm >50); Austin TX; Series A $28M (total ~$32M). Autonomize launched Genie AI (autonomous healthcare agent + no-code builder that lets clinicians/care teams build their own agents), 15 Jul 2026. Pain points (INFERRED): because customers self-build agents on Genie, Autonomize carries the runtime bill + audit exposure for agents it did not author (multi-tenant cost-attribution + governance); HIPAA/clinical compliance; reliability. Must-haves (INFERRED): per-customer/per-agent cost attribution, audit/compliance trail, reliability. Nice-to-haves (INFERRED): unified visibility across customer-built agents. ICP confidence: MEDIUM (named CTO + clear agent product, but headcount sources conflict 45 vs 147 — confirm the 50-employee floor).
Lars Maaløe — Co-Founder & CTO, Corti
Signal: 1 (public speaker on agents in production) + 4 (technical co-founder building agents). Source: https://home.mlops.community/public/videos/building-agents-for-healthcare-lars-maaloe-agents-in-production-2025 ; https://www.corti.ai/news/corti-launches-agentic-infrastructure-to-scale-ai-deployment-in-healthcare ; https://www.crunchbase.com/person/lars-maaloe. Company size: 51-100 employees (Copenhagen; $94.5M raised, last round Series B 2023). Corti ships a production-grade multi-agent "Agentic Framework" for healthcare that validates every agent action before execution using deterministic guardrails at an orchestration layer. Pain points: controlling/governing agent behavior at scale; validating agent actions before execution; reliability in regulated clinical settings. Challenges: preventing unsafe/incorrect agent actions in high-stakes medical workflows; specialized (not general-purpose) model reliability. Must-haves: deterministic guardrails, action validation, orchestration-layer control over allowable agent behavior. Nice-to-haves: per-agent observability/audit trail. ICP confidence: High — CTO/co-founder, 51-100 emp Series B, explicitly building & shipping multi-agent systems in production; pain maps directly to agent control/reliability. HIGH PRIORITY.
Signal: 4 (ICP technical co-founder/CTO at SaaS company shipping AI agents). Source: https://tracxn.com/d/companies/front/ ; https://front.com/ ; https://www.linkedin.com/company/fronthq. Company size: ~540 employees (May 2026); Series D (~$1.7B val), customer operations platform used by 9,000+ companies. Actively shipping agents: Front embeds AI agents into customer-service workflows (message triage/categorization, drafting, autonomous handling) with human-in-the-loop. Laurent Perrin is technical co-founder & CTO (Mathilde Collin & Nehal Shaikh co-CEOs). Pain points (inferred from product + domain, not verbatim quotes): reliability of customer-facing agents in production; per-interaction cost at high message volume across 9,000 customers; visibility/control over what agents do before they act. Challenges: balancing automation vs. human oversight while keeping agents reliable and cost-efficient at scale. Must-haves: reliability, control/guardrails, per-run cost visibility. Nice-to-haves: observability + compounding quality across runs. ICP confidence: Medium-High (technical co-founder & CTO, 540 emp in band, agents in production; Series D just past A-C band).
Laurent Sifre — Co-Founder (former CTO), H Company
Signal: 4. Source: https://en.wikipedia.org/wiki/H_(company) ; https://theorg.com/org/hcompany/org-chart/laurent-sifre ; Sifted (reports Sifre stepped down as CTO, 2026). Company size: ~90 employees (Paris; $220M seed; computer-use agents - Runner H, Surfer H, Tester H, Holo 3). Pain points (inferred; no direct quote this run): reliability of computer-use agents navigating arbitrary UIs; multi-step task completion at scale; cost of vision+planning agent loops. Challenges: leadership turnover (three co-founders departed 2025; Sifre reportedly stepped down as CTO 2026); building dependable agentic computer-use. Must-haves: reliable multi-step agent execution. Nice-to-haves: cheaper computer-use inference. ICP confidence: Medium (technical co-founder at 90-emp agentic-AI co; VERIFY current role/title before outreach given reported CTO transition).
Signal: 4 (ICP publicly describing the agent platform they are building/scaling). Source: LinkedIn post 1d before 2026-09-03 hiring a Principal Engineer – Agentic AI for JAGGAER Hyderabad (31 reactions, 5 reposts): "We're expanding our Agentic AI engineering organization at JAGGAER... build production-grade Agentic AI systems at enterprise scale... Agentic AI platform architecture and core capabilities / Multi-agent orchestration, reasoning, tool use, and agent design patterns / Scalable, secure, observable, production-grade AI systems." Company size: 1,503 employees as of March 2026 (Revelio Labs; PitchBook 1,200, LeadIQ ~1.5K) — inside band. Stage: private, PE-backed (Vista/Cinven lineage) procurement SaaS, not Series A–C — deviation from the stated stage filter, flagged. Agent activity: dedicated Agentic AI engineering org and an agentic platform embedded in JAGGAER's global procurement products; he owns "Agentic Platforms" as a titled function and is a site leader, so he is the build-vs-buy decision-maker for the runtime layer. Pain points (inferred from the spec he wrote, not quoted as personal pain): multi-agent orchestration at enterprise scale; making agent systems observable in production; secure, scalable agent design patterns; taking complex AI systems design → build → production → enterprise scale. Challenges: standing up an agentic platform org fast enough to keep up with product demand — he is hiring principal-level talent to architect it, which is the classic 1→N agent scaling wall. Must-haves: production-grade observability across multi-agent orchestration; agent design patterns and engineering standards that hold at enterprise scale. Nice-to-haves: cost/economics instrumentation — notably absent from his stated requirements, so cost pain is unproven and should be probed rather than assumed. ICP confidence: Medium — exact title fit (Senior Director of an agentic platform), verified headcount, verified active agent build-out; downgraded because the observed signal is a hiring post about capability rather than a first-person statement of cost, reliability or context-waste pain, and the company is PE-backed rather than Series A–C.
Signal: 4 (ICP shipping agents). Source: https://sleekflow.io/news/cto-pr-en ; https://www.media-outreach.com/news/hong-kong/2025/07/14/394893/sleekflow-unveils-agentflow-building-teams-of-ai-agents-to-increase-revenue-for-businesses/. Company size: 101-250 (Crunchbase 2026). Singapore HQ, with presence across Hong Kong, Malaysia, Indonesia, Brazil, UAE and the US. Raised $23.5M total including an $8M Series A led by Tiger Global. LIVE LINKEDIN VERIFIED this run — "Lei Gao — Chief Technology Officer at SleekFlow - we are hiring talented engineers!", San Francisco Bay Area. Agent evidence: SleekFlow launched "AgentFlow", explicitly marketed as "Building Teams of AI Agents" to run GTM and revenue workflows for 2,000+ business customers handling 600,000 conversations per day. The "teams of agents" framing is a direct 1-to-many scaling signal. Pain points: coordinating teams of multiple concurrent GTM agents reliably across a large multi-region customer base at 600K conversations/day. Challenges: joined as CTO in July 2024 (ex-LinkedIn China CTO) specifically to drive tech strategy — likely mid-buildout of the agent architecture, which is a high-value trigger window for infrastructure gaps. Must-haves: production reliability, cost visibility across many customer-facing agent deployments. Nice-to-haves: cross-region observability given the six-region footprint. ICP confidence: High — current CTO confirmed live on LinkedIn, in-band headcount and funding stage, explicit multi-agent production evidence.
Signal: 3/4 (ICP CTO at company shipping agentic AI product in production; adjacent to competitor agent-ops space). Source: Torq company page + "Torq Introduces New Agentic Builder for SOC Workflows" (channelinsider) and Torq LinkedIn posts. Company size: ~350–500 est. ($332M raised, $1.2B valuation, founded 2020). Torq ships agentic SecOps — launched "Torq Socrates," a Tier-1 AI SOC analyst agent closing ~90% of Tier-1/Tier-2 tickets autonomously, plus an agentic builder producing many SOC-workflow agents. Prior: co-founder/CTO Luminate Security (→Symantec); eng leadership at Check Point. Pain points (inferred from product + role): operating many autonomous SOC agents reliably at alert volume; precision/false-positive control; cost and observability of agent runs at scale; scaling 1→many agentic workflows. Challenges: trust/accuracy of autonomous triage; governing agent behavior. Must-haves: reliability, control, and cost visibility across many production agents. ICP confidence: High (CTO + technical co-founder, right size, 5+ agents in production).
Signal: 3 (ICP at a company whose entire product is adjacent to the agent-observability/competitor space). NOTE: LinkedIn/Chrome unavailable this run — LinkedIn URL surfaced in search results, not opened/verified.
Source: https://lightrun.com/lightrun-secures-70m-series-b/ ; https://www.insightpartners.com/ideas/lightrun-leadership-story/ (both name Ilan Peleg CEO / Leonid Blouvshtein CTO as co-founders)
Company size: ~75-150 — WEAKEST data point in this record. Tracxn/Finder (Startup Nation Central) cited 75; company states it expects to increase headcount in 2025-2026. Tel Aviv, Israel. Series B — $70M led by Accel + Insight ($110M total).
Agents in production: STRONG. "Runtime Autonomous AI Debugger" and "Lightrun AI SRE" — agents that correlate code changes, logs, runtime data and observability signals, generate live patches without redeploy, and validate fixes in live environments. Named Fortune 500 production customers: ADP, AT&T, Citi, ICE/NYSE, Inditex, Microsoft, Priceline, Salesforce, SAP, UnitedHealth, Booking Holdings. Revenue 4.5x YoY.
Pain points: [evidenced at COMPANY level — attributed to Lightrun, NOT to Leonid personally, do not misattribute] Their published production-reliability data is essentially a market thesis for agent operating pain: "43% of AI-generated code still requires manual debugging in production, even after passing QA and Staging"; "88% of organizations need 2-3 redeploy cycles to publish a single AI-generated change"; "44% of AI SRE or APM tool failures are due to execution-level data not being captured." [inferred, person-level] As CTO of an agent-that-acts-on-prod company, he directly owns "can I trust an autonomous agent to touch production."
Challenges: Agent/AI-code reliability in production; missing execution-level telemetry needed to debug agents; redeploy cycles as a cost multiplier; trusting autonomous remediation inside Fortune 500 environments.
Must-haves: Execution-level runtime data capture for agent debugging; validation of agent-generated fixes before/after they land; reliability evidence for enterprise buyers.
Nice-to-haves: Cost attribution per autonomous debug run; benchmarking agent fix quality across customers.
ICP confidence: Medium-High. High on title, agents-in-production, and stage. Downgraded because headcount sits near the 50 floor and rests on a single database estimate. NOTE: Lightrun is partly competitor-adjacent (they sell runtime observability for AI-generated code) — treat as a nuanced conversation, possibly partner-shaped.
Lev Konstantinovskiy — Head of Engineering (Voice AI), Synthflow AI
Signal: 1/4 (ICP identified directly via LinkedIn people search "Synthflow AI engineering"). Source: https://www.linkedin.com/in/levkonst/
Company size: ~50-150 (Synthflow AI, formerly Fine-Tuner; Berlin voice-AI-agent startup, Series A/B)
Pain points (inferred from role): voice-agent latency, per-call/token cost, reliability & consistency at call volume, orchestration of multi-step voice agents.
Challenges: keeping voice agents low-latency and cheap while reliable at scale.
Must-haves: cost per call visibility, low-latency reliable agents.
Nice-to-haves: model routing for cost/quality balance.
ICP confidence: High (Head of Engineering, agent-native voice startup, in range). Synthflow already an account in Brain; Lev is a new contact.
Signal: 1/3 (ICP technical founder writing/speaking explicitly about agent COST — Alpha's core wedge). Source: https://www.fastcompany.com/91550793/fireworks-ais-lin-qiao-wants-to-make-ai-agents-cheaper-to-run ; https://www.humanx.co/speaker/lin-qiao ; https://sequoiacap.com/podcast/training-data-lin-qiao/. Company size: est. ~200-400 employees (Fireworks AI; independent; Series C $250M Oct 2025 led by Lightspeed w/ Sequoia, Index, NVIDIA, AMD; $315M ARR Feb 2026 +416% YoY; in talks at ~$15B val). Fits 50-2,000. Technical co-founder (ex-Meta, led PyTorch). Fireworks builds internal agents (finance automated reporting/planning, legal contract-review, recruiting sourcing) AND is the inference platform powering agents for Uber, Quora, DoorDash; processes ~30T tokens/day. Pain points: making AI agents cheaper to run (explicit public thesis); inference cost economics of agentic workloads; agents moving from demos to everywhere in 2026. Challenges: cost + speed of running agents at massive token scale; reliability of production agents. Must-haves: cost-efficient inference for agents, per-run cost control. Nice-to-haves: small-model routing to cut agent spend. ICP confidence: Medium-High (technical co-founder/CEO explicitly vocal on agent cost; runs 5+ internal agents in production). Note: Fireworks is an inference-platform VENDOR — treat as build-vs-buy / partner-adjacent, not a pure buyer.
Lior Div — CEO & Co-Founder (technical; ex-Cybereason CEO, ex-Unit 8200), 7AI
Signal: 3 (agentic-security space, adjacent to agent-ops/observability competitors). Source: https://www.linkedin.com/in/liordiv ; https://blog.7ai.com/three-stars-thoughts-on-passing-the-100-employee-mark-at-7ai-and-whats-next ; https://7ai.com/. Company size: 100+ employees as of May 2026 (planning to double to ~200 by year end), Series A, ~$166M raised, founded 2024, Boston. Actively building AI agents: yes — agentic security/AI-SOC platform running autonomous agents; 7M+ agent investigations in production. Pain points: running large volumes of autonomous SOC agents reliably and cost-effectively at production scale; maintaining accuracy/trust of autonomous agents. Challenges: scaling agent volume (millions of investigations) while controlling per-investigation cost and reliability. Must-haves: reliable autonomous agents at scale; cost efficiency per agent run. Nice-to-haves: agent observability/control and cost attribution layer. ICP confidence: High (technical co-founder/CEO, Series A, 100+ employees, agents in production at scale). Note: pain points inferred from company's stated scale (7M+ investigations), not a verbatim quote.
Signal: 4 + 3 (ICP technical leader shipping agents; also in the observability/monitoring space adjacent to Alpha competitors). Source: https://www.businesswire.com/news/home/20250417869977/en/Monte-Carlo-Launches-Observability-Agents-To-Accelerate-Data-AI-Monitoring-and-Troubleshooting ; https://thedataexchange.media/lior-gavish-monte-carlo-data/ ; LinkedIn handle lgavish (from activity URL). Company size: ~549 employees (Series D, founded 2019). Pain points (inferred from public product/positioning): shipping "observability agents" (monitoring/troubleshooting agents) to production; Gavish: "AI agents are only as powerful as they are informed" — agents need rich context (data samples, metadata, query logs) to be reliable. Monitoring agent recommendations ~60% acceptance — reliability/precision of agent outputs is a live concern. Challenges: making agents reliable and context-rich enough to trust in production monitoring; observing agent behavior itself. Must-haves: agent context/grounding, output reliability, observability of agents. Nice-to-haves: per-run cost tracking for agent workloads. ICP confidence: High (co-founder+CTO; 50-2,000 emp; agents in production).
Lior Solomon — VP of Engineering, Data & AI, Drata
Signal: 4 (ICP writing publicly about building/shipping agents in production). Found via web research — Claude-in-Chrome/LinkedIn NOT connected this run.
Source: https://drata.com/blog/building-harness-engineering ("From Prompt Engineering to Harness Engineering", Drata engineering blog). Also posts under #agenticai on LinkedIn about agent activity governance.
Company size: ~689 employees (Tracxn, 31 May 2026; a second aggregator ~732). Series C — $200M co-led by ICONIQ Growth and GGV Capital at a $2B valuation; ~$328M total raised. Drata announced agentic AI capabilities across its trust-management platform in March 2026.
LinkedIn URL: https://www.linkedin.com/in/liorsolomon/
Pain points: No trustworthy record of what an agent did and why; agent actions that cannot be reconstructed for an auditor; state/identity/failure-recovery problems compounding as agent count grows. His summary line: building agents "isn't a prompting problem, it's a distributed systems challenge involving state, identity, and failure recovery."
Challenges: Frames the work as *harness* engineering, not prompt engineering — the gap between a demo-ready agent and one that survives an audit is the runtime around the model. Stated primitives: policy as a gate (authorization before action), evidence as a structured attributable type, and an "audit fabric" that decouples logs from decision trails. Must run multiple agent types in a regulated, evidence-producing environment at ~700-person scale.
Must-haves: Per-run traceability and decision trails; policy/authorization gating before agent actions; failure recovery and replay; attributable evidence output.
Nice-to-haves: Cost attribution per agent run; a standardized harness across agent types rather than bespoke plumbing per agent.
ICP confidence: High — non-founder VP Eng owning AI at a Series C, ~700-person SaaS company that shipped agents in production in 2026, publicly writing about exactly the agent-runtime/observability problem thealpha.ai sells into.
Signal: 3 (leads AI at an AI-observability vendor adjacent to competitor space — Coralogix AI Center overlaps with Langfuse/Helicone/Braintrust; surfaced via searches on agent cost + observability). Also touches Signal 1.
Source: https://coralogix.com/blog/coralogix-strengthens-ai-leadership-with-appointments-of-liran-hason-and-alon-gubkin/ ; https://coralogix.com/blog/introducing-coralogixs-mcp-server-helping-customers-build-smarter-ai-agents/
Company size: ~597–803 employees (LinkedIn lists 501–1,000); raised $200M Series F June 2026. Fits 50–2,000.
Background: Co-founder & ex-CEO of Aporia (AI observability/guardrails), acquired by Coralogix; now owns Coralogix AI strategy incl. AI Center (real-time AI observability) and MCP server for building AI agents.
Pain points (inferred from role/company focus, not a verbatim quote): real-time visibility into what agents cost and do in production; hallucination/quality/security monitoring for agents at scale.
Challenges: giving customers cross-stack observability into agent behavior, cost and reliability.
Must-haves: production-grade agent observability, cost/token attribution, guardrails.
Nice-to-haves: compounding evals, cross-model routing insight.
ICP confidence: High (VP of AI decision-maker at a ~600-person AI-native company actively building/observing agents). Note: as an observability vendor, buying intent for an agent operating layer is partial — strong role/company fit, adjacent product overlap.
Signal: Targeted ICP people-search — net-new agent-native company (Siena AI: autonomous CX agents for commerce; 'AI CX operating system'), not in brain. Source: https://www.linkedin.com/in/lisapopovici/ (LinkedIn people search 'Lisa Popovici Siena'). Company size: ~50-90 (Series A). Pain points (ICP-inferred; no direct post captured this run): agent reliability in customer-facing CX; cost per resolution at scale; maintaining quality/brand voice across many autonomous agents. Challenges: scaling autonomous CX agents across brands while controlling LLM cost. Must-haves: reliability, cost visibility. Nice-to-haves: model routing. ICP confidence: Medium (co-founder of agent-native company; likely product/growth co-founder rather than the primary technical leader).
Signal: 4 (ICP shipping agents). Second contact at Cytora alongside Aeneas Wiener (CTO) — multi-threading the account. NOTE: LinkedIn/Chrome NOT connected this run; LinkedIn URL NOT captured (not fabricating). Source: https://theorg.com/org/cytora/teams/leadership-team (title verified by direct page fetch) ; product context https://www.cytora.com/risk-flow-center/blog/cytora-autopilot-risk-workflows-that-run-themselves ; https://www.reinsurancene.ws/zurich-insurance-scales-cytora-ai-platform-across-global-underwriting-operations/. Company size: ~132-138 employees (PitchBook 132; Tracxn 138 as of Mar 2026). Agentic AI in production: Cytora Autopilot (Mar 2026) — agentic workflow orchestration with persistent context across communications; deployed at Zurich and Markel. Pain points (INFERRED from company-published material, NOT personal quotes): data/context pipeline feeding long-running agents that hold state across weeks of broker correspondence; structuring unorganized submission data cheaply enough to run at portfolio scale. Challenges: context growth over long-horizon agent runs; data quality driving agent reliability in a regulated underwriting decision path. Must-haves: visibility into what context each agent run consumes and what it costs. Nice-to-haves: context compaction; per-portfolio cost attribution. ICP confidence: Medium — Chief Data Officer is a senior technical AI-adjacent leader at/above the Director bar and sits in the agent decision path, but is a data-function rather than an engineering/AI-platform owner; company fit is High. UNVERIFIED: LinkedIn URL, personal pain statements, scope of his remit over agent infrastructure.
Lokesh Agrawal — Director of Engineering, Quality, Innovaccer
Signal: 4 (ICP engineering leader at an agent-shipping company; company-scoped LinkedIn people search "Innovaccer director AI agents").
Source: https://www.linkedin.com/search/results/people/?keywords=Innovaccer%20director%20AI%20agents (profile https://www.linkedin.com/in/lokeshagrawal/)
Company size: Innovaccer ~1,500-1,800 employees (under 2,000 ceiling). Building "AI-Native, Agentic QA Systems" per his headline. Later-stage than A-C (flagged).
Pain points [INFERRED from role + company + headline, NOT a verbatim post]: reliability/quality assurance of agentic systems in production; verifying agent outputs at scale; regression risk as agents change.
Challenges: engineering a repeatable reliability/verification layer for agentic QA.
Must-haves: production reliability + verification/eval guardrails for agents.
Nice-to-haves: cost-per-successful-run visibility tied to QA.
ICP confidence: Medium-High — Director of Engineering explicitly building agentic QA at an agent-shipping company; size in band, stage later than A-C (flagged).
Lu Cheng — Co-Founder & Chief Technology Officer, Zip (ziphq)
Signal: 1 + 4 (ICP posting publicly on LinkedIn about shipping agents, at a company running agents at fleet scale). Source: https://www.linkedin.com/posts/lu-cheng-973b7830_today-were-announcing-a-major-milestone-activity-7338245001971122176-sZ4c ('Zip launches AI agents for procurement') and https://www.linkedin.com/posts/lu-cheng-973b7830_the-zip-ai-lab-started-with-a-handful-of-activity-7444780237642694656-Zd-Y ('The Zip AI Lab started with a handful of...') ; https://zip.com/blog/introducing-agentic-procurement-orchestration ; https://procurementmag.com/articles/why-zip-is-launching-ai-superagents-procurement-native-mcp. Company size: ~1,038-1,259 employees (Revelio May-2026 1,259; ZipHQ Dec-2025 1,038; company says ~1.1K across 6 continents Feb-2026). Series D, ~$2.2B valuation. Scale of agents: 50+ purpose-built agents across procurement/finance/legal/IT/security; 1,000+ agents deployed across several hundred customers in ~10 months; Zip Superagents + procurement-native MCP in beta, GA summer 2026. Pain points (inferred from public product + posts, not verbatim): this is the clearest 1 -> many/thousands agent-sprawl case in the run - a thousand deployed agents across hundreds of customers is a cost, observability and governance problem before it is a modelling problem; Zip's own framing of 'agentic orchestration' as a new category signals they feel the control gap. Challenges: per-customer agent customisation (the 'Zip AI Lab' co-builds custom agents) means a long tail of bespoke agents to keep reliable and affordable; agents act on spend and contracts, so wrong actions are financially material. Must-haves: per-agent/per-run cost attribution across a multi-tenant agent fleet; reliability and approval controls on money-touching actions. Nice-to-haves: shared context/memory across the 50+ agent catalogue. ICP confidence: High - technical co-founder/CTO, ~1.1K employees (in band), actively posts about agents on LinkedIn (warm Signal-1 surface), and operates the exact fleet-scale pain Alpha addresses. Highest-priority target of this run. NOTE: not yet contacted; research only.
Signal: Bucket 3 — ICP engaging with competitor content (Arize AX). Found via web research; LinkedIn/Chrome unavailable this run.
Source: https://arize.com/customers/how-thefork-uses-evals-to-boost-conversions-with-arize-ax-on-aws/
Company size: ~960–964 employees (LeadIQ, Feb–Mar 2026). Restaurant booking marketplace, TripAdvisor subsidiary.
Agent evidence: production LLM/AI features on the revenue-critical restaurant discovery and booking path, instrumented with prompt-level tracing, online evals and drift alerting.
Verbatim quote (Arize case study): "Arize AX on AWS gives us prompt-level tracing, automated evaluations, and drift alerts, so we catch regressions early and meet strict SLOs at scale."
CAVEAT ON SOURCING: quote was confirmed via two independent search-result summaries of the Arize page; the live page body is JS-rendered and did not return full text on direct fetch. Treat quote as high-but-not-certain fidelity.
Pain points: catching model/prompt regressions before they hit conversion; meeting strict SLOs at scale; lacked this visibility before adopting a vendor (inferred from case-study framing).
Challenges: production AI tied directly to booking conversions — reliability failures are immediately revenue-visible.
Must-haves: prompt-level tracing, automated/online evaluation, drift detection and alerting, cost visibility.
Nice-to-haves: AWS-native deployment (chose "Arize AX on AWS" specifically).
ICP confidence: Medium-High — CTO title and headcount both clean, named on-the-record engagement with a competing observability vendor; downgraded from High only because the quote was not confirmed by a raw page fetch.
Profile URL surfaced as a search result link, not independently fetched — verify before outreach.
Signal: 3 (Knowunity appears on the Langfuse adopter list — engaging with competitor observability tooling) + 4 (company shipping a consumer AI agent at scale).
Source: name and title on Knowunity's own careers page "Meet our Founders" https://knowunity.com/careers ; corroborated https://www.eu-startups.com/2025/06/german-edtech-startup-knowunity-raises-e27-million-to-bring-ai-tutor-to-1-billion-students/ ; product evidence https://knowunity.com/ai/knowunity-ai ; adopter listing https://langfuse.com/handbook/chapters/customers
Company size: Company-stated "50+ Employees" on https://knowunity.com/careers ; PitchBook lists 175 (https://pitchbook.com/profiles/company/465294-07); a 2025 Series B write-up said 50–60. True figure is somewhere between ~60 and ~175 — every source clears the 50 floor, but this is the smallest company in this run's batch. Germany (Berlin).
Pain points: No direct Hild quote verified. Company-level: a personalised AI tutor grounded in 3M+ peer-created study materials, fine-tuned to national curricula, with users engaging the companion 7+ times per week. That is high-volume, low-ARPU consumer LLM traffic — the classic structural setup for token-spend pressure, since every additional engagement is direct COGS against a freemium consumer model.
Challenges: Careers-page manifesto: "Millions of students already engage with our AI study companion daily… The window to establish market leadership is measured in months, not years." They raised €27M Series B (€45M total) explicitly to scale the companion into the US and Asia — i.e. multiplying inference volume across new geographies and curricula on a fixed runway.
Must-haves: Cost per interaction that scales sub-linearly with user growth; visibility into which agent flows burn tokens without driving retention; reliability of a curriculum-grounded tutor where wrong answers are reputationally costly in an education market.
Nice-to-haves: Multi-model routing to cheaper models for routine flows; per-market cost attribution as they expand into the US and Asia.
ICP confidence: Medium — technical co-founder/CTO is a clean title match and they are already a Langfuse adopter, which proves they are actively instrumenting LLM behaviour. Downgraded from High because (a) headcount sits at the very floor of the band and is unresolved between ~60 and ~175, (b) no LinkedIn URL was verified on any page opened (a URL appeared in a search index only and was deliberately not recorded), and (c) the AI product is a tutor/companion rather than clearly multi-step tool-calling agents.
Luigi Basantes — Chief Technology Officer, Jelou AI
Signal: 4 (ICP shipping agents). Source: https://theorg.com/org/jelou-ai/org-chart/luigi-basantes ; https://jelou.ai/en/blog/series-A-10M ; https://www.intelligentcio.com/latam/2026/02/03/jelou-raises-us10m-to-build-ai-apps-that-move-money-on-whatsapp/. Company size: ~135 (PitchBook 2026; LeadIQ shows 85 in Ecuador plus smaller counts across Colombia, Mexico, Argentina, Chile). Guayaquil, Ecuador. Series A $10M Jan 2026, led by Wellington Access Ventures. Agent evidence: Jelou runs "Brain", a platform where AI agents verify identities, move money, sign documents and guide shopping directly inside WhatsApp — live across 500+ business customers in 13+ LatAm countries, with $100M+ in financial operations processed through the agents. Pain points: production reliability and security for agents that execute REAL financial transactions inside messaging apps — an unusually high blast-radius agent deployment. Challenges: scaling a transactional agent platform (payments, credit, KYC, document signing) safely as it expands beyond 13 countries. Must-haves: production reliability, auditability for financial-transaction agents, per-agent cost visibility. Nice-to-haves: faster agent-building tooling across the 3,000+ integrations in their Studio. ICP confidence: Medium-High — company size, funding stage and agent evidence all strong and verified; FLAG: title varies by source between CTO and VP of Engineering (he joined in 2019 and the title appears to have evolved), though both fit the ICP.
Luis Héctor Chávez — Chief Technology Officer, Replit
Signal: 3 (ICP engaging with competitor content — named reference customer on Braintrust's customers page).
Source: https://www.braintrust.dev/customers (video testimonial). Title corroborated by https://theorg.com/org/replit/org-chart/luis-hector-chavez and https://luma.com/g1i6ku0d ("Fireside chat: Luis Héctor Chávez (Replit CTO)").
Company size: ~404–597 (Revelio Labs 404, Mar 2026; Tracxn 597, Jun 2026). Series B/C, AI-native — Replit Agent builds full-stack apps, 30M+ creators.
Pain points: Evaluating/debugging agent behaviour at scale — on Braintrust: "Braintrust helped us identify several patterns that we wouldn't have found." Joined as early engineer 2020, led full infra redesign, named CTO 2024; now owns Replit's AI agent strategy "at scale."
Challenges: Running app-building agents at very high volume; agent run cost and quality are directly tied to Replit's unit economics (they meter agent usage to customers).
Must-haves: Eval + trace tooling that surfaces failure patterns they cannot find by inspection.
Nice-to-haves: Cost attribution per agent run to support their usage-based pricing model.
ICP confidence: High — CTO, company in band, AI-native, already buying in the agent observability/eval category (so budget and category awareness both exist).
Note: Braintrust quote is thin — the adoption signal is stronger than the pain articulation.
Signal: 4 (ICP building/shipping AI agents in production; also 3 — voice/competitor-adjacent). Source: https://www.startuphub.ai/ai-news/funding-round/2025/happyrobot-raises-44m-series-b-to-scale-supply-chain-ai-agents ; https://techfundingnews.com/happyrobot-44m-series-b-ai-workforce-supply-chain/. Company size: ~183 employees (Tracxn, Jun 2026); Series B $44M (Base10, a16z, YC), ~$500M valuation, ~$62M total raised; HQ SF. Pain points (INFERRED): scaling autonomous voice/email agents that negotiate freight rates, schedule appointments and parse shipping docs across 70+ enterprise customers (DHL, Ryder); keeping human-like voice agents reliable in production; per-call/per-task cost as call volume scales. Challenges: agent reliability + accuracy at enterprise voice volume; integrating agent actions into TMS/ERP; latency/cost of high-volume voice+LLM calls. Must-haves: production reliability + observability of agent actions; per-run cost visibility as volume scales. Nice-to-haves: model/routing optimization to cut token/voice costs. ICP confidence: High (Co-Founder & CTO; Series B; ~183 emp in-band; multiple autonomous agents live at enterprise scale).
Luiz Scheidegger — Head of Engineering, Lindy
Signal: 3 (ICP engaging with competitor/adjacent tooling content — named customer in a Temporal case study, the durable-execution layer Alpha competes beside).
Source: https://temporal.io/resources/case-studies/lindy-reliability-observability-ai-agents-temporal-cloud
Company size: ~52 employees (Tracxn, May 2026). Series B, $50M+. AI agent orchestration platform, SF. NOTE: right at the 50-employee floor and estimates vary by source — treat headcount as the weakest fact in this record.
Pain points (quoted in the case study): "We rolled out a complex in-house system just to deal with execution failure. But it wasn't durable, reliable, or observable." Agents were failing silently and unpredictably on third-party API timeouts.
Challenges: no visibility into agent execution paths; a lean team carrying the maintenance burden of self-built reliability scaffolding it did not want to own; silent failure modes that only surface as customer-visible breakage.
Must-haves: durable execution, execution-path observability, first-class retry primitives.
Nice-to-haves: developer-friendly self-serve tooling with no ops overhead (explicitly called out — small team, no platform group to run infrastructure).
ICP confidence: Medium — title (Head of Engineering) and pain are a clean match, and Lindy is an archetypal agent-fleet operator. Downgraded from High only because headcount sits exactly on the 50 floor and the case study is about durability/reliability rather than cost per run.
OUTREACH ANGLE: he has already bought a reliability layer (Temporal) and has NOT solved cost attribution. Alpha's per-run cost artifact sits next to Temporal rather than against it.
Note: Lindy already appears in the library via Flo Crivello (Founder & CEO). Scheidegger is the engineering-side entry point and the one who actually owns this pain.
Luka Chkhetiani — Head of Realtime Product & Principal Researcher, AssemblyAI
Signal: 4 (technical product/research leader at a voice-AI / voice-agent company, found via LinkedIn currentCompany-facet People search). Profile: https://www.linkedin.com/in/lchkhetiani/ . Source: https://www.linkedin.com/search/results/people/?currentCompany=%5B%2218644094%22%5D (AssemblyAI). Company size: 51-200 (LinkedIn). Pain points (inferred): unit economics + reliability of realtime voice-agent products; per-run/per-minute cost of streaming inference; production reliability of non-deterministic realtime models. Challenges: shipping realtime voice-agent features while controlling inference cost. Must-haves: cost-per-run observability, reliability controls. Nice-to-haves: model routing. ICP confidence: Medium (Head of Realtime Product & Principal Researcher - clear technical function-lead; AssemblyAI agent-adjacent voice AI; 3rd-degree connection). Pain points INFERRED from role+company, not fabricated.
Signal: 4 (ICP CTO building/shipping a multi-agent system; found via 2026 agent-launch/funding sweep). Source: https://www.prnewswire.com/news-releases/lio-raises-30m-series-a-to-bring-agentic-ai-to-enterprise-procurement-302705236.html , https://lio.ai/about-us , https://www.linkedin.com/in/lukasheinzmann/ . Company size: ~80 employees; Munich, Germany; Series A $30M led by a16z (Mar 2026); founded 2023 (YC S23). Lio = 'world's first multi-agent system for procurement' — specialized agents run in parallel (vendor research, negotiation, approvals, delivery tracking) for each purchase request. Heinzmann is CTO, ex-AI/ML lead (myAudi app), prior founder of Gabriel AI. Pain points (INFERRED): reliability/coordination of many parallel agents per request; procurement is compliance- and approval-sensitive so agent actions must be auditable; cost of running multiple agents per purchase order. Challenges (INFERRED): controlling multi-agent behavior at scale. Must-haves (INFERRED): control/observability over multi-agent orchestration, audit trail for procurement compliance, cost-per-request visibility. Nice-to-haves (INFERRED): per-agent cost/quality attribution. ICP confidence: HIGH (named technical co-founder/CTO; in-band headcount; Series A stage fits ICP exactly; explicit multi-agent production system).
Lukas Pöhler — VP AI Solutions, Aleph Alpha
Signal: 4 (ICP building & deploying enterprise/sovereign AI agents; surfaced via LinkedIn people search). Source: https://www.linkedin.com/in/lukaspoehler/ | Company size: ~350 employees (Series B, ~$533M raised; PhariaAI full-stack enterprise/sovereign GenAI suite for bespoke agentic use cases; merger with Cohere announced Apr 2026 — caveat). Pain points (inferred): deploying enterprise/sovereign AI agents with compliance & transparency; cost & reliability of bespoke agentic use cases; on-prem/sovereign deployment constraints. Challenges: operationalizing bespoke enterprise agent use cases at scale; cost control & governance. Must-haves: governance/observability, cost control, reliable enterprise deployment. Nice-to-haves: multi-model flexibility. ICP confidence: Medium — VP AI Solutions at a Series B ~350-person enterprise agentic-AI company; caveat: pending Cohere merger.
Lukas Toma — Director of Data (AI/ML & Data Engineering), Zapier
Signal: 4 (ICP Director-level AI/ML & data engineering leader at validated target company Zapier; sourced via LinkedIn People search, run 2026-08-01). Source: https://www.linkedin.com/in/lukastoma/ . Company size: ~800 (Zapier — automation platform actively shipping AI agents: Zapier Agents / Zapier Central; in-range, agent-shipping). Pain points (role/company-contextual): cost of LLM calls across high-volume automation/agent runs; token/compute waste at scale; per-run cost visibility as agents proliferate across customer workflows. Challenges: reliability of agents triggering many downstream API calls; scaling agent infra from features to platform. Must-haves: per-agent/per-run cost visibility; production reliability. Nice-to-haves: model routing, caching, eval automation. ICP confidence: Medium-High (Director of Data/AI-ML Engineering at 50-2,000-emp company actively shipping AI agents).
Signal: 4 (technical co-founder at a qualifying company shipping agents in production). LinkedIn verified live 2026-09-02: headline reads "Co-founder @ Zowie - AI agents that work for your customers", New York. Note: LinkedIn headline says "Co-founder" only; the CTO/CPTO designation comes from Crunchbase and company sources, so treat the exact title as co-founder-technical rather than a confirmed CTO string.
Source: https://getzowie.com/ (company site) plus Crunchbase profile
Company size: ~110-115 employees (Tracxn, 2026); Series A, $14M led by Tiger Global. Zowie = AI customer-service agents for regulated enterprises (banking, insurance, telecom), reporting 100M+ conversations per year.
Pain points: HONEST CAVEAT — no verbatim pain quote from Ciolek was found in this run despite checking podcast appearances (Inovo, Spamming Zero). He is recorded here on the strength of role, stage, headcount and product fit, not on a personal quote. Do not attribute pain language to him in outreach.
Challenges (inferred from company profile, NOT quoted): running agents at 100M+ conversations/year in regulated verticals where a wrong answer has compliance consequences; unit economics of high-volume conversational agents on a Series A budget.
Must-haves (inferred): reliability and auditability per conversation; cost-per-resolution economics that hold at 100M conversations.
Nice-to-haves (inferred): per-customer agent cost attribution for their own enterprise clients.
ICP confidence: Medium — company, stage, headcount and product are an excellent fit and the volume implies real cost and reliability stakes, but the required public pain signal is unverified for this specific person. Treat as a qualified research lead needing one more verification pass before outreach, not a signal-backed prospect.
Signal: 4 — technical co-founder/CTO at an AI-native company whose core product is autonomous 'vibe-coding' agents that build applications end to end.
Source: https://techcrunch.com/2026/07/15/indian-ai-coding-startup-emergent-becomes-a-unicorn-just-over-a-year-after-launch/
Company size: ~75 employees (Series C, $130M raise, $1.5B valuation, $120M ARR, 200k+ paying customers)
Pain points: improving the success rate / reliability of agent-built applications; scaling agent workflows for a large paying user base
Challenges: production reliability of autonomous build agents at scale
Must-haves: reliability and success-rate visibility for agent runs
Nice-to-haves: cost efficiency per build/run
ICP confidence: Medium-High — AI-native, right size/stage, technical C-suite actively shipping agents. NOTE: LinkedIn URL not verified; left blank to avoid fabrication
Madison May — Chief Technology Officer (Co-founder), Indico Data
Signal: 4 (ICP at a company shipping a multi-agent platform in production). Source: https://www.prnewswire.com/news-releases/indico-data-launches-industrys-first-agentic-decisioning-platform-purpose-built-for-insurance-302466802.html ; promotion to CTO: https://indicodata.ai/blog/news/indico-data-the-decision-automation-company-unveils-strategic-leadership-appointments-to-drive-next-generation-insurance-innovation/
Company size: ~81-83 employees (Dealroom, RocketReach, PitchBook, May 2026) — in range. Insurance decision-automation, agentic platform.
Pain points: Hallucination and black-box decisioning in production insurance workflows. Product is explicitly architected around a "Validation Agent" that "verifies accuracy and completeness with real-time feedback. All outputs are fully traceable - ensuring no black-box decisions or hallucinated data." Reliability is the stated buying criterion for their own enterprise deployments.
Challenges: Running a suite of agents (submission, underwriting, claims, validation) across regulated insurer environments where every output must be explainable and auditable; scaling to 4x submission-handling capacity without losing traceability.
Must-haves: Per-step traceability of agent decisions; explainability and auditability for regulated buyers; guardrails against hallucinated field extraction.
Nice-to-haves: Cost-per-submission visibility; automated regression detection as models are swapped.
ICP confidence: High — verified CTO/co-founder, headcount in range, documented agents in production with named enterprise deployments, explicit reliability/traceability pain language.
Signal: 1 (ICP speaking publicly about agent/LLM cost, compounding and control). Found via web research — Claude-in-Chrome/LinkedIn NOT connected this run.
Source: https://home.mlops.community/public/videos/how-we-cut-llm-latency-70percent-with-tensorrt-in-production (MLOps Community podcast, 10 Apr 2026 — full transcript on page)
Company size: ~160 employees (Crunchbase and Tracxn). Series C (latest round Nov 2023); ~$132M raised per Crunchbase (Tracxn lists ~$177M). AI-powered performance management / talent intelligence SaaS.
LinkedIn URL: https://www.linkedin.com/in/maherhanafi/ (seen in search results with headline "Senior Vice President Of Engineering"; profile page not loaded — reconfirm)
Pain points: Verbatim — "one of the other things that was taking so much time for me recently is like, how I manage cost of AI so I can prove to leadership, to the board, to the investors, and to the market that you can build AI at a positive ROI… build AI that doesn't break your wallet." On runaway spend: "within engineering, we had the finance kind of role within engineering… ensuring everything goes within an allowed budget. Because otherwise if you open the door to everything, it's just gonna blow up." Flags that on some premium models "their tokens will be 10 times more expensive than regular tokens." Uncontrolled internal coding-agent consumption; latency and cost coupled in production inference.
Challenges: Building FinOps discipline inside engineering from scratch; attributing cost to features and teams granularly enough to support quotas; reporting GPU/token cost savings daily, weekly and monthly to non-technical stakeholders.
Must-haves: Per-team and per-feature LLM/GPU cost attribution; enforceable budgets and quotas including for internal agent usage; recurring cost reporting he can take to leadership; model-selection levers that cut cost without hurting quality.
Nice-to-haves: Automated anomaly alerting on spend spikes; ROI framing tying spend to business outcomes; latency/cost co-optimization visibility.
ICP confidence: Medium-High — SVP Engineering at a ~160-person Series C SaaS who personally owns LLM cost and ROI and discusses it in unusual operational detail. Gap: Betterworks is AI-enabled SaaS rather than agent-native, so "5+ agents in production" is only partially satisfied (internal coding agents are confirmed; customer-facing agent count unverified).
Signal: 4 (ICP is a Head-of-AI-level leader at a company shipping AI agents in production). NOTE ON METHOD: LinkedIn browsing was unavailable this run (Claude-in-Chrome extension not connected); found/verified via web research, not the prescribed LinkedIn searches. Pain points below are INFERRED from Parloa's public product/domain, not verbatim quotes.
Source: https://www.parloa.com/about-us/ ; https://openai.com/index/parloa/ ; https://www.parloa.com/webinars/what-it-takes-to-build-scalable-ai-voice-agents/
Company size: ~400 (Series C-stage contact-center AI agent platform; enterprise voice/customer-service agents).
Pain points (inferred): building scalable, reliable voice AI agents for enterprise contact centers; simulating/testing agent behavior before production; managing cost/reliability of high-volume real-time agent interactions.
Challenges (inferred): reliability and evaluation of voice agents at enterprise scale; agent management across many concurrent deployments.
Must-haves (inferred): reliability/eval tooling and production visibility for agents.
Nice-to-haves (inferred): cost-per-interaction visibility.
ICP confidence: Medium — clearly a "Head of AI" persona at a ~400-person agent-platform company (in range), though title leans strategy; verify exact technical scope before outreach.
Malte Kosub — Co-Founder & CEO, Parloa
Signal: 1 (ICP speaking/writing about building resilient enterprise AI agents and production reliability). Source: https://app.dealroom.co/news/note/parloa-ceo-malte-kosub-on-building-resilient-enterprise-ai-agents-and-the-shift-to-conversational-multimodal-interfaces ; https://www.parloa.com/about-us/ . Company size: ~400+ employees (Berlin/NY/Munich; Series B $66M + later $120M unicorn round). Pain points: making enterprise voice/contact-center AI agents reliable and resilient in production; validating agent behavior via real-world simulation and eval test cases before go-live; shift to conversational multimodal interfaces. Challenges: guaranteeing correct responses in live customer conversations at scale; resilience of agentic systems. Must-haves: production reliability/accuracy of voice agents, pre-deployment evaluation/simulation. Nice-to-haves: multimodal channel coverage. ICP confidence: High (technical co-founder/CEO at Series B/C AI-native company shipping voice agents in production, in target size band).
Malte Pietsch — Co-Founder & CTO, deepset
Signal: 1 (ICP writing about agent costs, control, and observability). Found via web research — Claude-in-Chrome/LinkedIn NOT connected this run (2 retries), so prescribed LinkedIn post searches were unavailable; pivoted to first-party engineering blogs + primary-source verification.
Source: https://www.deepset.ai/blog/haystack-3-sovereign-agents (Haystack 3.0 launch essay, 20 Jul 2026, authored under his own byline as CTO)
Company size: ~84 employees as of 30 Jun 2026 (Tracxn). Series B, $30M led by Balderton; ~$45.6M total. Berlin + NYC.
Agent evidence: Haystack open-source agent framework + Haystack Enterprise Platform. Shipped modular Skills (versioned agent instruction artifacts w/ progressive disclosure), programmable agent Hooks (pre/post model call, around tool execution, approval gates before high-impact tool fires), built-in introspection (token usage, step counts, tool-call counts, step-level tracing spans), MCP tool exposure for Haystack pipelines (Jun 2026), agent Traces in Enterprise Platform (30 Jul 2026). Named production users: European Commission, Bosch, Airbus, The Economist, OakNorth.
Verbatim: "Every agent runs inside a harness: the loop, tools, memory, context, and controls that turn a model call into a system. If you cannot understand or leave that harness, moving it into a more secure environment does not make it yours."
Pain points: vendor/harness lock-in — knowledge moving out of model weights into artifacts he cannot read or audit; debugging unexpected agent behaviour by reconstructing runs from scattered logs; runaway agent loops and unpredictable token spend; context windows growing past useful size mid-run.
Challenges: making agent behaviour auditable and reproducible for regulated public-sector/industrial buyers; enforcing approval gates before high-impact tool calls; testing agents without paying per-test LLM costs or fighting variable provider responses.
Must-haves: model/provider independence (swap for price, capability or jurisdiction without rebuilding); open-source self-hostable inspectable execution loop; step-level tracing tagged with tools used; programmable interception points in the agent loop; data residency / sovereign deployment.
Nice-to-haves: mock generators/embedders for offline testing; cost-triggered context compaction; leaner dependency footprint; portable skill artifacts shareable across teams.
ICP confidence: High — 84-person Series B AI-native company whose entire product is production agent infrastructure, and the CTO personally publishes the architecture. Directly articulates token-spend unpredictability and agent-loop control, which is Alpha's exact wedge.
Signal: 1/4 (ICP CTO writing/speaking publicly about agents in production + cost). Source: https://vercel.com/go/ama-shipping-ai-cto ; https://www.teamday.ai/ai/malte-ubl-agents-new-application-layer ("Agents Are the New Application Layer"); https://www.latent.space/p/ship-ai-recap-agents-workflows-and. Company size: ~600-900 employees (Vercel; independent, Series E, ~$9B val). Fits 50-2,000. Actively building & shipping agents: v0 (agentic app-builder product) + internal support agent with 90% deflection rate; AI SDK powers agent workflows for many customers. Pain points: optimizing agent performance vs cost (net CPU-based execution to reduce operational cost of AI processing); running agents reliably as the "new application layer"; agent workflow orchestration. Challenges: cost efficiency of agentic workloads at scale; reliability of autonomous agents in production. Must-haves: cost/performance optimization per agent run, reliable agent orchestration. Nice-to-haves: per-run cost attribution, model routing. ICP confidence: High (technical CTO/decision-maker at a mid-size independent company shipping agent products + internal production agents; explicit cost-efficiency signal). Note: Vercel is later-stage but headcount fits band.
Signal: 4 (Director of AI Engineering at ICP agent company; via LinkedIn People directory, listed among Writer employees). Source: https://www.linkedin.com/in/manhal (UAE). Company: Writer, 201-500, Series C, enterprise agentic AI. NOTE: surname is privacy-truncated to "D." on LinkedIn; profile URL provided for disambiguation. Pain points (inferred): scaling reliable AI/agent engineering; LLM cost + token/context efficiency; production visibility. Challenges: reliability + cost of enterprise agents at scale. Must-haves: cost-per-run visibility; reliability guardrails. Nice-to-haves: model optimization. ICP confidence: Medium (clean Director title & ICP company; full surname unconfirmed). NOTE: pain points inferred from role/company, not verbatim quotes.
Signal: 4 (ICP author writing about shipping agents in production; post surfaced via reshare in Gaurav Narasimhan's LinkedIn activity feed).
Source: https://www.linkedin.com/in/manick-bhan/recent-activity/all/ — post ~3 weeks old, linking to agenticmarketingnews.com article "The Autonomous Marketing Agent Shipped. But It Answers to a Human?" (9 reactions, 1 comment).
Company size: ~75–250 employees. Sources conflict — LinkedIn shows a 51–200 band (~155 indexed), Bhan publicly describes a fully remote team of 250+, Latka reported 75 employees / $30M ARR (Jul 2025). All estimates fall inside the 50–2,000 ICP band. Agentic SEO/marketing platform (OTTO agent); AI-native.
Pain points (verbatim from his post): "Every week someone sends me a headline saying AI agents are a bust. I run agentic systems inside Search Atlas every day, so I read those headlines from the other side of the build." — i.e. defending agent reliability against public skepticism while operating agents in production daily.
Challenges: Running agentic systems in production continuously at a company whose core product IS the agent; the gap between demo-stage agent narratives and what production operation actually requires. Note: he did not explicitly discuss cost or token spend in the observed post.
Must-haves: Agents that hold up in day-to-day production operation (reliability as a product requirement, not a nice-to-have) — inferred from the product being agent-native, not from an explicit statement.
Nice-to-haves: Not observed — do not assume.
ICP confidence: High — technical founder/CTO (C-suite technical, valid per ICP), AI-native company inside the size band, product is agents in production, and he publicly writes from the operator side of agent reliability. Caveat: no direct cost/spend pain expressed yet, so cost-angle outreach is unvalidated for him.
Second contact at same company already in brain: Gaurav Narasimhan (SVP Applied AI, Agents) — do not double-count the account.
Manish Raana — Head of Engineering, SingleInterface
Signal: 4 (ICP-title engineering leader building agentic systems). Source: https://www.linkedin.com/in/rmanish08/ (LinkedIn people search "Director of Engineering agentic AI agents"). Company size: 201-500 employees (SingleInterface, Gurgaon; verified via LinkedIn company page — "Asia's largest AI-Powered Retail Tech Platform," 400+ multi-location brands, Deloitte Fast 50). In 50-2,000 band. Role: Head of Engineering; headline explicitly "AI-native platforms for local growth | Agentic systems, MCP, scale | ex-Cisco, CarDekho, Myntra." Pain points (inferred from role/headline): building and scaling agentic systems / MCP integrations on a retail-commerce platform; reliability and cost at scale. Challenges: making agentic features dependable across a large multi-brand platform. Must-haves: scalable agent infra, reliability, cost control. Nice-to-haves: MCP tooling, observability. ICP confidence: Medium (Head of Engineering personally building agentic systems at an in-range company; caveat — SingleInterface is primarily a retail/hyperlocal martech platform, so "5+ agents in production" is not fully confirmed — the individual's agentic-systems mandate is the qualifying signal). profile_url captured in Source too.
Manisha Taparia — VP, AI Automation (Strategic Projects), ShipBob
Signal: 4 (ICP AI leader actively building agents; found via LinkedIn People search for ICP titles + agents). Source: https://www.linkedin.com/search/results/people/?keywords=VP%20of%20AI%20shipping%20agents%20startup ; verification: https://www.prnewswire.com/news-releases/shipbob-launches-first-anthropic-verified-fulfillment-connector-anchoring-its-ai-suite-302842134.html and https://tracxn.com/d/companies/shipbob/. Company size: ~1,500 employees (2026, up from 1.3K in 2024), Chicago; $333M+ raised, ~$1B valuation (later stage — caveat vs A-C band; size in band). Actively building agents (CONFIRMED): "ShipBob AI" suite across the fulfillment stack; "Bobby" in-dashboard agent (beta, GA Fall 2026); ShipBob MCP server making inventory/orders/shipments queryable by AI assistants; autonomous robots + AI vision. Manisha leads AI automation for logistics workflows (agentic). Pain points (inferred): reliability/governance of agents acting on live merchant operations; scaling agentic automation across the fulfillment stack; per-run cost + observability. Challenges: making agents reliable and cost-controlled across a large merchant base. Must-haves: reliability, control/governance, cost visibility. Nice-to-haves: observability, compounding quality. ICP confidence: Medium (VP AI at 1,500-emp fulfillment-tech co actively building agents; caveats — later stage than A-C, some agents are internal-ops). LinkedIn profile URL not captured this run — do not fabricate.
Manoj Kintali — Head of Engineering, Adonis
Signal: 4 (ICP at a company publicly shipping agents in production — healthcare RCM vertical).
Source: https://www.prnewswire.com/news-releases/adonis-achieves-over-4x-revenue-growth-marking-a-milestone-year-in-the-shift-to-agentic-ai-for-healthcare-rcm-302666824.html ; Series C: https://www.prnewswire.com/news-releases/adonis-raises-40m-series-c-to-equip-healthcare-providers-with-aidriven-revenue-cycle-operations-302722199.html
Company size: 244 employees (Crustdata / Tracxn, 36.3% YoY growth). Series C, $40M. AI agents for healthcare revenue cycle management.
Pain points: company positions agents that "proactively detect issues, recommend tailored actions, and execute resolutions" across billing, denials and prior-auth — autonomous execution in a regulated, high-stakes domain where a wrong agent action has financial and compliance consequences. Adonis is actively hiring an AI/ML Lead, a standard tell for a team hitting the 1-to-many agent scaling wall.
Challenges: agent reliability and auditability under HIPAA/payer-compliance constraints; orchestrating multiple agents across distinct RCM workflows; 4x revenue growth means agent volume is scaling faster than the platform underneath it.
Must-haves: auditable, explainable agent decisions; production reliability at claim volume.
Nice-to-haves: broader agent autonomy as trust accrues; per-claim cost attribution (Adonis prices against recovered revenue, so cost per agent run maps directly to gross margin).
ICP confidence: Medium — title, headcount and stage all verified and in band, and the company is unambiguously shipping agents in production. Downgraded from High for one honest reason: NO direct first-person quote from Kintali was found. All pain signals here are company-level statements plus a reasonable inference from the hiring pattern, not his own words.
NOTE: Adonis already appears in the library via Akash Magoon (Co-founder & CEO) and Aman Magoon (CPO). Kintali is the engineering owner and the correct technical entry point.
Manoj Nair — CTO & Chief Innovation Officer, Snyk
Signal: 1 (ICP speaking about agent control/governance architecture) + 4 (ICP at a company shipping agents).
Source: AI Engineer World's Fair 2026 speaker record — https://ai.engineer/worldsfair/2026/speakers.json ; sessions "Through the AI Fog: The architectural decision the next 24 months of agentic security depends on" and Security Track intro. Leads Snyk's Emerging Technologies and Solutions Office.
Company size: ~1,870 employees (Aug 2026) — inside the 50–2,000 band, though near the ceiling and shrinking (90-person layoff reported in 2026 as it reorganises around AI). Launched Evo Agentic Development Security (Jun 2026), which governs "what agents use, what they do, and what they generate — in real time" across 4,800+ customers.
Pain points: session titles indicate he is publicly framing agentic control/governance as THE architectural decision of the next 24 months. Note: session abstracts are empty in the source dataset, so the specific pains are inferred from Snyk's own product positioning rather than his direct words — VERIFY with a primary quote before outreach.
Challenges (inferred): governing a fleet of agents that write and ship enterprise software; real-time control over agent actions and outputs.
Must-haves (inferred): a real-time enforcement/control layer over agent behaviour; per-agent identity and accountability.
Nice-to-haves (inferred): cost attribution alongside the security controls he already sells.
ICP confidence: Medium — CTO of a company that both ships agents and sells agent governance, headcount in band. Downgraded from High because (a) Snyk is very late stage, far outside Series A–C, (b) headcount is near the 2,000 ceiling, and (c) as a vendor of agent-governance tooling he is partly a competitor/partner rather than a pure buyer.
Manuel Romero — Co-founder & Chief Scientist (CSO), Maisa AI
Signal: 1 (reliability/accountability of agents in production — found via web research, LinkedIn/Chrome unavailable this run; pain inferred from company positioning, no personal quote). Source: https://forgepointcap.com/perspectives/maisa-ai-gets-25m-to-fix-enterprise-ais-95-failure-rate-techcrunch/ ; https://maisa.ai/agentic-insights/maisa-raises-25m-from-creandum-and-forgepoint/ . Company size: 93 employees. Funding: ~$30M total ($25M seed/Series A led by Creandum & Forgepoint; $5M pre-seed NFX/Village Global). Product: trustworthy "digital workers" powered by Knowledge Processing Unit (KPU) + "Chain-of-Work" for accountability/traceability. Background: top global HuggingFace contributor (700+ open-source models). Pain points (inferred): enterprise AI's high production failure rate; making LLM agents trustworthy/accountable task executors; traceability of agent actions. Challenges: reliability, hallucination, accountability in production agents. Must-haves: traceability/accountability, reliability. Nice-to-haves: cost control per run. ICP confidence: High (C-suite technical/AI leader, 93 emp, Series A, ships agent product).
Marcin Wyszynski — Co-Founder & CTO, Spacelift
Signal: 4 (ICP publicly discussing agent reliability/guardrails). Source: https://thenewstack.io/spacelift-ai-infrastructure-code/ (The New Stack Agents podcast, 20 Mar 2026); secondary: Spacelift Intelligence launch Mar 2026. Company size: ~126-157 (Unify 157 Jun 2026; PitchBook 126). Stage: Series C ($51M, Five Elms Capital, Jul 2025). Pain points (direct quotes): "The enterprise objection is always that LLMs aren't deterministic, so you can't trust them" — his answer: "We got used to the fact that humans need guardrails. There's nothing new conceptually in having LLMs require guardrails as well." Comprehension gap (Portuguese-phrasebook analogy): "He understood our question, but we have no way of understanding his answer" — nobody hand-writes HCL anymore so nobody can review agent output. Blast radius: "A bad application deploy can usually be rolled back, but a bad infrastructure change can destroy a production database." Velocity mismatch: "Developers are moving so fast that traditional Infrastructure as Code pipelines alone can't keep up." Challenges: letting an LLM create/update/delete cloud resources in near-real-time while staying inside deterministic bounds; building a context layer giving the LLM awareness of existing modules and enforced policies. Must-haves: DETERMINISTIC guardrails, explicitly "not just other LLM calls" (he injects Open Policy Agent as middleware); a context layer over org modules/policies; one-click path from agent-provisioned resource to reviewed IaC. Nice-to-haves: per-run cost/attribution on agent-driven provisioning; customer-selectable LLM (already shipped: Bedrock-Anthropic or Google Gemini) for governance compliance. ICP confidence: High — technical co-founder, Series C, right headcount, strongest agent-reliability quotes in this run.
Marco Badorrek — Director, Solution Engineering EMEA, Parloa
Signal: 4 (Director-level technical leader at ICP agent company; via LinkedIn People directory, Parloa employees). Source: https://www.linkedin.com/in/marcobadorrek. Company: Parloa = AI agent management platform for contact centers; 201-500 emp, Series C. Pain points (inferred from role): deploying/architecting reliable voice AI agents at enterprise customers; per-customer LLM cost; production reliability. Challenges: scaling agent deployments; cost/reliability per customer. Must-haves: production visibility + cost-per-conversation tracking. Nice-to-haves: faster deployment tooling. ICP confidence: Medium (Director-level & technical, but customer-facing Solution Engineering function rather than core product engineering). NOTE: pain points inferred from role/company, not verbatim quotes.
Marija Nakevska — Chief Product & Technology Officer, Pleo
Signal: 1 (public statements by ICP about agent control/transparency) + 4 (ICP shipping a fleet of agents).
Source: https://fintech.global/2026/06/11/pleo-launches-ai-agents-for-autonomous-spend-management/ ; https://www.businesswire.com/news/home/20260611774454/en/ ; https://blog.pleo.io/en/ai-at-pleo
LinkedIn: https://www.linkedin.com/in/marija-nakevska-221a5b164 (probable match, not login-verified — verify before outreach)
Company size: ~949-953 employees (Revelio Labs, Apr & Jun 2026). Copenhagen; European spend management; late-stage/Series C+.
Agent evidence: 5-agent fleet announced 11 Jun 2026 — Policy Agent (LIVE), Pleo MCP server, AP Agent, Treasury Agent, Accounting Agent. Beta from Jul 2026, broad rollout after.
Pain points: QUOTED — "Finance software is shifting from passive dashboards to AI agents that actively work on behalf of users, safely and transparently"; "They want AI working for them, delivering better control... The MCP server foundation... enables advanced AI capabilities while keeping customers in control of their data." Control + transparency named as the gating concern, twice, unprompted.
Challenges: moving 4 more agents from announced to live at ~950-person scale; agents act on real company money, so every action needs an audit trail; agents must escalate only what warrants human attention (their own stated bar) — that is a precision/eval problem.
Must-haves: per-action audit trail and provable control over autonomous spend actions; safe escalation/human-in-the-loop; customer-data isolation.
Nice-to-haves: per-agent/per-run cost attribution as the fleet goes from 1 live to 5 live.
ICP confidence: HIGH — CPTO (CTO-level) at a ~950-person company with a named 5-agent fleet mid-rollout, and she is publicly framing the problem as control, which is our exact wedge.
Caution: Pleo announced layoffs one day after the agent launch (thenextweb.com) — cost pressure is real but the moment may be politically sensitive; do not lead with headcount replacement.
Mario Claudio Martone — Head of Applied Research, EliseAI
Signal: 4 (ICP senior technical leader at a 50-2,000-employee company actively shipping AI agents; surfaced via LinkedIn People search after Signal 1-3 content searches returned mostly non-ICP consultants/vendors). Source: https://www.linkedin.com/in/mcmartone/. Company size: ~400+ (EliseAI; conversational AI agents for housing/healthcare). Pain points (inferred, NOT verbatim): model/agent accuracy & reliability in production; controlling token/compute cost of research-to-production agents; evaluation. Challenges: translating applied research into reliable, cost-efficient production agents. Must-haves: eval + cost/reliability observability. Nice-to-haves: per-run cost attribution. ICP confidence: High (Head of Applied Research = Head-of-AI-type leader at qualifying agent company).
Markel Sanz Ausin — Director of LLM, Hippocratic AI
Signal: 4 (ICP technical AI leader at a company shipping AI agents in production; found via LinkedIn people search "Hippocratic AI engineering leader agents" this run).
Source: https://www.linkedin.com/search/results/people/?keywords=Hippocratic%20AI%20engineering%20leader%20agents — profile summary: "Director of LLM at Hippocratic AI since August 2025, after Staff Research Scientist and Senior Research Scientist roles at Hippocratic AI."
Company size: ~312 employees (Tracxn, May 2026; PitchBook 190, other sources ~264 — all inside the 50–2,000 band). Hippocratic AI = AI-native healthcare company, Series C ($126M at $3.5B valuation, Nov 2025, led by Avenir Growth; $404M total), explicitly expanding patient-facing generative AI agents across healthcare — a genuine multi-agent production shop (Polaris platform, voice agents at call volume).
Pain points: NO public pain post captured this run — do not treat as expressed. Qualification is role + company based.
Challenges (inferred from role scope, flagged as inference): owning LLM quality/behaviour for high-volume patient-facing voice agents implies per-call inference cost pressure, latency budgets, and regression/eval burden as agent count grows.
Must-haves: unverified this run.
Nice-to-haves: unverified this run.
ICP confidence: High — Director-level AI leader (at/above ICP floor), AI-native Series C company, ~312 employees, agents demonstrably in production at scale. Next run: check his recent activity and Hippocratic engineering blog/talks for an actual cost/reliability signal before outreach.
Markus Ring — Director AI Transformation, NiCE Cognigy, NiCE Cognigy
Signal: 4 (ICP senior AI leader at a company shipping AI agents). Source: https://www.linkedin.com/in/ringmarkus (found via NiCE Cognigy People directory, filter 'Director'). Company size: 201-500 employees (verified); Cognigy ships contact-center AI agents. Pain points (inferred from Director AI Transformation remit): operationalizing agents in production for enterprise customers, reliability at scale, cost predictability. Challenges: scaling agent programs while keeping quality and cost under control. Must-haves: production observability and cost-per-run metrics. Nice-to-haves: automated evals. ICP confidence: Medium (Director-level, AI-focused, verified agent company; role partly customer-facing).
Signal: 4 (ICP technical leader at company actively building agentic AI in production). Source: https://www.fiercehealthcare.com/ai-and-machine-learning/nabla-banks-70m-series-c , https://www.nabla.com/blog/70m-series-c , https://tracxn.com/d/companies/nabla . Company size: ~143 employees (Feb 2026), Series C ($70M, total ~$120M, led by HV Capital). Paris-based clinical AI. Pain points: expanding from ambient documentation into a multi-agent clinical model — agents that initiate EHR actions and adapt across care settings, where reliability and correctness are safety-critical; scaling from documentation to many agent types. Challenges: production reliability of agents taking real actions in EHRs; controlling behavior across diverse clinical roles/settings; cost of high-volume clinical LLM calls. Must-haves: reliability + guardrails on agents that take actions, observability into agent decisions. Nice-to-haves: per-run cost visibility, model routing across providers. ICP confidence: High — CTO/co-founder at ~143-person Series C AI-native company explicitly building out multiple production AI agents. (Note: CEO Alex LeBrun now also leads AMI; CTO Martin Raison is the primary technical ICP contact at Nabla. profileUrl is company page.)
Signal: 4 (senior technical leader at agent-native company; found via company-targeted LinkedIn people-search + web verification). Source: https://www.linkedin.com/search/results/people/?keywords=Parloa%20engineering%20AI%20agents (profile https://www.linkedin.com/in/beheim/). Company size: ~300 est. — Parloa, Series C agentic voice-AI CX platform (~$120M raised). Pain points [INFERRED from verified role+company, NOT verbatim]: agent cost blowout scaling live voice agents across many enterprise CX deployments; reliability of real-time voice agents in production; per-call/per-run cost visibility. Challenges: keeping multi-agent voice systems reliable + affordable at enterprise scale (latency vs cost). Must-haves: production reliability; cost-per-run visibility before scaling 1->many. Nice-to-haves: agent observability/eval tooling; model routing. ICP confidence: High (VP Eng, agent-native ~300-emp Series C; NET-NEW person at a previously-covered company).
Signal: 4 (VP Eng at ICP agent company; via LinkedIn People directory). Source: https://www.linkedin.com/in/beheim (Parloa = AI agent management platform for contact centers). Company size: 201-500 (LinkedIn), Series C. Pain points (inferred): scaling reliability of voice AI agents in enterprise contact centers; cost of high-volume LLM calls per conversation; latency/quality tradeoffs. Challenges: reliability at enterprise scale; observability across many concurrent agents. Must-haves: production monitoring + cost control per conversation. Nice-to-haves: model routing / cheaper inference. ICP confidence: High. NOTE: pain points inferred from role/company, not verbatim quotes.
Signal: 1 (ICP writing/speaking publicly about agent cost, token spend and agent observability).
Source: LinkedIn people search (linkedin.com/in/matryer) + Grafana Labs agentic-ops launch coverage — helpnetsecurity.com/2026/07/28/grafana-assistant-ai-capabilities/ and businesswire.com/news/home/20260727216919/en/ (Grafana Labs Ships Six Tools That Power Agentic Operations From Planning to Production).
Company size: ~1,800 employees as of April 2026 (Revelio/jobsbyculture); remote-first, 50+ countries. Note: valuation ~$9B, later-stage than the stated Series A–C band, but headcount is inside the 50–2,000 window.
Pain points: no per-run visibility into where tokens — and therefore dollars — are going across agent activity; agents that "loop through tool calls burning tokens"; agents that pass conventional health checks (usage/latency/errors all green) while still being broken — hallucinating a policy or leaking a credential into a log.
Challenges: conventional observability signals are insufficient for agentic systems; needs token usage AND the conversation itself as first-class telemetry; correlating aggregate spend down to individual agent activity.
Must-haves: per-agent token/cost attribution; drill-down from a "10,000-foot view" of time/tokens/dollars into specific agent runs.
Nice-to-haves: eval and benchmark integration alongside cost telemetry.
ICP confidence: High — Senior Director-level AI leader at a 1,800-person company actively shipping multiple agent products (Grafana Assistant, agentic investigations, six agentic-ops tools shipped July 2026), and personally on record about token-cost visibility.
CAVEAT for outreach: Grafana Labs also ships LLM/agent observability (Grafana Cloud Agent Observability), so there is partial competitive/partner overlap with thealpha.ai — treat as a design-partner or ecosystem conversation rather than a straight displacement pitch.
Matan Grinberg — CEO & Co-founder (technical), Factory AI
Signal: 1 (ICP writing/speaking about agent reliability at scale). Source: https://www.mckinsey.com/capabilities/mckinsey-technology/our-insights/paving-the-road-for-ai-agents-interview-with-factory-ceo-matan-grinberg ; also LangChain Interrupt talk "Building Reliable AI Agents". Company size: ~60-150 (Series C, $150M raise at ~$1.5B valuation, Khosla-led; lean coding-agent startup). Note: Factory CTO Eno Reyes already in brain — this is the technical co-founder/CEO, distinct person. Pain points: coding agents "fail in production not because the models lack capability but because the surrounding enterprise context is too messy for an agent to operate reliably"; directing most new funding at long-horizon agent reliability. Challenges: long-horizon agent reliability, operating agents reliably against messy enterprise context. Must-haves: reliability tooling for long-running agents, control over agent behavior. Nice-to-haves: cost/token efficiency on long agent runs. ICP confidence: Medium-High — technical co-founder and decision-maker at an in-range AI-native company; caveat: CEO rather than pure eng-leadership title.
Matan-Paul Shetrit — Director of Product Management, Writer
Signal: 1 (ICP quoted on agent cost visibility). Source: https://venturebeat.com/orchestration/writer-says-its-new-palmyra-x6-model-cuts-ai-agent-costs-by-52-as-token-spending-surges (Aug 13, 2026). Employer confirmed live on LinkedIn 2026-08-31 — headline reads "Product @ Writer | We are hiring!", San Francisco, ~6K followers (exact profile URL not exposed in search results; left blank rather than guessed). Director-of-Product-Management title comes from the VentureBeat attribution, not from his LinkedIn headline — verify before outreach. Company size: 250+ employees (Writer's own Series C announcement, writer.com/blog/series-c-funding-writer, Nov 2024). Stage: Series C, $200M at $1.9B (Nov 2024). Pain points: Shetrit told VentureBeat the biggest barrier to enterprise AI expansion is NOT model capability but the cost around it. The article frames Writer's new governance tooling — centralized agent-usage visibility, per-workflow analytics, consumption alerts, spending limits — against the problem that nobody in the C-suite knows what the agents are spending. Asked whether spend controls implied surprise bills, he reframed visibility as the thing that lets CIOs/CISOs approve broader adoption. Challenges: per-workflow cost attribution across many agents; making agent spend legible to non-engineering buyers so adoption isn't gated on cost uncertainty. Must-haves: consumption alerts, hard spending limits, per-workflow analytics. Nice-to-haves: chargeback/showback by team, exec-level cost reporting. ICP confidence: Medium (Director level and Series C / 250+ employees both qualify, and the cost-visibility signal is exactly on-thesis — but he is a PRODUCT leader, not engineering, and Writer is already heavily represented in the People Library, so treat as a champion/coach rather than the technical buyer). Pair with Waseem AlShikh (CTO, already in library) as the technical counterpart.
Mateusz Marszałek — Head of Engineering, Tidio
Signal: 4 (ICP senior technical leader at a company shipping AI agents). Found via LinkedIn people search targeting mid-size customer-service AI companies (Tidio ships the "Lyro" AI support agent).
Source: https://www.linkedin.com/search/results/people/?keywords=Tidio%20Lyro%20AI%20engineering
Company size: ~150-200 employees (SMB customer-service platform; Lyro autonomous AI agent in production). Within ICP 50-2,000.
Pain points (inferred from role/company context, not a direct quote): Lyro AI agent handling large volumes of automated conversations across many small-business customers; per-conversation token cost directly affecting unit margins; accuracy/reliability of autonomous resolutions at scale.
Challenges: keeping per-conversation LLM cost low enough for SMB pricing; visibility into cost per agent run across a large customer base.
Must-haves: cost-per-run/observability; reliability controls for autonomous resolutions.
Nice-to-haves: model routing / prompt-cost optimization; margin dashboards per customer.
ICP confidence: Medium-High (Head of Engineering at a mid-size company with an autonomous support agent in production; title is leadership-level).
Matt Ffrench — Co-Founder & CTO, Fyxer AI
Signal: 4 (ICP technical leader building/shipping AI agents). Source: https://uk.linkedin.com/in/matt-ffrench ; https://www.madrona.com/fyxer-ai-productivity-tools-for-email-and-meetings/. Company size: ~126 employees (London; Series B $30M led by Madrona; AI executive-assistant agents for email and meetings; hit $10M ARR in 6 months). Pain points (inferred from role/product; no direct quote captured this run): running email/meeting agents reliably at scale; per-user LLM cost as usage grows; agent output quality and trust. Challenges: scaling agent reliability and unit economics across a fast-growing user base. Must-haves: cost control per user, reliability. Nice-to-haves: agent personalization that improves over time. ICP confidence: High (technical co-founder and CTO at ~126-emp Series B company shipping AI agents).
Signal: 4 — building end-to-end AI agents for accounting/tax/audit; ~30% of Top 25 accounting firms deploy Basis agents to complete complex workflows end-to-end. Co-founder: Mitchell Troyanovsky. Source: https://siliconangle.com/2026/02/24/ai-accounting-startup-basis-secures-100m-1-15b-valuation-firms-adopt-agent-based-workflows/ ; https://www.crunchbase.com/organization/basis-d271. Company size: estimate ~50-150 (NOT officially disclosed; inferred from Series B $100M at $1.15B valuation, Feb 2026; founded 2023; actively expanding headcount). Pain points: reliability/accuracy of agents doing high-stakes, audited financial work end-to-end; controlling agent behavior across many firm deployments. Challenges: agent reliability at scale; cost per workflow/run. Must-haves: reliability + auditability of agent output. Nice-to-haves: cost visibility per engagement. ICP confidence: Medium (technical co-founder/CEO; company size inferred from Series B scale, not officially confirmed — verify headcount before outreach).
Matt Henderson — VP of Research, PolyAI
Signal: 4 (VP Research at an AI-native voice-agent company; found via LinkedIn people search of PolyAI; ex-Google, Apple, Reka).
Source: https://www.linkedin.com/search/results/people/?keywords=PolyAI%20head%20of%20engineering%20director%20AI
Company size: ~250-350 employees; PolyAI, Series D, enterprise voice AI agents. In-band on size; stage past Series C noted.
Pain points (inferred): model/agent quality and reliability tradeoffs vs. inference cost in production.
Challenges: balancing agent capability with cost/latency at scale.
Must-haves: eval/observability into agent behavior and cost.
Nice-to-haves: research tooling for agent trajectories.
ICP confidence: Medium — VP Research (senior technical leader, more research than ops-buyer), agent-native company, in-band size.
Signal: 4 (ICP technical leader at validated target company EliseAI; sourced via LinkedIn People search, run 2026-08-01). Source: https://www.linkedin.com/in/matthew-lowe000/ . Company size: ~250-500 (EliseAI — Series D, property-management AI agents for housing; validated in-range). Pain points (role/company-contextual, not from an individual post this run): scaling conversational property-management agents in production; reliability and cost-per-conversation; agent quality / hallucination control in a regulated housing context. Challenges: moving from 1 -> many agents across leasing/maintenance/payments workflows; consistent behavior at high volume. Must-haves: cost-per-run visibility, reliability monitoring/observability. Nice-to-haves: eval tooling, model routing to cheaper models. ICP confidence: High (Director of Engineering at 50-2,000-emp agent-native company).
Matt Nassr — Head of Global Data Engineering & AI Transformation, Optiver
Signal: 1 & 4 — spoke at AI Engineer World's Fair 2026 side event 'Agents that compound' on building shared foundations for agentic software development (context, evaluation, execution, governance) and scaling beyond a single agent; also featured in Driver.ai Optiver case study on agentic coding.
Source: https://www.ai.engineer/worldsfair/2026 (Optiver 'Agents that compound') ; https://www.driver.ai/blog/optiver-customer-story
Company size: ~1,600 employees
Pain points: getting one agent working is easy but scaling to many is hard; context/token waste ('all manual context efforts are pretty much out of date the moment they shipped'); no shared foundation across a fleet of agents
Challenges: standardizing context, evaluation and governance across a global agentic-development program
Must-haves: shared context/eval/governance layer that works across many agents
Nice-to-haves: automated context maintenance so context isn't stale on ship
ICP confidence: Medium — seniority (Head of Eng/AI) ✓, headcount in 50–2,000 ✓, actively scaling agents ✓; caveat: Optiver is a proprietary trading firm, not a Series A–C SaaS/AI-native company
Signal: Bucket 1 (ICP VP Engineering at agent company) — web research pivot (LinkedIn/Chrome unavailable). Source: https://leadiq.com/c/kognitos/6231381daf7fcf885364cda5/employee-directory ; https://www.kognitos.com/about-us/ ; https://www.kognitos.com/. Company size: ~50-150 employees (Mountain View HQ; eng/CS teams across NA, Europe, India). Kognitos is a "deterministic agentic AI platform" automating business processes via natural-language agents; founded by Binny Gill (ex-Nutanix CTO). Pain points: reliability/determinism of business-process agents in production; cost control and guardrails for autonomous agents; scaling agents across many enterprise processes. Challenges: keeping agent behavior deterministic, auditable and cost-efficient at scale. Must-haves: reliability/guardrails, per-run cost visibility. Nice-to-haves: cross-process agent observability. ICP confidence: Medium — VP of Engineering at a 50-2,000 emp company shipping agentic automation; company size at lower bound of range. (Note: Kognitos' Binny Gill (CEO) and Neeraj Mathur already in library; Strathman is net-new.)
Matthew Payne — VP of Engineering & Head of AI R&D, Domo
Signal: 4 (ICP title at an agent-shipping company), found via LinkedIn people search "VP of Engineering AI agents" rather than via a pain post. WEAK SIGNAL — see below.
Source: https://www.linkedin.com/search/results/people/?keywords=VP%20of%20Engineering%20AI%20agents and https://www.linkedin.com/in/matthewpayne2/recent-activity/all/
Company size: Domo — headcount not confirmed by search this run; publicly understood to be roughly 900–1,000 employees, i.e. inside the 50–2,000 band, but treat as UNVERIFIED and confirm before outreach. Domo is a publicly traded BI/data SaaS, not Series A–C. Agent activity confirmed: Domo shipped Agent Catalyst for building autonomous AI agents with end-to-end workflows, plus DomoGPT, FileSets and a Semantic Layer (domo.com press release; SDxCentral coverage).
Pain points: NONE DIRECTLY EXPRESSED. His own most recent original post is ~1 year old and is about engineering leadership generally, not agents. The only agent-adjacent activity observed is a repost (~4 months old) of a colleague's post: "We just got the Domo MCP setup and connected to Claude ... holy smokes." Do not attribute agent cost or reliability pain to him — it was not stated.
Challenges: Not observed.
Must-haves: Not observed.
Nice-to-haves: Not observed.
ICP confidence: Medium — strong on structural fit (VP of Engineering + Head of AI R&D is squarely in the target title range; company is inside the likely size band and demonstrably shipping agents via Agent Catalyst), but weak on behavioural signal: he is not publicly writing about agent cost, reliability or scaling. Treat as a fit-based lead requiring discovery, not a signal-based warm lead. Verify Domo headcount before working the account.
Signal: 4 (technical co-founder at Series B agent-native company). Attention raised $30M Series B (June 2026) to build "the AI system that runs revenue teams"; deploys an army of AI sales agents; 500+ customers incl. Abridge, Scale, Lovable, Preply, BambooHR.
Source: https://www.globenewswire.com/news-release/2026/06/23/3315763/0/en/Attention-Raises-30M-Series-B-to-Build-the-AI-System-That-Runs-Revenue-Teams-Not-Just-Records-Them.html
Company size: ~94-98 employees (Series B, NYC).
Pain points: scaling many sales agents reliably across CRM/revenue workflows; keeping agent output consistent.
Challenges: moving from recording calls to autonomous, reliable multi-agent action at scale.
Must-haves: reliability/consistency of agents in production; cost control as agent count grows.
Nice-to-haves: observability/visibility across agent runs.
ICP confidence: Medium-High (technical co-founder/CTO at Series B, agent-native company squarely in size range; untapped account).
Max Lowenthal — Director, Agent Product, Decagon
Signal: 4 (ICP product leader for agents at agent-shipping company). Source: https://www.linkedin.com/company/decagon-ai/people/?keywords=director. Company size: 501–1K employees (per Decagon LinkedIn). Company context: Decagon ships customer-experience AI agents (concierge CX) in production for enterprises. Pain points (inferred from role/ICP): owning the agent product means being accountable for agent reliability, resolution quality, and the economics of each agent run at enterprise volume. Challenges: balancing agent autonomy/quality against cost per run and predictability as deployments scale. Must-haves: visibility into what agents cost and do per run; reliability guarantees. Nice-to-haves: tooling to compare agent configurations on cost/quality. ICP confidence: High — Director-level, agent-focused product leadership at a strong ICP agent company in size band.
Maximilian Eber — Co-Founder & Chief Product and Technology Officer (CPTO), Taktile
Signal: 1 (ICP publishing directly about agent cost and reliability). Source: https://labs.taktile.com/ and https://labs.taktile.com/team — named co-author on all three Taktile Labs agent benchmarks; title also confirmed https://londontechweek.com/speakers/maximilian-eber. Company size: 224 (Tracxn, Jun 2026). Stage: Series C — $110M led by Goldman Sachs Growth Equity, 24 Jun 2026 (Balderton, Index, Tiger Global, YC participating). Pain points (first-person, STRONGEST COST SIGNAL OF THIS RUN): Taktile Labs' stated benchmark design principle — "Every benchmark is designed around the KPIs business teams care about most: accuracy, cost per decision, and latency." KYBench (Apr 2026, Eber co-author) evaluates "detection accuracy, evidence quality, reliability across runs, and cost efficiency across eight frontier models and 31 configurations." Research track 05: "choose the right tool for each task while balancing cost, risk, and performance." Research track 03: "Agentic systems based on LLMs are stochastic and hard to inspect." PIBench (Jun 2026) tests prompt-injection resistance across 16 frontier models in agentic underwriting. Challenges: proving agent reliability and unit economics to regulated financial institutions under SR 11-7 model risk management; managing a model x configuration matrix where cost per decision swings widely. Must-haves: per-decision (per-run) cost attribution; run-to-run reliability/variance measurement; auditability and inspectability of stochastic agent behaviour. Nice-to-haves: automatic model routing to hit cost/latency targets; prompt-injection and guardrail telemetry; cost/latency KPIs surfaced to business owners not just engineers. ICP confidence: High — technical co-founder at a 224-person Series C who personally publishes benchmarks whose headline metrics are literally cost-per-decision and reliability-across-runs. TOP PRIORITY for outreach.
Mayank Goyal — Co-Founder & Chief Scientist, Leena AI
Signal: 4 (ICP at company building/shipping enterprise agents; Leena AI ships agentic "AI Colleagues" across ~500 enterprise customers). Source: https://theorg.com/org/leena-ai?person=mayank-goyal and https://leena.ai/about-us . Company size: ~268 employees (YC/Bessemer/Greycroft; ~$40M raised). Role note: Chief Scientist = technical AI decision-maker (Head-of-AI equivalent). Pain points (inferred from public product positioning, not a verbatim quote): making enterprise agents reliable and accurate across IT/HR/Finance/Procurement workflows and many customer environments. Challenges: agent quality/reliability at scale; controlling model behavior and cost across a large deployed base. Must-haves: production reliability/accuracy, control over agent behavior. Nice-to-haves: cost/token visibility per agent run. ICP confidence: Medium-High (technical co-founder & Chief Scientist at an AI-native agent company in the target size band; profile URL is TheOrg rather than a confirmed direct LinkedIn).
Signal: 4 (ICP AI engineering leader at a company shipping agents at scale). Source: LinkedIn people search "Head of AI Engineering agents production cost per run platform" -> profile https://www.linkedin.com/in/mayur-patel2/ ("Head - AI Engineering @ Cognite | Executive leadership | Advisor | Investor", United States). Corroborating: https://www.automation.com/article/cognite-announces-cognite-atlas-ai (Cognite Atlas AI — industrial agent workbench + low-code industrial agent builder extending Cognite Data Fusion), https://www.reveliolabs.com/companies/cognite/employees. Company size: 898 employees worldwide as of March 2026 (Revelio Labs), up 44.4% from 622 in 2023; >$170M revenue FY2025; ~700% YoY increase in Atlas AI customers. Pain points (INFERRED from role + company trajectory, no verbatim quote captured this run): supporting a ~700% YoY jump in agent customers on an industrial agent workbench; cost and reliability of agents that automate workflows and decision support against live operational/industrial data; per-customer and per-run economics as agent volume compounds. Challenges: keeping agent accuracy and reliability high in safety-relevant industrial operations while agent count and customer count scale fast. Must-haves: reliability in production, visibility into agent run cost as customer count multiplies. Nice-to-haves: observability/eval tooling across a multi-tenant agent fleet. ICP confidence: Medium-High (Head of AI Engineering — clear ICP title; 898-emp company squarely in band and demonstrably shipping agents in production; caveat: pain points are inferred, and Cognite is later-stage than the A-C band). NOTE: Geir Engdahl (Co-Founder & CTO, AI) is already in the People Library for Cognite — this is a distinct person, not a duplicate.
Signal: 4 (ICP technical leader building/shipping AI agents). Source: https://www.prnewswire.com/news-releases/jasper-appoints-melody-meckfessel-as-chief-technology-officer-302197647.html ; https://www.jasper.ai/agents ; https://getlatka.com/companies/jasper.ai. Company size: ~300-700 employees (San Francisco/Austin; ~$131M raised, Series A, ~$1.5-1.8B valuation; shipping marketing agents that execute research/setup autonomously). Pain points (inferred from role/product; no direct quote captured this run): moving from content generation to reliable agent EXECUTION of marketing work; agent trust/quality across 900+ enterprise customers; cost/observability of agents at scale. Challenges: making autonomous marketing agents dependable and on-brand; scaling agent infra (she previously ran Google Cloud DevOps/infra as VP Eng). Must-haves: reliability, observability, control of agent output. Nice-to-haves: agents that improve with feedback. ICP confidence: High (CTO / senior technical leader, ex-Google VP of Engineering, at a ~300+ emp company actively shipping AI agents; 50-2000 range). LinkedIn URL not captured this run - do not fabricate.
Michael Bargury — Co-founder & CTO, Zenity
Signal: 3-adjacent / 4 (web-research pivot — Chrome/LinkedIn NOT connected this run). Source: https://zenity.io/authors/michael-bargury ; https://theaiinsider.tech/2026/08/11/zenity-closes-125m-to-secure-the-era-of-1-billion-ai-agents/ . Company size: ~230 employees (~150 in Israel); Series C, $125M (Aug 2026); founded 2021 by ex-Unit 8200 / ex-Microsoft. Focus: security & governance for AI agents (agent-facing platform; builds agentic detection). Note: adjacency caveat — Zenity secures others' agents rather than shipping business agents itself, so cost/reliability pain applies indirectly; strong technical leader at a right-size, agent-native company and a useful ecosystem/competitor-signal contact. Pain points (inferred): visibility and control over large fleets of agents; agent behavior reliability/guardrails. Challenges: governing agents at enterprise scale. Must-haves: agent observability/control. Nice-to-haves: cost attribution per agent. ICP confidence: Medium (Co-founder/CTO at a ~230-employee agent-native company; security/governance adjacency lowers direct-buyer fit).
Michael Bevilacqua — VP, AI Product Management, Adeptia
Signal: 1 (VP of AI Product; quoted on 'the cost of almost right AI agents' in a LinkedIn discussion). Source: https://www.linkedin.com/in/michael-bevilacqua/. Company size: ~208 employees (AI-native data automation platform with workflows for AI agents). Pain points: cost of 'almost right' AI agents; agent accuracy in production data/integration workflows. Challenges: getting agents production-accurate; predictable agent cost. Must-haves: agent accuracy/reliability, cost predictability. Nice-to-haves: natural-language-driven workflow orchestration. ICP confidence: Medium (VP of AI Product at right-sized AI-native company; recently joined; signal via quoted commentary).
Signal: Bucket 1 (ICP writing publicly about agent/LLM cost in production) — STRONGEST cost signal found this run.
Source: https://inworld.ai/resources/host-open-source-llms-production (authored by Michael Ermolenko, CTO & Co-founder) ; https://inworld.ai/resources/what-is-inworld-ai ; https://www.crunchbase.com/organization/inworld-ai
Company size: ~87–107 employees (Crustdata/Growjo/PitchBook, Jun 2026; headcount trending down YoY). ~$120–123M raised across 6 rounds; $56M Series A led by Lightspeed (Aug 2023), investors incl. Microsoft M12, Samsung NEXT, Intel Capital, LG Tech Ventures. Founded 2021 with Ilya Gelfenbeyn and Kylan Gibbs (all ex-API.AI/Google Dialogflow).
Agent activity (verified): Inworld runs realtime voice/character AI agents at consumer scale. Inworld Realtime Inference serves Inworld-optimized Gemma 4, DeepSeek V3.2/V4 and MiniMax-M2.5 on NVIDIA B200s.
Pain points (verbatim-adjacent, from his own published piece): he frames the 2026 production problem as choosing between self-hosting on cloud GPUs, managed inference providers, and routing layers — and states the goal is running open-source LLMs "at consumer-scale cost with realtime latency." That is textbook agent cost blowout + latency tension, written by the CTO himself.
Challenges: consumer-scale unit economics — every user turn is an inference call, so cost per session directly caps the business model; simultaneously bound by realtime latency, which rules out naive cost reductions.
Must-haves: cost-per-session/per-run visibility; model routing and fallback across open-source + hosted models; latency-preserving cost controls.
Nice-to-haves: automated model-tier selection per request; context/token-waste reduction in long-running character sessions.
ICP confidence: High — CTO + technical co-founder, 87–107 emp (in range), agent-native company, and he is actively publishing on exactly the cost/latency/routing problem thealpha.ai addresses. Note: headcount is near the lower bound and declining — verify before heavy investment.
Michael Laccetti — Senior Director, Engineering, Gorgias
Signal: 4 (senior technical leader at a company actively shipping AI agents). Source: https://rocketreach.co/michael-laccetti-email_4848088 | https://neuron.com/profile/2736879053/michael-laccetti | company AI context: https://www.gorgias.com/ai-agent. Company size: ~450–525 employees (verified — Gorgias ~521 as of Mar 2026); Series C ($530M valuation, ~$72M ARR). Gorgias ships an AI Agent automating 60%+ of ecommerce support for 15,000+ brands. Pain points (inferred from Sr Director of Engineering at an agent-native CX company): production reliability of the AI Agent at scale, LLM cost governance as automation volume grows, visibility into failure modes and cost per resolved ticket. Challenges: scaling agent engineering org and infra; balancing automation rate vs. quality/hallucination risk. Must-haves: reliability guardrails, cost-per-run attribution, eval/observability tooling. Nice-to-haves: model routing, prompt/context optimization to cut token waste. ICP confidence: Medium (title = Director+, company squarely in ICP; individual pain inferred from role + company, not quoted). METHOD NOTE: Chrome/LinkedIn NOT connected this run; found via verified web research. LinkedIn URL not directly verified (profile_url is an aggregator listing).
Michael Mac-Vicar — Co-Founder & Chief Technology Officer, Enter (getenter.ai)
Signal: 4 (ICP shipping agents). Source: https://www.riotimesonline.com/enter-brazilian-legal-ai-unicorn-1-2-billion-may-2026/ ; https://www.getenter.ai/en/about-us. Company size: ~100-150 (Series B press coverage May 2026 — growing from ~100 toward 150). Sao Paulo, Brazil. Series B, $1.2B valuation — LatAm's first AI unicorn. Agent evidence: Enter runs a multi-stage pipeline of specialized AI agents that analyze lawsuits, organize evidence, detect fraud, recommend settlements and draft legal defenses — processing 250,000+ legal cases per year for clients including Nubank and Itau. This is a genuine 5+ agent production fleet, not a single copilot. Pain points: running a multi-stage agent pipeline (analysis to evidence to fraud detection to settlement to drafting) reliably at very high case volume in a regulated legal context. Challenges: scaling headcount and agent infrastructure together (100 to 150 employees) while maintaining accuracy on legal outputs that carry real liability. Must-haves: production reliability, auditability and explainability of agent decisions, cost visibility per case processed. Nice-to-haves: faster iteration tooling across the multi-agent legal pipeline. ICP confidence: High — current CTO/co-founder, in-band headcount, clear multi-agent production evidence, Series B stage fits. LatAm is an under-covered geography in the People Library.
Signal: 1/4 (leads engineering on an agentic AI platform; publicly discussing orchestrating AI agents & the shift to agentic software dev — InfoQ podcast on 'factory architects' orchestrating agents).
Source: https://www.turintech.ai/news/former-docker-engineering-leader-joins-turintech-to-help-scale-artemis---its-ai-engineering-platform-for-the-agentic-era ; https://itbrief.asia/story/michael-parker-joins-turintech-to-lead-artemis-ai-expansion
Company size: AMBIGUOUS — sources range ~44 (LinkedIn/Crunchbase) to ~120 (growth estimate); Series A, ~$15M revenue. Borderline vs the 50-employee floor — flagged.
Background: Joined from Docker (dev tooling) Feb 2026 as VP Eng to scale Artemis, TurinTech's AI engineering platform for the agentic era / automated code optimization.
Pain points (inferred from role/company focus): orchestrating/controlling AI agents in software delivery; reliability and cost of agentic dev workflows.
Challenges: scaling an agentic platform; managing agent behavior and cost at volume.
Must-haves: agent orchestration + control, reliability at scale.
Nice-to-haves: cost/perf optimization of agent runs.
ICP confidence: Medium (clear role/technology fit; company size sits on the 50-employee boundary — confirm headcount before prioritizing).
Signal: 4 (ICP writing publicly about building and coding with AI agents). Source: https://www.silenteight.com/, https://tracxn.com/d/companies/silent-eight/, https://www.cnbc.com/2023/09/22/hsbc-backed-startup-is-using-ai-to-help-banks-fight-financial-crime.html — title verified live on LinkedIn 2026-08-30. His LinkedIn headline is itself the pitch: "CTO @ Silent Eight | Responsible GenAI & LLMOps for FinCrime & Risk | 26 yrs building enterprise software that ships". Company size: 108 employees (Tracxn, Apr 2026); Singapore HQ, he is based in Warsaw; $61.2M raised across 5 rounds, Series B $40M. Pain points: the company states it has been "delivering AI Agents to leading global banks in live environments" since 2018; agents execute decisions across sanctions screening, AML, fraud, trade surveillance and due diligence for HSBC, Standard Chartered, AIA and Mashreq. Product branded "Iris 7 — Policy-Bound Agentic AI". He has posted articles titled "The year of AI agents has begun" and "Building the Future of Automated Coding at Silent Eight". Challenges: cost-per-decision and silent model failure in regulated production where a wrong adjudication is a compliance event; he owns both R&D and delivery ops. He has publicly defended explainability: "at Silent Eight, we've ensured our AI is 'white box' – AKA, totally transparent. That allows our clients to easily comply with regulations and be prepared for auditing." Must-haves: auditable, explainable per-run traces; predictable cost per adjudication at bank volume. Nice-to-haves: cross-client cost benchmarking. ICP confidence: High — cleanest fit in this run: CTO, right size and stage, agents in regulated production for named tier-1 banks, and he self-identifies with LLMOps.
Signal: 1 (ICP-adjacent technical leader whose company is publishing directly about agent cost/control; found via web research — LinkedIn/Chrome unavailable this run). Source: https://getzowie.com/blog/the-economics-of-scaling-ai-agents (Zowie, 5 Aug 2026, "Cheaper tokens, bigger bills: the economics of scaling AI agents"); https://getzowie.com/blog/inside-ai-architecture (authored by Partyka); https://www.linkedin.com/posts/zowieai_join-zowie-cto-micha%C5%82-partyka-for-a-deep-activity-7359226608336941059-_89v (Zowie company page: "Join Zowie CTO Michał Partyka for a deep..."); title corroborated by ZoomInfo and theorg.com. Company size: ~110-115 employees (PitchBook 115; Tracxn ~110). Series A (~$19.6M raised, $14M Series A May 2022). AI-native customer-service agent platform; >100M conversations/year; ships Orchestrator, Agent Studio, Supervisor, Tester and Traces — i.e. 5+ agent surfaces in production with enterprise customers (Allianz, Decathlon, InPost, KRUK). Pain points: inference/token spend rising faster than token prices fall ("Every enterprise running agentic AI in production is watching the same number climb... What constantly goes up is the bill"); agentic loops consuming 5-30x the tokens of a single exchange; model routing still bills every decision; long-tail/rare cases are where cost and risk both spike. Challenges: keeping per-decision cost flat while volume scales; preventing hallucinations from reaching production in regulated refund/eligibility/identity workflows; proving cost predictability to enterprise buyers. Must-haves: deterministic control over policy-sensitive decisions, per-run cost visibility, agent tracing/regression testing before release. Nice-to-haves: cross-model portability (customers bring their own frontier or small model), finer-grained routing analytics. ICP confidence: High (CTO at a 110-person Series A AI-native company with many agents in production; company is publishing verbatim on agent cost blowout and control — the exact thealpha.ai wedge).
Signal: 4 (ICP writing/speaking about building agents in production at scale). Source: https://www.youtube.com/watch?v=snroDwX1-JU (Evaluating and improving Replit Agent at scale) + https://claude.com/code-with-claude/session/sf-evaluating-and-improving-replit-agent-at-scale. Company size: ~150-300 (est.; $150M ARR, $3B valuation, $250M raise Sept 2025). Pain points: closing the gap between eval/testing and production performance for Replit Agent; reliability incidents at scale (July 2025 agent deleted a live prod DB and misrepresented actions); running tens of thousands of agents in parallel under 4x load. Challenges: turning non-deterministic agent behavior into reliable production experiences; offline/online eval loop maintenance. Must-haves: production reliability + safeguards, evaluation harness (ViBench), visibility into agent behavior. Nice-to-haves: compounding overnight eval gains, agent-manager orchestration tooling. ICP confidence: High (Head of AI at a mid-size company shipping agents at massive scale; explicit reliability + eval pain).
Signal: 1 (also 4) — writes and teaches (Stanford lecturer; Maven course "AI Software Development: From First Prompt to Production Code") about running agents at scale: multi-agent pipelines, parallel subagent batches, and automated retrospectives.
Source: https://www.linkedin.com/in/mihaileric + https://maven.com/the-modern-software-developer/ai-course
Company size: ~50 employees (Monaco, AI-native revenue/GTM platform; Series A led by Founders Fund → Series B led by Benchmark, $85M+ total).
Pain points: reliability of multi-agent pipelines in production; cost of running parallel subagent batches; getting agents from prototype to production quality.
Challenges: scaling agent reliability while controlling spend.
Must-haves: visibility + control over multi-agent execution and cost.
Nice-to-haves: automated retrospectives / evals.
ICP confidence: High — Head of AI at a ~50-person AI-native Series B actively shipping agents.
Mike Freedman — Co-Founder & CTO, TigerData (TimescaleDB)
Signal: 4 (ICP building explicitly for agent workloads). Source: https://x.com/michaelfreedman/status/1980646019361571049 (his own Agentic Postgres announcement); role confirmed https://www.tigerdata.com/blog/author/mike; product thesis https://www.tigerdata.com/blog/postgres-for-agents. Company size: ~153 (Unify, Jun 2026). Stage: Series C ($110M Tiger Global; $180M+ total, >$1B valuation). Pain points: "Agentic Postgres... a big step toward what we think the future of data infrastructure looks like in the age of AI. Agentic Postgres is the first database built for agents." Product thesis IS the pain — agents want hundreds of throwaway database copies to test against, loop on, and discard, so they built Fluid Storage zero-copy forks, an MCP server, and a terminal-native CLI. Direct response to agent-loop resource blowout. Challenges: serving an "army of AI agents" each wanting isolated disposable state; hybrid keyword+vector search for agent memory in real time; making per-agent DB spin-up cheap enough not to blow the bill (they market "no pay per query or pay per index"). Must-haves: instant cheap isolation per agent run; MCP-native access; agent memory/retrieval co-located with operational data. Nice-to-haves: per-agent-run cost visibility across forked environments; guardrails on agent-initiated resource creation. ICP confidence: High — CTO/co-founder, Series C, 153 people, entire 2025-26 product bet is agent infrastructure. NOTE: cost pain is structural/product-level, not a first-person complaint.
Mike Gozzo — Chief Product & Technology Officer, Ada (ada.cx)
Signal: 4 — Leads product + engineering + AI/ML for Ada's autonomous customer-service AI agent; surfaced via podcast/interview coverage of building 'the world's first autonomous AI agent for customer service.'
Source: https://www.ada.cx/blog/q-and-a-with-mike-gozzo-ada-s-chief-product-officer/ ; theorg.com (CPTO, Ada) ; Crunchbase crunchbase.com/person/michael-gozzo (LinkedIn /in slug not confirmed — left blank rather than guessed)
Company size: ~300–400 (Series C, ~$1.2B val, Toronto)
Pain points: scaling autonomous CS agents reliably across 100s of enterprise customers; agent quality/safety; per-conversation cost & resolution economics
Challenges: keeping autonomous agents accurate & on-policy in production; visibility into what agents actually do
Must-haves: agent observability/evals, reliability guardrails, cost-per-resolution visibility
Nice-to-haves: multi-agent orchestration, model routing
ICP confidence: High — CPTO (owns AI/ML + engineering), Series C, 50–2000 employees, autonomous agents in production
Mike Murchison — CEO & Co-founder, Ada
Signal: 1 (ICP engaging on agent cost/reliability at scale). Source: https://www.ada.cx/blog/q-and-a-with-mike-murchison-ada-s-co-founder-and-ceo/ (plus The Peel / MAD Podcast interviews). Company size: ~300-400 (Toronto Series C, ~$190M raised, ~$1.2B valuation; 350+ enterprise customers). Note: Ada co-founder David Hariri already in brain — Mike Murchison is the distinct CEO/co-founder. Pain points: Ada now "processing 1.5 trillion tokens monthly" powering customer-service agents — massive token volume implies cost-visibility and reliability pressure; frames enterprise-readiness gaps where cost of failure is "failed implementations, wasted investment." Challenges: reliability and ROI at very high token volume, proving agents outperform human teams, enterprise-readiness. Must-haves: cost visibility at scale, reliability guarantees. Nice-to-haves: per-run/per-resolution cost attribution. ICP confidence: Medium — decision-maker co-founder at an in-range company running massive agent token volume; caveat: CEO rather than eng-titled.
Mike Myer — Co-Founder & CEO (technical; ex-CTO RightNow, led Eng through IPO), Quiq
Signal: 4 (ICP technical co-founder at a company shipping AI agents). Method: web/funding research + leadership verification. Source: SiliconANGLE 2026-05-11 'Quiq extends its AI agent platform into voice as enterprise rollouts move past pilots', https://www.quiq.com , https://www.linkedin.com/in/mikemyer. Company size: 107 employees (Mar 2026). Agent activity: Quiq enterprise AI agent platform (chat + newly voice), enterprise rollouts moving past pilots (confirmed shipping in production). Technical founder: software engineer (Bell Labs), led Engineering at RightNow from day 1 through 2004 IPO (Oracle acq. $1.8B), ran Product/Eng at Dataminr. Pain points (INFERRED, not verbatim): moving enterprise agents from pilot to production reliably; extending agents to voice; accuracy + cost at enterprise scale. Challenges: production reliability past the pilot 'last mile'. Must-haves: reliability + cost visibility per agent run. Nice-to-haves: cross-channel (voice) agent consistency. ICP confidence: Medium-High (technical co-founder/CEO, right size 107, agent-native shipping to production).
Signal: 4 (ICP technical decision-maker building & shipping agents in production). Source: https://www.linkedin.com/posts/liming_excited-to-share-that-ive-joined-artisan-activity-7307420792789553152-bkK_ , https://theorg.com/org/artisan-ai-yc-w24/org-chart/ming-li , https://www.artisan.co/blog/artisan-series-a , https://techcrunch.com/2025/04/09/. Company size: ~168 employees (per prior Alpha scan; scaled fast from ~35 humans in Apr 2025), Series A $25M, YC W24, SF. AI-native platform building "Artisans" — human-like autonomous digital workers; ships Ava, an autonomous AI sales rep (outbound SDR agent) in production. Li is CTO (ex-VP of Technology at Deel; led eng at Rippling, TikTok, Google; joined with 4 senior Rippling engineers to scale AI infra & autonomy). Pain points: scaling autonomous outbound agents in production reliably; expanding agent autonomy and product line across verticals; cost/reliability of high-volume autonomous agent runs. Challenges: reliability & quality of autonomous sales agents at scale; building AI infra for growing agent fleet. Must-haves: production reliability, control over autonomous agent behavior, cost visibility. Nice-to-haves: per-run cost/perf analytics. ICP confidence: Medium-High — CTO / clear technical decision-maker at an AI-native company (~168 employees, in band) actively shipping autonomous agents in production.
Signal: 4 (technical leader building agent systems at an agent-native company). Source: LinkedIn people search (https://www.linkedin.com/search/results/people/?keywords=Cresta%20director%20engineering%20agents). Background: ex-Senior Staff Engineer Google, ex-Chief Architect Hello Group, former Founder/CTO — heavyweight leading agent engineering. Company size: ~400 (Cresta; unified platform for human + AI agents for CX; Series D — in 50-2,000 band). Pain points (INFERRED): building/scaling multi-step LLM & agent systems in production; retry/reliability and cost per agent run. Challenges: reliability of multi-step agents, token/cost efficiency. Must-haves: cost-per-run visibility + reliability at scale. Nice-to-haves: model comparison/routing on real workloads. ICP confidence: Medium (agent-native, right-size; title is a functional 'Lead' — seniority flagged, but scope + pedigree are Director-equivalent; stage past Series C).
Signal: 1 (ICP speaking on agent cost & reliability). Source: AI Engineer World's Fair 2026 talks "AI Architects: Tokenmaxxing" and "I Let Agents Refactor My Codebase for 3 Weeks. Then I Read the Code" (https://www.ai.engineer/worldsfair/schedule); LinkedIn profile "Mingsheng Hong — Head/VP of AI, Ironclad" (VP of AI since Jan 2026, ex-Microsoft Azure Data research scientist). Company size: ~600 (legal contract AI, $200M+ ARR; large applied-AI org shipping agents). Pain points: "tokenmaxxing" / runaway token spend; trusting agent output (agents refactoring code that still must be human-reviewed); raw throughput is a misleading metric — argues for "trusted throughput". Challenges: controlling agent cost while scaling; validating agent output quality/reliability in production. Must-haves: agent cost control plus output-trust/reliability signals. Nice-to-haves: throughput metrics that account for correctness. ICP confidence: Medium-High — exact VP/Head of AI title at an active agent-builder; note company is later-stage (beyond Series C) but within 50-2000 headcount. Profile URL not verified to a clean /in/ slug.
Minna Song — Co-Founder & CEO (technical co-founder, MIT CS), EliseAI
Signal: 4 — technical co-founder at a company actively shipping AI agents in production. | Source: https://www.cnbc.com/ (EliseAI / Minna Song, 2026 CNBC Changemaker) and https://www.linkedin.com/company/eliseai | Company size: EliseAI ~300-800 employees (housing + healthcare conversational AI agents; unicorn) — within 50-2,000 ICP band. Confirmed ICP account (brain already holds EliseAI CTO Tony Stoyanov, VP Eng Ryan St Pierre, 3 Directors of Eng, Head of Applied Research). | Pain points: scaling conversational/voice AI agents across housing and healthcare in production; reliability and accuracy in regulated verticals; cost of running agents at high conversation volume. | Challenges: keeping agents reliable and compliant at scale across two regulated verticals. | Must-haves: reliable, accurate production agents; compliance controls. | Nice-to-haves: per-run cost visibility as agent volume grows. | ICP confidence: Medium — genuinely new name (not previously in brain) and technical co-founder (MIT CS, built initial product) at a strong ICP company. Caveat: current function is CEO rather than a core engineering-leadership role, and EliseAI's eng leaders are already captured, so marginal add value; qualifies via the "technical co-founder" C-suite criterion. NOTE: profile URL left blank — not fabricating a LinkedIn URL. Found via web research fallback (LinkedIn browsing unavailable this run).
Mirron Rozanov — Senior Director of Engineering, AI Platform, Gong
Signal: 4 (ICP Sr. Director of Engineering, AI Platform, at an agent-shipping company). Source: https://www.linkedin.com/in/mirronrozanov/ (found via LinkedIn people search 'Gong director engineering AI agents'). Company size: Gong ~1,100-1,300 employees (independent; revenue-intelligence platform now shipping AI agents for revenue teams). Pain points (inferred from role, not a verbatim quote): building/scaling the AI platform powering Gong's agents; reliability and cost of LLM features at scale. Challenges: production reliability, latency, and cost of an AI platform serving many customers/agents. Must-haves: AI-platform observability, cost control, reliability. Nice-to-haves: per-feature / per-agent cost attribution. ICP confidence: High (Sr. Director Eng owning the AI Platform at an in-range, independent agent-shipping company).
Signal: 4 (technical founder building/shipping agents). Source: https://techcrunch.com/2025/10/09/reflection-raises-2b-to-be-americas-open-frontier-ai-lab-challenging-deepseek/ and https://sequoiacap.com/podcast/training-data-misha-laskin/. Company size: 187 employees (May 2026, up from ~60 in Oct 2025). Ex-DeepMind (co-founder Ioannis Antonoglou created AlphaGo). Building Asimov, a code-research agent fusing long-context search, team-wide memory, and multi-agent planning. Pain points: coordinating multi-agent planning reliably, giving agents shared/team-wide memory, context retrieval efficiency ("spend less time spelunking for context"). Challenges: scaling multi-agent systems in constrained-to-open environments. Must-haves: agent memory + context management, multi-agent coordination visibility. Nice-to-haves: cost control on long-context retrieval. ICP confidence: Medium (frontier lab, but actively ships an agent product w/ multi-agent architecture; 187 employees fits size band). Note: personal LinkedIn /in/ URL not confirmed — company page used to avoid fabrication.
Signal: 4 (ICP Director-of-Engineering at a company confirmed to be actively building AI agents). Source: LinkedIn profile above — found via people search scoped to Vic.ai. Company: Vic.ai, autonomous accounting/finance AI agents. Company size: 51-200 employees (confirmed via LinkedIn company page, New York). Pain points (inferred from role + company, NOT a verbatim quote): delivering reliable autonomous-accounting agents in production; managing engineering cost/velocity as agent count grows; observability into agent runs and failures. Challenges: scaling agent infrastructure and reliability with a mid-size eng org; cost control on LLM spend. Must-haves: production reliability, per-run cost/usage visibility, evals. Nice-to-haves: unified agent monitoring & control layer. ICP confidence: Medium-High (Director of Engineering at a 51-200 emp agent-native company; engineering leadership, at/above Director).
Signal: 4 (ICP building & shipping autonomous agents in production). Source: https://openai.com/index/basis/ , https://www.businesswire.com/news/home/20260224020999/ , https://www.linkedin.com/in/mitchelltroyanovsky/ . Company size: ~76 (Series B, $100M at $1.15B val; Accel/Khosla/GV; autonomous accounting agents used end-to-end by ~30% of top-25 US accounting firms across accounting, tax & audit). Pain points: reliability/trust of autonomous agents in high-stakes structured accounting work; turning model progress into trusted production agents; scaling agents across client workflows. Challenges: agent accuracy & auditability; scaling multi-step agent workflows; cost of complex reasoning. Must-haves: reliable/auditable agents; production-grade control. Nice-to-haves: per-workflow cost visibility. ICP confidence: Medium (co-founder w/ deep AI/ML product background; title varies across sources; ~76 emp Series B AI-native shipping agents).
Mo Shaker — Sr. Director of Engineering, Agentic AI, Writer
Signal: 4 (Sr Director Eng explicitly owning Agentic AI at ICP company; via LinkedIn People directory). Source: https://www.linkedin.com/in/mo-shaker-12812230 (Writer = enterprise agentic AI platform). Company size: 201-500, Series C. Pain points (inferred): scaling enterprise agent deployments reliably; agent cost at enterprise volume; token/context efficiency. Challenges: reliability + cost as enterprise agent usage grows. Must-haves: cost-per-run visibility + reliability guardrails. Nice-to-haves: model optimization / cheaper inference. ICP confidence: High (near-perfect title fit). NOTE: pain points inferred from role/company, not verbatim quotes.
Moe Haidar — Head of Agentic AI & Engineering, Nexthink
Signal: 4 (non-founder senior technical leader — "Head of Agentic AI & Engineering" — at a company actively shipping agentic AI; via people search 'Head of Engineering AI agents startup').
Source: https://www.linkedin.com/in/moehaydar/ ; Nexthink (Digital Employee Experience) — agentic AI, analytics & automation; Gartner MQ Leader 2026.
Company size: ~1,160 employees (2026).
Pain points: building/scaling agentic AI that autonomously diagnoses and resolves digital-workplace issues across millions of endpoints; reliability of autonomous agents at very large scale.
Challenges: 0->1 building agentic systems inside a large established platform; large-scale systems/architecture for agents.
Must-haves: reliable agent orchestration, production observability at scale.
Nice-to-haves: cost-per-run/action visibility.
ICP confidence: Medium — Head-of-Agentic-AI-Engineering title and ~1,160-emp size both fit, and Nexthink is actively shipping agentic AI. Caveat: Nexthink is a mature PE-owned (Vista Equity) company, not a Series A-C or AI-native startup, so company-type/stage is a partial ICP mismatch.
Moiz Virani — Co-founder & CTO, Momentum.io
Signal: 4 (ICP technical co-founder/CTO building revenue AI agents). Source: https://www.momentum.io/about ; https://ca.linkedin.com/in/moiz-virani-69a9871a ; LeadIQ (headcount ~132 as of Feb 2026). Company size: ~132 (across 5 continents). Stage: ~$13M raised (Series A range). Actively building agents: Momentum is an AI revenue-orchestration platform that runs agents over every customer interaction (calls/meetings) to produce structured CRM data, summaries, and revenue signals. Pain points (inferred from product + domain, NOT verbatim quotes): reliability/accuracy of agents processing large volumes of customer conversations; cost of running LLM agents over every interaction; visibility into per-run cost and agent output quality. Challenges: scaling agent processing across all customer interactions reliably and affordably. Must-haves: reliability, cost-per-run visibility, accuracy. Nice-to-haves: compounding intelligence across accounts. ICP confidence: Medium (technical co-founder & CTO; ~132 employees fits band; builds agents in production; funding modest/early).
Moritz Kröger — Director, Forward Deployed Engineering, Parloa
Signal: 4 (ICP technical leader found via LinkedIn people search at a named agent company). Source: https://www.linkedin.com/search/results/people/?keywords=Parloa%20director%20engineering%20AI%20agents . Profile: https://www.linkedin.com/in/moritz-kroeger/ . Company size: ~300-450 (Series C $120M; voice AI agents for contact centers). Pain points (inferred from role+company, no direct quote): deploying voice agents into customer production reliably; latency/cost per call as deployments scale. Challenges: bridging pilot->production for enterprise voice agents; reliability and cost governance across customer fleets. Must-haves: production reliability + per-call cost/latency visibility. Nice-to-haves: observability + guardrails to speed forward-deployment. ICP confidence: Medium-High (Director of FDE deploying agents into production at a 50-2,000-emp voice-agent company). Found by scheduled task icp-prospect-signal-scanner run 2026-08-03.
Moritz Kröger — Director, Forward Deployed Engineering, Parloa
Signal: 4 (ICP senior technical leader at a 50-2,000-employee company actively shipping AI agents; surfaced via LinkedIn People search after Signal 1-3 content searches returned mostly non-ICP consultants/vendors). Source: https://www.linkedin.com/in/moritz-kroeger/. Company size: ~400-490 (Parloa; enterprise voice AI agent management platform, voice/chat/messaging). Pain points (inferred, NOT verbatim): forward-deployed enterprise voice-agent rollouts; per-call cost + latency; production reliability/monitoring across channels. Challenges: per-customer customization at scale; keeping agents reliable in production. Must-haves: production monitoring + cost control per deployment. Nice-to-haves: faster deployment tooling. ICP confidence: High (Director FDE; ex-McKinsey).
Moritz Kröger — Director, Forward Deployed Engineering, Parloa
Signal: 4 (Director FDE at ICP agent company; via LinkedIn People directory). Source: https://www.linkedin.com/in/moritz-kroeger (ex-McKinsey). Company size: 201-500, Series C. Pain points (inferred): deploying agents into customer production environments reliably; per-customer cost blowout; debugging agent failures in prod. Challenges: scaling FDE deployments; cost/reliability per customer. Must-haves: production visibility + cost-per-run tracking. Nice-to-haves: faster deployment tooling. ICP confidence: Medium-High (FDE is delivery-focused but core-technical and owns prod agents). NOTE: pain points inferred from role/company, not verbatim quotes.
Signal: 4 (ICP technical co-founder building/shipping agents in production; also Signal 1 — public POV on agentic engineering). Source: https://siliconangle.com/2026/04/14/german-startup-synera-lands-40-million-automate-engineering-workflows-ai-agents/ ; https://www.synera.io/press/synera-raises-40m-series-b-to-scale-agentic-ai-engineering-for-global-manufacturers | Company size: ~102 employees (Tracxn, May 2026); Series B, $40M led by Revaia (Apr 2026), $58.1M total; 60+ enterprise customers across 15 countries (NASA, Airbus, BMW, Volvo, Brose, L'Oréal, Miele, Stihl). Founded 2018 (Bremen/Germany). Platform ("JARVIS for engineers") deploys teams of AI agents that autonomously execute product-development workflows across design/simulation/optimization; integrates 80+ CAD/CAE tools; runs on-prem. | Pain points: (inferred from public positioning) getting agentic pilots into reliable production — cites Gartner that only ~41% of manufacturing AI prototypes reach production; orchestrating multi-agent workflows across fragmented legacy engineering tools; on-prem data-residency constraints. | Challenges: reliability/accuracy of autonomous multi-step engineering agents at enterprise scale; cost of running long, tool-heavy agent workflows. | Must-haves: production reliability + per-run visibility for agent workflows; on-prem/data-residency support. | Nice-to-haves: cost-per-run analytics, cross-department orchestration. | ICP confidence: High — technical co-founder/CEO (PhD) at a 50–2,000-emp company shipping 5+ agents in production; softer only in that role is CEO not CTO, but ICP explicitly allows technical co-founders.
Muayad Sayed Ali — Director of Engineering, Writer
Signal: 4 (company-scoped LinkedIn people search — ICP technical leader at agent-shipping company). Source: https://www.linkedin.com/in/muayad (via https://www.linkedin.com/company/getwriter/people/?keywords=engineering). Headline: 'Director of Engineering at Writer.com | Helping companies leverage generative AI'. Company size: 201-500 employees (Writer; enterprise generative-AI + agentic platform / AI agents, Series C). Pain points (INFERRED from role+company): reliability and cost control of generative-AI/agent features in production; scaling agent workflows for enterprise customers. Challenges: shipping reliable, cost-efficient agentic features across enterprise deployments. Must-haves: reliability guardrails + per-run cost visibility. Nice-to-haves: token/model optimization. ICP confidence: High (Director of Engineering at an independent 201-500-emp agent-shipping SaaS; net-new alongside existing Writer entries Waseem AlShikh and Dan Bikel).
Muayad Sayed Ali — Director of Engineering, Writer
Signal: 4 (Director of Engineering at ICP agent company; via LinkedIn People directory). Source: https://www.linkedin.com/in/muayad (Writer.com = enterprise generative/agentic AI). Company size: 201-500, Series C. Pain points (inferred): engineering reliability + scaling of the agent product; infra cost of LLM workloads; visibility into per-feature/agent cost. Challenges: scaling eng org + infra cost efficiency. Must-haves: cost observability + reliability. Nice-to-haves: automated cost optimization. ICP confidence: High. NOTE: pain points inferred from role/company, not verbatim quotes.
Muddu Sudhakar — Co-Founder & CEO (technical; PhD; serial founder, prior exits to Splunk/EMC), Aisera
Signal: 1 & 4 — agentic AI platform running autonomous agents across IT/ops/CX in production; frequent public keynotes/interviews on agentic AI at scale. Source: https://www.crunchbase.com/organization/aisera ; https://aisera.com/company/. Company size: ~311 employees; $174M+ raised. Pain points: production economics of multi-step agentic loops (5-30x token multiplier vs chatbots); reliability at enterprise scale; visibility/control of cost across many production agents. Challenges: cost observability and governance across a large fleet of agents; agent reliability. Must-haves: cost + performance observability for agents. Nice-to-haves: model routing/optimization. ICP confidence: Medium (311 emp, actively runs agents in prod; NOTE: Series D — slightly beyond stated Series A–C stage, flagged for review).
Mukund Jha — Co-founder & CEO (prev Co-founder/CTO at Dunzo; ex-Google engineer), Emergent
Signal: 1 (agent cost/scaling — autonomous coding/"vibe coding" agents; found via web research, LinkedIn/Chrome unavailable this run; pain inferred from company positioning, no personal quote). Source: https://techcrunch.com/2026/07/15/indian-ai-coding-startup-emergent-becomes-a-unicorn-just-over-a-year-after-launch/ ; https://www.dealstreetasia.com/stories/emergent-unicorn-489228 . Company size: ~275 employees (Bengaluru + SF HQ). Funding: Series C $130M at $1.5B valuation (Jul 2026); earlier $70M (SoftBank, Khosla). Product: platform that builds software (CRMs, ERPs, apps, internal tools) via autonomous AI agents. Pain points (inferred): compute/token cost of autonomous multi-step coding agents; reliability of long agent runs; scaling many agents in production. Challenges: cost + reliability as agent workloads scale. Must-haves: cost visibility per run, reliability. Nice-to-haves: observability, control. ICP confidence: High (technical co-founder/CEO, 275 emp, Series C, coding agents). NOTE: brother & CTO Madhav Jha already in People Library — Mukund is net-new.
Munil Shah — Chief Product, Technology and Customer Officer (CPTO), Talkdesk
Signal: 4 (ICP author writing about agent evaluation/reliability at scale). Surfaced via Yunjing Ma's activity feed — she reposted him, so this is also a Signal 2/4 crossover: ICP author + ICP engagement from a VP at the same company.
Source: https://www.linkedin.com/in/yunjing-ma-99637819/recent-activity/all/ — original post by Munil Shah ~3 months old, linking to talkdesk.com "AI Agent Evaluation: From Signal to Fix | Talkdesk" (102 reactions, 6 comments, 6 reposts).
Company size: ~1,300–1,400 employees (Revelio/LeadIQ/Unify report 1,373–1,400 in 2026, down from ~1,700 in 2024). Inside the 50–2,000 band. CCaaS / Customer Experience Automation platform shipping AI agents in production. Note: Talkdesk is late-stage ($10B valuation, well past Series C) — it fits the size and agent-shipping gates but not the Series A–C gate.
Pain points (verbatim from his post): "One of the biggest gaps we see in the market today is that teams are deploying AI agents without a rigorous way to evaluate behavior, reliability, tool usage, and customer outcomes at scale."
Challenges: Evaluating agent behavior, reliability and tool usage at scale; connecting agent-level signals to customer outcomes; moving from detected signal to actual fix (framing of the linked piece, "From Signal to Fix").
Must-haves: A rigorous, scalable way to evaluate deployed agent behavior — explicitly named as the market's biggest gap.
Nice-to-haves: Not separately expressed.
ICP confidence: High on role and agent activity — CPTO is a C-suite technical/product owner, company is inside the size band and demonstrably shipping agents, and the pain is stated in his own words. Medium caveat on fit: cost/token-spend was not the pain he raised (his angle is evaluation and reliability, not spend), and the company is later-stage than the stated Series A–C preference.
Note: colleague Yunjing Ma (VP of AI Engineering, Research & Applied AI, Talkdesk) is already in the brain from the 2026-07-30 run — same account, do not double-count.
Signal: 1 (ICP writing/speaking about cost, safety and reliability of production voice agents). Source: https://hippocraticai.com/munjal-shah/ ; https://www.crunchbase.com/person/munjal-shah . Company size: ~150-300 (raised $278M total; backers incl. General Catalyst, a16z, Kleiner Perkins). Pain points: lowering healthcare cost via patient-facing voice agents; safety, clinical accuracy and reliability of real-world voice interactions (Polaris 5.0); scaling non-diagnostic agent services under strict safety/regulatory constraints. Challenges: guaranteeing safe, accurate, empathetic agent behavior in regulated healthcare settings at scale. Must-haves: safety/reliability, clinical accuracy, cost-effective scalable inference. Nice-to-haves: empathy/naturalness, regulatory compliance tooling. ICP confidence: High (technical serial-founder CEO shipping healthcare voice agents in production; well-funded, target size band). Note: LinkedIn profile URL not confirmed, left blank.
Mustafa Ahamed — VP, Product Management, Aisera
Signal: 4 (VP of Product, AI-focused, at agentic-AI company; found via LinkedIn people-search 'Aisera agentic AI'). Source: https://www.linkedin.com/in/mustafaahamed/ (headline: VP Product Management - Aisera). Company size: 201-500 (LinkedIn-verified). Building agents: yes - Aisera ships agentic AI for enterprise service/IT. Pain points (inferred from role/company, not a quoted post): scaling many production agents across enterprise workflows; cost/ROI visibility per agent run; reliability of enterprise agent deployments. Challenges: proving agent ROI and controlling cost at enterprise scale. Must-haves: per-agent cost + reliability visibility. Nice-to-haves: governance across a growing fleet of agents. ICP confidence: Medium (VP Product-AI at 201-500-emp agentic-AI company; pain inferred from role).
Nagasai Pallapotu — Director of Technology, Kore.ai
Signal: 4 (ICP Director-level technical leader explicitly building agentic platforms). Source: targeted ICP search of Kore.ai; profile https://www.linkedin.com/in/nagasai-pallapotu-9856a038/ (headline 'Director of Technology at Kore.ai - Building Agentic platforms with voice Agents, multi agents, conversational AI, bots, LLMs, contact centre automations'). Company size: ~1,000 (Kore.ai, enterprise agentic AI platform). Pain points [INFERRED from his own headline + role]: building/scaling voice + multi-agent platforms in production; reliability across many enterprise contact-center deployments; cost/latency of multi-agent orchestration. Challenges: multi-agent orchestration reliability; contact-center automation at scale. Must-haves: per-run visibility + reliability for multi-agent systems. Nice-to-haves: agent cost analytics. ICP confidence: Medium-High (Director-level technical leader explicitly building agentic/multi-agent platforms at a ~1,000-emp agent company).
Nanda Kumar Kante — Associate Vice President – Technology (CoE Head, Conversational AI Bots Design & Development), Kore.ai
Signal: Signal 4 (LinkedIn people search at named agent company; Director→VP level. Sourced by role/company match, not observed posting)
Source: https://www.linkedin.com/search/results/people/?keywords=Yellow.ai%20VP%20engineering%20OR%20head%20of%20AI
Profile: https://www.linkedin.com/in/nandakumarkante/
Company size: ~1,000-1,200 (Kore.ai, enterprise agentic AI platform, Series-stage, actively shipping agents)
Pain points: (inferred, not directly observed) Scaling conversational bots into agentic workflows across an enterprise CoE; keeping reliability and cost predictable as agent count grows
Challenges: (inferred) Governance and per-run cost visibility across many enterprise bot/agent deployments
Must-haves: (inferred) Cost + reliability visibility per agent/run; guardrails at scale
Nice-to-haves: (inferred) Unified observability across the agent fleet
ICP confidence: Medium-High (AVP = above Director; Kore.ai is a confirmed in-ICP agent company in the 50-2,000 band; multi-threaded account)
Nanda Santhana — Co-Founder & CEO (technical), DataBahn.ai
Signal: 4 (web-research proxy — LinkedIn/Chrome unavailable this run; found via Jul 2026 Series B announcement + company profile). Source: https://www.databahn.ai/press-releases/databahn-raises-40-million-series-b-led-by-insight-partners ; headcount via https://theorg.com/org/databahn-ai . Company size: ~85 employees (2026); Series B, $40M led by Insight Partners; founded 2023. Technical co-founder: ex Oracle Tech Fellow, founding member & SVP at Securonix, MS CS (USC).
Pain points (from company positioning): enterprise telemetry/data volume and cost blowing up as it feeds AI/agent-driven operations; pipelines need to be automated and reliable ("agentic data control plane").
Challenges: controlling data volume + cost feeding AI, keeping pipelines reliable, giving teams visibility.
Must-haves: cost control + reliability for data feeding agents; observability.
Nice-to-haves: automation of pipeline ops.
ICP confidence: Medium-High — technical co-founder/CEO at confirmed 50-2,000-emp company; agentic data control plane maps to cost/observability pain, though DataBahn is more agent-infra than an agent-app shipper.
Natalie Meurer — Head of Agent Engineering, Sierra
Signal: 4 (ICP publicly writing/speaking about operating agents in production; engaged with other ICP leaders in the agent-ops discourse at AIEWF 2026). NOTE: LinkedIn/Claude-in-Chrome NOT connected this run — sourced via web research + primary-source verification.
Source: https://www.latent.space/p/forward-deployed-engineers-aiewf (Latent Space, Jul 1 2026); AIEWF 2026 talk "The Dirty Secret of Forward Deployed Engineering" (https://www.youtube.com/watch?v=Byv311hdoHE); https://sierra.ai/author/natalie-meurer.
Company size: Sierra ~690-855 employees (Revelio Labs Dec 2025: 690; Tracxn Jun 2026: 855). She leads a global team of 120+ engineers building conversational AI agents for enterprise customer service.
Pain points: enterprise customers cannot keep track of what their agent estate actually does — "Every enterprise we work with wants to know how it can maintain everything its agentic ecosystem is capable of doing"; orchestration sprawl — "It needs to manage all the integrations and all the teams that contribute to the agent."
Challenges: scaling agent engineering across 120+ engineers and many enterprise deployments; keeping ownership/visibility as the number of integrations and contributing teams grows; the FDE label has become so stretched it "defies coherent definition" (org/role clarity problem).
Must-haves: a coherent operating layer showing what every deployed agent does, costs and touches; maintainability of a growing agent ecosystem across many teams.
Nice-to-haves: standardised agent engineering practice/tooling so each customer deployment isn't bespoke.
ICP confidence: High — Head-of-Engineering-level leader (120+ engineers) at an AI-native company with many agents in production; headcount squarely in band; pain is agent estate visibility and maintainability.
Signal: 4 (ICP publicly presenting about building/deploying agents in production). Source: https://www.zenml.io/llmops-database/evolution-of-forward-deployed-engineering-and-agent-engineering-in-the-ai-era (2026 talk, video https://www.youtube.com/watch?v=Byv311hdoHE). Company size: ~500-600 (Sierra Technologies; $15B valuation, ~$950M Series E May 2026; enterprise CX AI agents adopted by ~half the Fortune 50). Pain points: deploying agents into production and guaranteeing measurable outcomes under outcome-based pricing (completed sales, resolved inquiries); production reliability/accountability shifting from uptime to business outcomes; cost optimization of agent runs (talk explicitly tagged cost_optimization). Challenges: ensuring reliable outcome delivery per run; monitoring agents against outcome metrics and iterating fast; emerging specialization ("harness engineering"). Must-haves: production monitoring tied to outcomes, per-run cost/observability, rapid iteration. Nice-to-haves: platform/self-serve enablement so more engineers can deploy/manage agents. ICP confidence: High (Head of Agent Engineering — senior technical AI leader — at a 50-2,000-emp company definitively shipping agents in production). LinkedIn URL not captured this run — do not fabricate.
Natalie Rutgers — SVP of Product, Deepgram
Signal: 1/4 (senior product leader at an agent-active company; surfaced via LinkedIn people/leadership search this run). Source: https://www.linkedin.com/in/natalierutgers/ and https://deepgram.com/company/leadership. Company size: ~250-300 (Deepgram, voice AI platform; Series B; ships Voice Agent API + powers 200k+ devs building production voice agents). Pain points [INFERRED from role+company, not a verbatim quote]: unit economics of voice agents (cost per call/session); making agent behavior reliable and measurable for customers; packaging cost/latency transparency into the product. Challenges: giving customers visibility/control over what each agent interaction costs; reliability at production scale. Must-haves: per-session cost + reliability metrics customers can trust. Nice-to-haves: cost-per-completed-task benchmarking across models. ICP confidence: Medium (SVP Product, AI-focused, at an agent-active 50-2000-emp company; product not eng, and Deepgram is voice-AI infra).
Signal: 4 (ICP technical co-founder shipping an agentic platform in production). Source: https://www.prnewswire.com/news-releases/sight-machine-launches-agentic-manufacturing-platform-that-understands-your-plant--and-improves-it-every-run-302797542.html ; https://www.sightmachine.com/news-agentic-manufacturing-platform ; https://sightmachine.com/company ; headcount 66 (LeadIQ, Feb/Mar 2026; ~66 across 4 continents), Crunchbase bracket 51-100, LinkedIn bracket 51-200 — NOTE Tracxn showed 43 as of Dec 2025, so the company crossed the 50-employee ICP floor during 2026. LinkedIn: https://www.linkedin.com/in/nateoostendorp/. Company size: ~66 (2026) — LOWEST IN THE BRAIN'S ACCEPTABLE BAND, re-verify before outreach. $85.5M raised over 6 rounds. Vertical: industrial AI / manufacturing operations. Agent evidence: agentic manufacturing platform launched 11 Jun 2026; the press release headline itself is 'improves it, EVERY RUN', and the product has AI agents working directly with process experts to connect data, build the Semantic Model and improve performance run over run; integrates with Azure, Microsoft Fabric, Teams, NVIDIA Omniverse and Databricks. Note the company also has a Chief AI Officer & co-founder, Kurt DeMaagd — a second, possibly better-targeted entry point at the same company. Pain points (INFERRED from public product/positioning, not verbatim quotes): their whole claim is that agents get better with every run, which is a compounding-intelligence claim — meaning they must measure what each run did and cost in order to substantiate it; agents reason over high-cardinality plant signal data, which is a context/token-waste problem by construction. Challenges (INFERRED): they compressed work that 'takes months of effort by data engineers' into days of agent work, so agent runs are now doing expensive engineering labour and the cost per run is being compared against a consulting alternative; small engineering org (~66 people) with no room to build an internal harness. Must-haves (INFERRED): per-run cost and context visibility to back the 'improves every run' claim; reliability on plant data where a wrong agent conclusion changes production. Nice-to-haves (INFERRED): run-over-run comparison tooling. ICP confidence: MEDIUM — role and agent evidence are strong and the compounding-intelligence framing is the closest public match to thealpha.ai's own thesis found this run, but headcount (~66) sits just above the floor and moved up from 43 within the last two quarters, so size is the reason this is not HIGH. Also worth noting: Oostendorp co-founded Slashdot and was site architect at SourceForge — a genuinely technical, open-source-native buyer who will evaluate a harness on architecture, not on sales narrative.
Neal Lathia — Co-founder & CTO (ex-Monzo, built ML infrastructure), Gradient Labs
Signal: 4 (ICP building/shipping specialist agents in production). Source: https://gradient-labs.ai/about; Wikipedia/Sifted coverage. Company size: "more than 40 employees" — BORDERLINE at the 50-employee floor; fast-growing AI-native fintech agent company (~$42.6M raised, 32M+ end users, customers incl. Wise, Zego, Plum, Current, Stash, Pockit). CTO/technical decision-maker responsible for the multi-agent architecture powering regulated financial-services support. Pain points: engineering reliability + strict compliance in long-running multi-agent workflows; latency/accuracy tradeoffs; controlling cost per resolution. Challenges: production reliability where a wrong answer is unacceptable (finance); shared context/memory across specialist agents. Must-haves: reliability/observability, compliance/audit trails, cost control per run. Nice-to-haves: multi-agent context sharing. ICP confidence: Medium — technical decision-maker, excellent pain fit; flag headcount near the 50 floor (verify before prioritizing).
Nebojša Miletić — VP Engineering (Chief of Staff to the CPTO), Parloa
Signal: Bucket 4 (ICP leader building/shipping agents at a qualifying company). NOTE: LinkedIn/Chrome not connected this run; found via web research + verification.
Source: https://theorg.com/org/parloa/org-chart/nebojsa-miletic ; https://theorg.com/org/parloa/teams/leadership-team-1 ; corroborated rocketreach + LeadDev. VP Engineering & Chief of Staff to the CPTO at Parloa (Berlin); prior Director of Payments & Fraud Tech / Principal Engineer at Wayfair. Non-founder (founders: Malte Kosub CEO, Stefan Ostwald CTO — both already in library).
Company size: ~300–450 employees (Series C, $120M raise).
Ships agents: Parloa builds enterprise voice/conversational AI agents deployed in production contact centers; heavy agentic-engineering internal practice ("Claude Kitchen").
Pain points: voice-agent latency + reliability at contact-center scale; per-conversation/token cost; multi-region scaling of many concurrent agents.
Challenges: maintaining reliability and cost efficiency as agent volume and languages scale.
Must-haves: reliability guardrails and per-conversation cost visibility for production voice agents.
Nice-to-haves: unified observability across the agent estate.
ICP confidence: Medium-High — non-founder VP Engineering verified on multiple sources; qualifying voice-agent company in size range.
Signal: 4 (senior ICP AI leader at an agent-native company; surfaced via LinkedIn people search on agentic-automation companies during ICP company review). Source: https://www.linkedin.com/in/mathur-n/ (company: https://www.linkedin.com/company/kognitos/about/). Company size: 51-200 (verified on LinkedIn). Company: Kognitos — agentic process automation for large enterprises ('deterministic AI / English as Code'), actively shipping enterprise AI agents. Pain points: making probabilistic LLM agents reliable & deterministic in production; controlling/ auditing agent execution for large enterprises; cost & governance of enterprise automation agents. Challenges: reliability + auditability of agents at enterprise scale; scaling agent deployments beyond pilots. Must-haves: deterministic/observable agent execution, governance & guardrails, enterprise-grade reliability. Nice-to-haves: cost-per-run visibility, multi-agent orchestration. ICP confidence: High (C-suite technical AI leader — Chief AI Officer, ex-UiPath/VMware/Goldman — at a 51-200 agentic-automation company shipping agents).
Neha Gupta — Senior Director of AI, Uniphore
Signal: 1 (ICP AI leader at an agent-shipping company; found via web research — LinkedIn/Chrome unavailable this run). Source: https://www.uniphore.com/blog/meet-neha-gupta-director-of-ai/ ; https://theorg.com/org/uniphore/teams/artificial-intelligence-division. Company size: ~700–1,100 employees (within 50–2,000 band); Uniphore ships agentic AI / conversational AI agents in production for enterprise CX and automation. Neha Gupta is Senior Director of AI (previously leadership roles at LinkedIn, Uber, Adobe). Pain points (INFERRED from role/company context — NOT verbatim): productionizing AI agents reliably, managing model/token cost at enterprise scale, quality + cost visibility across agent workflows. Challenges: scaling AI agent quality and cost-efficiency across large enterprise deployments. Must-haves: production observability + cost control per agent. Nice-to-haves: routing/optimization. ICP confidence: Medium (clean "Senior Director of AI" ICP title at a 50–2,000-emp agent company; caveat: Uniphore is late-stage/Series E-plus rather than Series A–C). NOTE: pain points inferred, not quoted.
Niall O'Higgins — Director of AppSec, SRE & Infrastructure, Replit
Signal: 4 (ICP technical leader at agent-native company building secure agent platforms). Source: https://www.linkedin.com/in/niallohiggins/ (LinkedIn, 2026); title per theorg.com (Director AppSec/SRE/Infra, Oct 2025). Company size: ~200-400 emp (Replit; ships Replit Agent). Pain points: security/reliability of AI-generated code at scale; prompt injection at scale; abuse across millions of daily agent actions; sandbox/isolation for untrusted agent code. Challenges: guardrails for non-deterministic agents; defending an agent platform where 'the adversary is AI'; infra + SRE cost of massive agent volume. Must-haves: reliability guardrails + observability for agents in production, secure agent-native infra. Nice-to-haves: automated defense tooling, cost-efficient sandboxing. ICP confidence: High (Director-level infra/security leader at agent-native company; leads internal AI, cloud infra, SRE; verified current).
Nicholas Arcolano — Head of AI & Research, Jellyfish
Signal: 1 (ICP writing about agent costs/reliability on LinkedIn). Source: https://www.linkedin.com/in/arcolano — LinkedIn posts "AI Token Costs: Uneven and Rising for Developers", "20 million PRs later: The truth about AI productivity"; speaker at AI Engineer World's Fair 2026. Company size: ~250 (Series C engineering-management SaaS, Boston, building & measuring AI agents for software engineering). Pain points: per-developer token consumption up ~18.6x in 9 months; massive agent parallelism inflating spend; ~37% gap between lab benchmark and production agent performance; can't tie agent spend to business value ("Whether extreme spend pays off comes down to the ultimate business value of shipped code, which most companies still can't measure"). Challenges: measuring ROI of AI/agent investments; per-agent-run cost attribution; separating productive vs wasteful token use. Must-haves: per-agent/per-run cost attribution tied to outcomes. Nice-to-haves: agent productivity-vs-spend benchmarking. ICP confidence: High — exact Head of AI title, builds/measures agents, repeatedly public on cost & reliability pain.
Signal: 2/4 (ICP engaging with agent-development content on a practitioner channel — LangChain's Max Agency podcast with Harrison Chase — talking directly about agent reliability, evals and multi-model spend). NOTE: LinkedIn/Claude-in-Chrome NOT connected this run — sourced via web research + primary-source verification.
Source: https://www.langchain.com/blog/benchling-max-agency-podcast (Jun 11 2026, Max Agency podcast with LangChain CEO Harrison Chase); https://decodingbio.substack.com/p/scaling-bio-009-benchlings-ashu-singhal; speaker listing https://www.syntheticbiologysummit.com/2026/speakers/nicholas-larus-stone.
Company size: Benchling ~793-884 employees (PitchBook 2026: 811; Revelio Labs Dec 2025: 884). Launched Benchling AI (agent-backed intelligence layer) Oct 2025; he leads the company's scientific agents. Joined via Benchling's acquisition of Sphinx Bio, which he founded.
Pain points: runs the SAME task across MULTIPLE model providers to cross-check answers — a direct multiplier on inference spend with no per-run cost framing discussed ("Each of them will make slightly different errors... being able to ask different model providers, we found gives us much better performance"); evals alone are insufficient for scientific work, so reliability depends on manual production trace review; context engineering trade-offs (SQL vs file-based harnesses).
Challenges: evaluating non-verifiable tasks where clean benchmarks don't exist; a weekly rotating "fire chief" plus PM/engineer trace review as the primary reliability mechanism — human-expensive and unscalable; multi-agent architecture and agent-managed memory/skills.
Must-haves: automated detection of failure modes in production traces (currently human labour); a defensible way to justify multi-provider redundancy cost.
Nice-to-haves: per-run cost attribution across model providers; better context/memory management for long scientific workflows.
ICP confidence: High — Head of AI at a company in the 50-2,000 band with agents shipped in production; pain points map directly to cost multiplication and no automated visibility into agent behaviour.
Signal: 4 (technical co-founder of AI-native company shipping an agent platform). Source: https://betakit.com/cohere-pitches-security-and-productivity-with-general-release-of-north-enterprise-ai-platform/ ; https://en.wikipedia.org/wiki/Cohere ; https://www.thetwentyminutevc.com/nick-frosst. Company size: ~350-500 (Cohere; enterprise/sovereign AI; ships North, a platform where enterprises build/run custom AI agents). Pain points: making enterprise agents secure, reliable, and cost-efficient in production; sovereign/on-prem deployment constraints. Challenges: competing on efficiency vs larger labs; production reliability and cost control for customer-deployed agents. Must-haves: reliability, security/governance, cost efficiency. Nice-to-haves: model-level token efficiency. ICP confidence: Medium (technical co-founder at 50-2,000-emp AI-native company shipping an agent platform; note Cohere is model-lab-adjacent rather than a pure vertical agent deployer, so slightly softer fit). LinkedIn URL not captured this run — do not fabricate.
Signal: 4 (technical co-founder talking publicly about building and shipping agents).
Source: podcast "Nico Laqua & Emily Yuan, Corgi Founders on AI Native Insurance" — https://open.spotify.com/episode/7jmLtsCrUunLXXZzf9TWdB
Company size: ~250 employees. AI-native full-stack insurance carrier; $108M combined seed + Series A at a ~$630M valuation.
Pain points: building toward agents that handle underwriting, claims and pricing end-to-end rather than routing to outsourced human claims teams; his stated end state is "AI agents talk directly to other AI agents ... with zero human latency" — i.e. agent-to-agent volume with no human in the loop to catch cost or correctness drift.
Challenges: compute-speed underwriting and claims automation while remaining licensed and auditable across US states; agent decisions must be defensible to insurance regulators; as a carrier, every agent run is a direct loss-ratio input.
Must-haves: auditable, regulator-defensible agent decision trails; reliability at claims volume.
Nice-to-haves: agent-to-agent transaction capability; per-policy cost attribution.
ICP confidence: Medium — headcount and funding verified and in band, company is AI-native and unambiguously shipping agents. Downgraded from High for two honest reasons: (1) the title is CEO rather than a CTO/VP Eng/Head of AI title, though he is the technical co-founder and the library already carries several technical-CEO records on the same basis; (2) the pain signals come from a podcast summary, and I could not verify a per-run cost or reliability complaint in his own words.
LinkedIn URL: not found in any source actually read this run — left blank rather than guessed.
Nicolae Rusan — Co-Founder, Clay (clay.com)
Signal: 4 — Technical co-founder of Clay, which ships Claygent AI research agents for GTM at scale.
Source: https://research.contrary.com/company/clay ; https://www.clay.com/claygent ; Crunchbase crunchbase.com/person/nicolae-rusan (LinkedIn /in slug not confirmed)
Company size: ~1,167 (Series C, $100M Aug 2025, $3.1B val)
Pain points: running research agents (Claygent) cost-effectively across a very large user base; reliability/accuracy of agent outputs; token spend at scale
Challenges: agent output quality & cost control at scale
Must-haves: cost/token visibility, agent reliability/eval
Nice-to-haves: orchestration
ICP confidence: Medium — technical co-founder at a Series C agent-shipping company; exact current technical title not fully confirmed
Signal: 1 (technical ICP at AI-native company shipping agents in production). Source: https://www.ambiencehealthcare.com/blog/ambience-healthcare-announces-nikhil-buduma-as-new-chief-executive-officer ; Modern Healthcare "How Abridge, Hippocratic AI, Ambience are building nursing tools" (2026). Company size: ~150–300 employees (est.), ~$1B valuation, AI-native healthcare; platform live across outpatient, ED, and inpatient settings and 100+ specialties. Pain points: production-grade reliability of clinical agents in regulated environments; coding-aware documentation agents at scale; expanding agents upstream into revenue-cycle management (RCM). Challenges: reliability/accuracy of clinical agents across specialties and health systems; compliance in regulated clinical settings; scaling multi-step clinical agents. Must-haves: production reliability + accuracy in clinical/regulated environments; control over agent behavior. Nice-to-haves: cost efficiency of multi-step clinical agents. ICP confidence: High — technical co-founder and CEO (deep-learning researcher, authored "Fundamentals of Deep Learning") at an AI-native company shipping clinical agents in production. Profile URL left blank (not fabricating LinkedIn URL).
Nikhil Cheerla — Co-founder & CTO, Nooks
Signal: 4 (ICP building/shipping agents in production). Source: https://www.geekwire.com/2026/the-rise-of-vertical-ai-agents-and-the-startups-racing-to-build-them/ | Company size: ~400 employees (2026), Series B ($43M, Kleiner Perkins); AI-native sales platform serving 1,000+ companies. Pain points: has "injected [agents] into almost every part of the stack" and now runs many domain-specific agents end-to-end (identify accounts, find contacts, draft emails, assist on live calls); wrestling with data/context quality — says incumbents "don't know what's good and what's bad in that data," so Nooks is architected to "collect very high quality data" to capture the full context behind each agent decision. Challenges: scaling many vertical agents across the product while keeping outputs high-quality; capturing decision context. Must-haves: high-quality context/data feeding agents; visibility into agent behavior at scale. Nice-to-haves: agent-to-agent collaboration, background/bulk agent orchestration. ICP confidence: High (technical co-founder/CTO at a 50–2,000-employee AI-native company actively running 5+ agents in production).
Signal: 4 (ICP is a senior technical leader at a company shipping AI agents at scale). NOTE ON METHOD: LinkedIn browsing was unavailable this run (Claude-in-Chrome extension not connected), so this person was found and verified via web research rather than the prescribed LinkedIn post/people searches. Pain points below are INFERRED from Vapi's public product/domain, not verbatim quotes from the individual.
Source: https://techcrunch.com/2026/05/12/vapi-hits-500m-valuation-as-amazon-ring-chose-its-ai-platform-over-40-rivals/ ; https://www.globenewswire.com/news-release/2026/05/12/3292882/0/en/vapi-raises-50m-series-b-as-it-reaches-1-billion-calls-powering-the-next-generation-of-enterprise-voice-ai.html ; https://vapi.ai/library/building-voice-ai-with-nikhil-gupta-of-vapiai
Company size: ~100-150 (Series B, $50M raised May 2026, ~$500M valuation; 1M+ developers, 2.7M+ agents created, 1B+ calls).
Pain points (inferred): per-call/per-run cost economics across millions of concurrent voice agents; latency and reliability at massive scale; visibility into what each agent run costs.
Challenges (inferred): keeping voice-agent reliability high while scaling call volume 1->1B+; controlling inference spend as agent count compounds.
Must-haves (inferred): granular cost-per-run and reliability observability for agents in production.
Nice-to-haves (inferred): cost attribution per customer/agent, token-waste reduction.
ICP confidence: High — technical co-founder & CTO of a ~100+ person Series B voice-AI-agent platform (well within 50-2,000 emp, actively shipping agents in production at scale).
Nikhil Mungel — Head of AI R&D, Cribl
Signal: 1 (ICP publicly discussing agent cost, compounding spend and control). LinkedIn profile verified live 2026-09-02: headline reads exactly "Head of AI R&D at Cribl", San Francisco.
Source: https://redmonk.com/videos/nikhil-mungel/ (RedMonk MonkCast episode with Kate Holterhoff)
Company size: ~1,396-1,458 employees (Revelio Labs / trackers, 2026) — inside the 50-2,000 band. Cribl = data/observability pipeline platform, now shipping AI copilot and agent-driven features. Stage note: last primary round was a Series D (2022), ~$3.5B valuation, so later-stage than the Series A-C target.
Pain points: Agent-generated telemetry (logs, traces, tool-call chains) piling up "at a volume nobody's pipeline was built for"; token prices falling while total spend climbs; users defaulting to the priciest model; cost visibility disconnected from actual agent behaviour; confusion between credit-based and outcome-based billing.
Challenges: No standard schema for agent telemetry; attributing spend to a specific agent run rather than an aggregate bill; pipelines designed for machine logs now absorbing agent reasoning traces.
Must-haves: Agent-specific telemetry standards (he points at OCSF-style schemas), cost-per-agent-run attribution.
Nice-to-haves: Standardised efficiency metrics for "token factory" throughput; guardrails that stop default-to-frontier-model behaviour.
ICP confidence: Medium — title and pain signal are strong and on-message, headcount fits the band, but Cribl is late-stage and is fundamentally an observability/data-pipeline vendor rather than a company whose core product is customer-facing agents. That makes him a credible buyer for internal agent spend, and simultaneously a partial competitor — worth flagging to whoever runs outreach.
Signal: 1 (ICP writing about agent reliability/eval). Source: https://www.artificiallawyer.com/2026/05/06/harvey-launches-legal-agent-bench/ and https://www.harvey.ai/company. Company size: ~340-500 (AI-native legal agents; scaling agents across law firms/enterprises, raised at $11B). Pain points: measuring and improving agent reliability in production — Harvey built "Legal Agent Bench" to benchmark how well legal agents actually perform. Challenges: proving agents are reliable/accurate enough for high-stakes legal work; closing the eval gap between demo and production. Must-haves: rigorous agent evaluation/benchmarking, reliability guarantees. Nice-to-haves: cost/latency visibility per agent task. ICP confidence: Medium (applied-research leader driving agent reliability at an AI-native company in size band; research role rather than pure eng decision-maker, and company is later-stage). LinkedIn slug not confirmed — profile_url blank to avoid fabrication.
Nikola Mrksic — Co-Founder & CEO, PolyAI
Signal: 4 (ICP senior technical leader at a company actively shipping AI agents; found via LinkedIn people-search, not an observed post). Source: https://www.linkedin.com/in/nikola-mrksic/ | Company size: ~200-300 (Series C; enterprise voice AI agents). Pain points (INFERRED from role + company product focus — no observed quote): scaling agents reliably in production; controlling per-run LLM/agent cost as volume grows; visibility into what agents cost and how they behave across customers. Challenges: keeping agent accuracy/guardrails consistent at scale; token/context efficiency; cost blowout as they move from pilots to broad production. Must-haves: production reliability + observability into agent cost and behavior. Nice-to-haves: per-run cost attribution, context/token optimization, spend guardrails. ICP confidence: High (Technical co-founder (Cambridge NLP/dialogue PhD); headline explicitly 'AI agents for enterprises').
Nikola Mrkšić — Co-Founder & CEO, PolyAI
Signal: 4 (ICP building/shipping agents in production). Source: https://www.linkedin.com/in/nikola-mrksic/ and https://app.dealroom.co/news/note/ (PolyAI scaling enterprise voice agents to $50M ARR). Company size: ~372 employees, AI-native enterprise voice-agent company; 100M+ people interact with its agents yearly, deployments equivalent to 1,000+ FTEs per client. Technical co-founder — PhD, Cambridge Machine Intelligence Lab. NOTE: PolyAI closed a Series D (Dec 2025), slightly beyond the stated Series A-C band, but headcount (372) and AI-native agent-shipping profile fit the ICP. Pain points (inferred from role + massive agent volume, not a verbatim quote): per-conversation/token cost at 100M+ interactions, voice-agent reliability and containment, controlling agent behavior at scale. Challenges: keeping unit economics healthy as volume compounds. Must-haves: cost-per-run visibility, reliability at scale. Nice-to-haves: cross-deployment observability. ICP confidence: Medium — strong role/company fit, one stage past stated funding band.
Signal: 1 (ICP speaking/writing about scaling reliable enterprise voice agents). Source: https://app.dealroom.co/news/note/polyai-ceo-nikola-mrk-i-on-scaling-enterprise-voice-agents-from-science-fiction-sales-to-50m-arr-and-the-future-of-ai-contact-centers . Company size: ~250-300 employees (Series C; ~$50M ARR; 100M+ interactions/yr). Pain points: minimizing hallucinations in long multi-turn voice conversations; reducing ASR word-error-rate in contact-center calls; reliability/accuracy of voice agents at scale. Challenges: building a proprietary reliable voice stack (ASR + conversational LLMs) across 45 languages; latency and accuracy in live calls. Must-haves: production reliability, low hallucination rate, accurate speech recognition. Nice-to-haves: multilingual coverage, voice customization. ICP confidence: High (technical co-founder/CEO, ex-Apple/VocalIQ, Cambridge PhD, shipping voice agents in production at target size).
Nikos Alexakis — Director of Software Engineering (Founding Team), Aisera
Signal: 4 (ICP senior technical leader at a 50-2,000-employee company actively shipping AI agents; surfaced via LinkedIn People search after Signal 1-3 content searches returned mostly non-ICP consultants/vendors). Source: https://www.linkedin.com/in/nikosalexakis/. Company size: ~250-340 (Aisera; agentic AI platform automating IT/HR/customer service/ops). Pain points (inferred from role+company, NOT verbatim quotes): controlling LLM/agent spend across high-volume automated resolutions; agent accuracy & reliability at enterprise scale; multi-domain multi-agent orchestration. Challenges: scaling a founding-team-built platform to enterprise multi-agent workloads; guardrails/quality. Must-haves: production reliability + per-run cost visibility. Nice-to-haves: unified observability across the agent fleet. ICP confidence: High (Director of SW Eng + founding team at an agentic-AI company squarely in range).
Signal: 4 (ICP AI leader at an agent-shipping company; company-scoped LinkedIn people search "Innovaccer director AI agents").
Source: https://www.linkedin.com/search/results/people/?keywords=Innovaccer%20director%20AI%20agents (profile https://www.linkedin.com/in/nilavghosh/)
Company size: Innovaccer ~1,500-1,800 employees (under 2,000 ceiling). Healthcare data+AI platform now shipping agentic AI ("Agents of Care" / healthcare AI agents). NOTE: later-stage (Series F/unicorn) than the A-C target guideline.
Pain points [INFERRED from role + company, NOT a verbatim post]: reliability/governance of healthcare agents in production; cost-per-run at enterprise data volumes; scaling from few to many agents across care workflows.
Challenges: safety, governance and reliability requirements are high for clinical/healthcare agents; cost control at scale.
Must-haves: production reliability + governance layer for agents; per-run cost visibility.
Nice-to-haves: verification/eval tooling for agent outputs.
ICP confidence: Medium — Senior Director of AI at an agent-shipping company that fits the size band but is later-stage/larger than the core Series A-C profile (flagged).
Nimish Hathalia — Senior Director of Engineering, Aisera
Signal: 4 (company-scoped LinkedIn people search of agent-native company Aisera). Source: https://www.linkedin.com/search/results/people/?keywords=Aisera%20director%20engineering | Company size: ~300-500 (est; Series D). Pain points (INFERRED from role+company): scaling reliable enterprise agentic-AI in production; cost blowout across many agents; visibility into per-run/per-agent cost. Challenges: 1->many-agent scaling reliability. Must-haves: production reliability + governance; cost observability per run. Nice-to-haves: context/token-waste reduction. ICP confidence: High (Senior Director of Engineering at agent-native 50-2,000-emp company; confirmed via explicit '@ Aisera Inc.' headline; net-new).
Nirmal Mukhi — VP / Head of Engineering & Chief Architect, ASAPP
Signal: 1 (ICP senior technical leader at an AI-native company shipping agents in production; found via web research — LinkedIn/Chrome not connected this run). Source: https://www.asapp.com/author/nirmal-mukhi ; https://www.zoominfo.com/p/Nirmal-Mukhi/8952605835 (listed as VP Head of Engineering) ; https://www.asapp.com/press/asapp-unveils-generativeagent-tm-a-generative-ai-application-capable-of-fully-automating-contact-center-interactions . Company size: ~389 employees (Mar 2026); ASAPP is an AI-native contact-center CX platform, ~$400M raised, ~$1.6B valuation, shipping GenerativeAgent (positioned to automate up to ~99% of contact-center interactions across voice/chat). Pain points: production reliability of autonomous agents on live customer interactions; quality drift and the last-mile from pilot to production-grade automation; per-interaction cost/observability at very high volume. Challenges: making agents reliable, governable and explainable in regulated/enterprise contact centers; scaling ML infra behind many concurrent agent runs. Must-haves: reliability, governance/control, per-run cost visibility. Nice-to-haves: token/context efficiency, compounding quality across runs. ICP confidence: High (senior technical AI leader — VP/Head of Engineering & Chief Architect — at a ~389-emp AI-native company actively shipping agents in production; solidly in the 50–2,000 band).
Signal: 4 (ICP technical co-founder whose entire product is a customisable autonomous agent fleet). Source: https://raft.ai/ ; https://www.raft.ai/resources/press-releases/raft-raises-30m-in-series-b-funding-to-transform-global-supply-chain-execution-with-ai ; https://tracxn.com/d/companies/raft (134 employees as of 30 Jun 2026) ; https://www.indexbox.io/blog/freight-forwarding-at-a-crossroads-why-agentic-ai-demands-a-growth-strategy-not-just-cost-cuts/. LinkedIn: https://uk.linkedin.com/in/nisargam (note: a second profile https://www.linkedin.com/in/nisarg-mehta-01881987/ also surfaces — /in/nisargam is the one his own posts publish under; confirm before outreach). Company size: 134 (Tracxn, 30 Jun 2026). Stage: Series B, $30M led by Eight Roads with Bessemer, Episode 1, Dynamo Ventures. London; founded 2017 as Vector.ai, rebranded to Raft. Agent evidence: Raft markets itself as letting freight forwarders and customs brokers 'launch intelligent, fully customizable AI agents that work autonomously' across the whole shipment lifecycle — i.e. every customer spins up their OWN agents on Raft's platform, so agent count scales with the customer base, not with Raft's roadmap. Pain points (INFERRED from public product/positioning, not verbatim quotes): document-heavy customs and freight-audit workflows are token-expensive per shipment, and Raft's customers are in a margin-compressed industry, so cost per agent run maps straight onto Raft's own gross margin; a customer-configurable agent platform means Raft cannot predict or bound what its users' agents will do or spend. Challenges (INFERRED): multi-tenant agent governance — attributing cost and failures to the right customer's agents; customs is a regulated, auditable domain so every autonomous action needs a defensible trail. Must-haves (INFERRED): per-tenant, per-agent run cost attribution and hard spend limits; audit trail for autonomous customs actions. Nice-to-haves (INFERRED): benchmarking agent configurations across tenants to steer customers off pathological setups. ICP confidence: HIGH — technical co-founder/CTO, Cambridge MEng, ex-Microsoft/Entrepreneur First, in-range headcount, Series B, and the cost/reliability pain is structural to the business model rather than incidental. He also speaks publicly on AI in logistics (WESCCON, NBCBA), so there is a warm content surface.
Signal: 4 (ICP technical founder publicly discussing making AI reliable in business-critical production settings). Source: https://www.klarity.ai/post/klarity-raises-70-million-in-series-b ; podcast: https://open.spotify.com/episode/1LSWE4jP2ZXB1fFkV6bfc9
Company size: ~130-162 employees at Series B (2024); Tracxn lists 51-200 as of Jul 2025 — in range. Series B, $70M (total $90M). AI document review / accounting workflow agents.
Pain points: Reliability of AI in business-critical settings is his stated working problem — "working on challenges of making AI technology reliable in a business-critical setting." Customers include OpenAI, Zoom, Cloudflare, Intercom, so error tolerance is near-zero and every agent output is auditable by a finance/accounting team.
Challenges: Document-review agents must be deterministic enough for revenue recognition and audit; scaling agent throughput across enterprise customers without per-customer quality regressions.
Must-haves: Reliability guarantees / accuracy measurement in production; ability to catch and explain agent errors before they reach an auditor.
Nice-to-haves: Per-customer cost visibility as agent volume scales; faster eval loops when swapping models.
ICP confidence: High — verified co-founder/CTO, headcount and Series B stage squarely in range, agents in production at named enterprise logos, and an on-record statement of production-reliability pain.
Signal: 3 (ICP engaging with competitor content — he is the named customer in a Helicone case study). Title verified live on LinkedIn people search 2026-09-02: "Current: Sr. Director of AI at QA Wolf", Los Angeles CA.
Source: https://www.helicone.ai/blog/langchain-qawolf . LinkedIn URL href-verified from search results.
Company size: 51-200 (QA Wolf LinkedIn company page, read live 2026-09-02); ~248 per Tracxn (Apr 2026). Seattle, Series B ($36M, Jul 2024).
Pain points: Publicly stated that LangChain's built-in tooling was "limiting" once they needed custom visualizations, alerting and anomaly detection, and a usable UX for large inputs — so they bolted Helicone on as the telemetry layer. QA Wolf runs Playwright test-generation agents at very high volume, so trace payloads are huge and per-run economics matter.
Challenges: agent telemetry at high run volume; traces with very large inputs/outputs; anomaly detection on cost and failure; framework tooling that does not scale past the demo.
Must-haves: per-run trace + cost attribution that survives large payloads; alerting and anomaly detection on spend and failure rate.
Nice-to-haves: usable UX over huge inputs; framework-agnostic instrumentation so they are not locked to LangChain.
ICP confidence: High — Sr. Director (Director+), 51-200/~248 employees (in band), Series B, US, and a documented, quoted observability pain on production agents. Already evaluating/using a direct competitor (Helicone) = active budget and an incumbent to displace. Net-new name (QA Wolf already in library).
Signal: 4/1 (ICP technical/product founder building & shipping autonomous supply-chain agents; found via funding-announcement primary sources, not LinkedIn — Chrome/LinkedIn unavailable this run). Source: https://news.crunchbase.com/transportation/freehand-pando-enterprise-supply-chain-spend-management-startup/ and https://www.globenewswire.com/news-release/2026/07/29/3335085/0/en/Freehand-Raises-75M-to-Scale-AI-Teams-Managing-Supply-Chain-Spend-for-Fortune-500-Companies.html. Company size: ~200 (confirmed, July 2026). Stage: Series B ($75M, co-led Battery Ventures + NewRoad, July 2026; $100M total). Agent-native (founded 2024); AI agents run supply-chain spend for Meta, Unilever, J&J, Pfizer, Cardinal Health, Dunkin' — 5+ agents in production. Role note: repeat founder who previously built the Pando supply-chain platform; CEO but product/technical founder (co-founder Abhijeet Manohar already in the People Library). Pain points: reliability/trust of autonomous procurement & spend agents at Fortune 500 scale; unifying unstructured docs + structured ERP data (their "Category Context Graph"); procure-to-pay accuracy. Challenges: making multi-step agents trustworthy in high-stakes spend; scaling across many large enterprises. Must-haves: reliable multi-step agents, audit/controls, accuracy. Nice-to-haves: per-run cost/token visibility. ICP confidence: Medium-High (technical co-founder/CEO; company clearly shipping agents in production; role slightly business-leaning).
Nitzan Bar — Chief Technology Officer, Hyro
Signal: 4 (ICP building/shipping AI agents in production). Source: https://www.hyro.ai/blog/hyro-ai-agents-for-healthcare-raise-45-million/ ; https://www.mobihealthnews.com/news/hyro-raises-45m-ai-agent-platform-healthcare ; https://tracxn.com/d/companies/hyro/ . Company size: ~107 employees (NYC/Tel Aviv; Cornell Tech spinout 2018; Series B $50M + $45M growth round Oct 2025, $95M total). Ships responsible AI voice & chat agents for healthcare systems (call-center automation, patient access) at enterprise scale. CEO Israel Krush, COO Rom Cohen; Nitzan Bar is CTO. Pain points (INFERRED): reliability/safety of patient-facing agents in a regulated setting; cost of running voice+chat agents across large health-system call volume; visibility into what each agent run costs and does. Challenges (INFERRED): scaling agent deployments across many hospitals without accuracy/cost regressions. Must-haves (INFERRED): reliable, controllable agents + per-run cost visibility. Nice-to-haves (INFERRED): fleet observability across voice/chat agents. ICP confidence: Medium-High (named CTO, ~107 emp in-band, Series B/growth stage, agents in production; recent growth round nudges stage past classic C). LinkedIn URL not surfaced — not fabricated.
Niyati Chhaya — Co-Founder & VP-AI, Hyperbots
Signal: 1 (ICP speaking publicly about agent reliability and accuracy in production — YourStory interview).
Source: https://yourstory.com/2025/12/new-york-bengaluru-startup-hyperbots-agentic-ai-native-co-pilots-automate-core-finance-workflows
Company size: ~115 employees (range cited 101–250). Hyperbots — HQ New York, engineering centre in Bengaluru, India. Series A, $6.5M (May 2025), ~$8.5–9M total raised. Agentic AI-native co-pilots automating finance/accounting workflows (procure-to-pay, order-to-cash).
Pain points: (1) single-model approaches fail on messy real-world inputs — "You're dealing with math, tables, handwritten data and implicit accounting rules all at once. That combination is still largely unsolved with single-model approaches"; (2) sustaining accuracy consistently in production under resource constraints — "Ensuring we could maintain that level of accuracy consistently, even with limited resources, has been the most demanding part of the journey"; (3) agents write back into ERP systems, so errors are financially consequential.
Challenges: orchestrating a mixture-of-experts pipeline of specialised models into one coherent agentic flow; maintaining a claimed 99.8% production accuracy as volume grows; doing this on a small Series A budget.
Must-haves: consistent, measurable accuracy across a multi-model agent pipeline; full audit trails when agents write to source systems; reliability guarantees under cost constraints.
Nice-to-haves: judgment-level reasoning beyond deterministic workflow automation.
ICP confidence: Medium-High — VP-AI is a clean ICP title, technical co-founder, headcount and Series A stage in range, AI-native agent company. Caveat: pain is framed around accuracy/reliability rather than cost specifically; HQ is US with India engineering, so geography is mixed. LinkedIn URL not found in search results — not guessed.
Noa Flaherty — Co-Founder & CTO, Vellum
Signal: 1/3 (writes about rigor, reliability & evaluation for enterprise AI and agent workflows). Source: https://www.businesswire.com/news/home/20250710009580/en/Vellum-Raises-$20M-Series-A-to-Bring-Rigor-Speed-and-Reliability-to-Enterprise-AI-Development ; https://www.vellum.ai/team . Company size: ~40-60 (Series A $20M 2025, ~$25.5M total; YC W23). Pain points: reliability/rigor in agent workflows, evaluation of non-deterministic behavior, controlling agent outputs, cost of iteration. Challenges: bringing rigor + speed + reliability to enterprise agent development. Must-haves: evals + reliability + observability for agent workflows. Nice-to-haves: faster iteration/deployment. ICP confidence: Medium (Series A AI-native agent dev/eval platform; competitor-adjacent).
Oege de Moor — Founder & CEO (technical; creator of GitHub Copilot, ex-Semmle founder, ex-Oxford professor), XBOW
Signal: 1 (ICP-technical founder writing publicly about autonomous agents operating in live production; posts e.g. "We build XBOW because we must" and on agents in production). Source: https://www.linkedin.com/in/oegedemoor , https://xbow.com/leadership , and https://sequoiacap.com/founder/oege-de-moor/ . Company size: ~190 employees (founded Jan 2024; ~$237M raised incl. $120M Series C). Pain points (grounded in public statements/product): running autonomous AI security agents continuously and safely in live production environments at scale (reached #1 on HackerOne, thousands of verified vulns). Challenges: making autonomous agents operate reliably/safely at machine speed across enterprise production apps; scaling agent runs. Must-haves: production reliability and safety, scale of concurrent agent runs. Nice-to-haves: cost/observability per agent run. ICP confidence: High (deeply technical founder/CEO at an AI-native company shipping agents in production, in size band; strong signal-1 public writing about production agents).
Signal: 4 (ICP publicly writing about scaling agents in production). NOTE: LinkedIn/Chrome not connected this run. LinkedIn URL not verified.
Source: https://torq.io/news/torq-seriesd/ (Series D announcement, Jan 2026)
Company size: 350+ employees, hiring toward ~550 by end of 2026 (company Series D press release). Series D, $140M at $1.2B valuation; $332M total. Israel/US. Agentic SOC automation (HyperSOC), Fortune 100/500 customers.
What he said (verbatim, short): self-service agent builder "enables teams to manage 100X more alerts."
Pain points: alert-fatigue at scale; going from a handful of automations to enterprise-wide autonomous agent coverage; maintaining trust as agent count grows.
Challenges: scaling from human-triaged SOC to agent-run SOC without losing auditability; letting non-engineers build agents safely.
Must-haves: self-service agent builder, triage automation, agent memory/context graph.
Nice-to-haves: cross-agent context sharing; deeper per-agent economics.
ICP confidence: Medium — strong 1→many agent scaling signal and headcount in band, but Series D is later-stage than the target A-C band and the title is CEO (co-founder, security-technical background) rather than a pure engineering/AI leadership title.
Ofir Goren Bar — Chief Technology Officer, aiOla
Signal: 3 (surfaced via live LinkedIn people search on aiOla while working the groundcover competitor-content thread). Source: LinkedIn people search read live 2026-08-31 — "Ofir Goren Bar · CTO at aiOla · Israel" (exact profile URL not exposed in search results; left blank rather than guessed). Supporting company-level evidence: https://www.groundcover.com/customer-stories/aiola. Company size: ~70 per the groundcover case study; 51–200 per Startup Nation Finder / Crunchbase-derived. Stage: Series A2, $25M (Jul 2025), $58M total. Pain points: company-level, per the groundcover case study — aiOla had no tracing or APM before adopting observability tooling, no visibility into how/when/at what scale its voice agents are invoked inside customer applications, and rising data-volume and egress costs at scale. IMPORTANT CAVEAT: these pain points are documented from the company case study and from his VP Platform Engineering (Roy Sela, id 1059), NOT authored or quoted by Goren Bar personally — no first-person signal from him was found this run. Challenges: real-time voice-agent SLAs in third-party environments; instrumenting agent workflows across many services without R&D friction. Must-haves: agent-workflow observability, per-run visibility, data residency. Nice-to-haves: unified infra + app observability, non-ingestion-based pricing. ICP confidence: Medium (CTO title and employer verified live on LinkedIn, company is a genuine production-agent company at Series A in the size band — downgraded from High only because the pain signal is company/colleague-sourced rather than first-person). Pair with Roy Sela (id 1059) — same account, Sela carries the first-person quote.
Oleg Poleshuk — Director of Engineering, AIOps, SoundHound AI
Signal: 4 (senior technical leader; AIOps focus is directly adjacent to agent cost/observability). Source: https://www.linkedin.com/in/olegpoleshuk/ (LinkedIn people search "SoundHound AI director engineering agents"). Company size: ~750-950 (SoundHound AI). Ships agents: YES (Enterprise AI / agentic). Pain points (inferred from AIOps role): observability, cost, and reliability of AI systems/agents in production; no single pane for what agents cost per run. Challenges: instrumenting agent ops at scale. Must-haves: per-run cost + trace visibility. Nice-to-haves: automated anomaly/cost alerting for agents. ICP confidence: Medium (title + company confirmed; AIOps directly on-theme; no individual post observed).
Signal: Targeted ICP people-search — agent-native company (AI SDR / autonomous sales agents) absent from brain. Source: https://www.linkedin.com/in/oleg-zaremba/ (LinkedIn people search). Company size: ~50-120 (YC, Series A). Pain points (ICP-inferred; no direct post captured this run): LLM cost blowout as outbound volume scales; token waste in per-lead personalization; agent output quality/reliability at high concurrency. Challenges: controlling per-message LLM spend while preserving deliverability/quality; running many concurrent sales agents reliably. Must-haves: cost-per-run visibility, reliability at volume, model routing. Nice-to-haves: eval/observability tooling. ICP confidence: High (CTO & co-founder, agent-native, YC).
Signal: 2 (non-ICP-authored podcast content on agent dev/costs where this ICP was the guest). Source: https://www.signalcast.app/episode/eye-on-ai/318-olek-paraska-how-ai-is-fixing-the-biggest-bottleneck-in-construction ; https://theorg.com/org/togal-ai/org-chart/olek-paraska. Company size: ~59-65 (2026 headcount trackers) — at the LOW end of the 50-2,000 band. LIVE LINKEDIN VERIFIED this run — "Oleksandr Paraska — CTO @ Togal.ai — AI-powered construction takeoff & estimating · CV/ML · agentic AI · shoniko.com", Bonn Germany. Note his own LinkedIn headline explicitly says "agentic AI". Agent evidence: on the Eye on AI podcast he describes AI agents handling RFI generation, floor-plan version comparison and spec-document parsing — a perception + reasoning agent architecture in construction estimating. Pain points: reliability of agents reasoning over unstructured architectural drawings; cost of running agentic pipelines per takeoff/estimate. Challenges: small engineering team scaling agent coverage across every step of the estimating workflow. Must-haves: agent reliability tooling, cost-per-task visibility given a thin engineering team. Nice-to-haves: eval/observability for mixed computer-vision + LLM agent pipelines. ICP confidence: Medium — role, company and agent evidence all solid and live-verified, but FLAG: funding is a convertible note/SAFE (~$17.5M total raised), not a clean Series A-C round, and headcount sits at the very bottom of the band.
Omid Nejati — Senior Director, Platform Engineering, Hippocratic AI
Signal: Bucket 4 (ICP leader building/shipping agents at a qualifying company). NOTE: LinkedIn/Chrome not connected this run; found via web research + verification.
Source: https://theorg.com/org/hippocratic-ai-1/org-chart/omid-nejati ; https://www.zoominfo.com/p/Omid-Nejati/2359843556 ; https://www.linkedin.com/in/omidnejati/ . Senior Director, Platform Engineering at Hippocratic AI (San Francisco); AI & API management in HIPAA-compliant environments; prior Sr. Manager Eng @ Tonal, Kubernetes/Cloud @ PlayStation. Non-founder (founder/CEO Munjal Shah — already in library).
Company size: ~250–350 employees (Series B/C, ~$1.6B valuation).
Ships agents: Hippocratic AI builds generative voice AI healthcare agents on the Polaris model family, deployed for patient-facing tasks.
Pain points: token/compute cost of running large agent fleets; reliability/safety of agents in a regulated healthcare setting; scaling the agent platform; compliance-grade observability.
Challenges: keeping the platform reliable and cost-efficient while agents scale in a safety-critical domain.
Must-haves: platform-level cost visibility and reliability/guardrail controls for agents.
Nice-to-haves: centralized observability/governance across the agent platform.
ICP confidence: Medium-High — Senior Director (platform) verified across sources; platform lead is the ideal buyer for an agent operating layer; AI-native voice-agent company in size range.
Signal: 4 (ICP building/shipping AI agents in production). Source: https://research.contrary.com/company/instabase ; https://tracxn.com/d/companies/instabase/ ; https://docs.instabase.com/welcome . Company size: ~165 employees (SF; founded 2015; AI Hub platform). Ships secure, federated AI agents for document/content understanding and workflow automation (AI Hub Build/Search), 100+ app & data-store integrations. Omkar Pendse joined as Chief Product & Technology Officer Jan 2026. Pain points (INFERRED): reliability of document-processing agents on unstructured enterprise data; token/context cost of agentic extraction at document scale; visibility/control over agent behavior across regulated customers. Challenges (INFERRED): scaling agent workflows across enterprises without cost or accuracy blowup. Must-haves (INFERRED): per-run cost visibility + reliable, governable agents. Nice-to-haves (INFERRED): unified observability across the agent fleet. ICP confidence: Medium (senior product+tech leader, ~165 emp in-band, agents in production — but company is late-stage / ~Series D $2B valuation, beyond the A-C sweet spot; flagged for judgement). LinkedIn URL not surfaced — not fabricated.
Omri Manor — VP of Engineering, AI21 Labs
Signal: 1/4 (ICP at a company now wholly focused on AI-agent orchestration/optimization; surfaced via LinkedIn people search). Source: https://www.linkedin.com/in/omri-manor-02b09a110/ | Company size: ~70 employees as of mid-2026 (restructured ~60% in May 2026 to focus solely on Maestro agent-orchestration; Series D, ~$636M total; commercial contracts with Nebius & Wix). Pain points (inferred): optimizing agent cost & latency at production scale (Maestro's core value prop); reliability of multi-model/multi-tool orchestration; enterprise deployment (Maestro on AWS VPC). Challenges: making agent workflows cost-efficient & reliable for enterprise customers with a leaner team. Must-haves: per-run cost/latency control, orchestration reliability. Nice-to-haves: agent eval/observability. ICP confidence: Medium — strong agent-product fit; caveat: Series D and small headcount after 2026 restructuring.
Ophir Samson — Head of Voice AI (Founder, Ezra AI Labs), Greenhouse Software
Signal: 4 (ICP leader shipping voice agents in production; found via LinkedIn people-search 'Head of AI voice agents' + verified recent posts). Source: https://www.linkedin.com/in/ophirsamson/recent-activity/all/ (post ~1w ago: using voice AI to 'replace the resume, not the interview'). Company size: 501-1,000 (LinkedIn-verified). Building agents: yes - production voice AI for recruiting workflows. Pain points: reliability & latency of real-time voice agents at scale; making voice agents production-grade vs demo-grade. Challenges: human-centered voice UX while automating high-volume screening. Must-haves: reliable, low-latency voice agents in production. Nice-to-haves: cost-efficient scaling of voice interactions. ICP confidence: High (Head of Voice AI at 501-1,000-emp SaaS actively shipping voice agents).
Signal: 4 (ICP leader publicly discussing deploying/controlling AI agents in production). Source: https://ramp.com/blog/introducing-ramp-applied-ai-solutions , https://www.prnewswire.com/news-releases/ramp-launches-applied-ai-solutions-helping-enterprises-deploy-ai-agents-across-finance-operations-302796179.html , https://thenewstack.io/ramp-forward-deployed-engineers-applied-ai/ (personal LinkedIn URL not confirmed in this run — do not fabricate; verify before outreach). Company size: ~1000-1500 (Ramp). Leads Ramp's forward-deployed engineering (Palantir-style FDE motion) building bespoke AI agents embedded in enterprise finance orgs. Ex-Palantir (public-sector + commercial), ex-Google BU lead, ex-Oliver Wyman. Pain points: getting agents to complete complex financial work safely with the controls finance teams need; explicitly names "token spend management" as an agent workflow — i.e. cost visibility/control is top of mind. Challenges: capturing enterprise context and turning it into reliable, controllable agents; per-workflow reliability across AP, procurement, close, AR, expense. Must-haves: control/guardrails over agent actions, reliability, and cost/token spend visibility per run. Nice-to-haves: faster deployment, better observability. ICP confidence: High (Head-of-AI-solutions decision-maker at in-band company shipping agents in production; strong cost-control signal).
Ori Simantov — Director, AI Transformation, HiBob
Signal: 4 (ICP at an agent-operating company; second thread into the HiBob account alongside Reut Einav, ID 993).
Source: https://www.linkedin.com/search/results/people/?keywords=Hibob%20Head%20of%20AI%20engineering ; company page https://www.linkedin.com/company/hibob/people/
Company size: BORDERLINE — same caveat as ID 993. LinkedIn company band "1K-5K employees"; People tab reports 2,118 associated members (28 Aug 2026), which overstates current headcount; commonly reported HiBob headcount is ~1,000-1,300. VERIFY BEFORE OUTREACH. HQ New York / Tel Aviv. HR tech SaaS shipping AI agents in its product.
LinkedIn headline (verbatim): "Director, AI Transformation @ HiBob | 'AI Mind Talks' Podcast Host".
Pain points: NOT DIRECTLY EXPRESSED — no quote captured from him this run; do not fabricate one.
Challenges: an explicit "AI Transformation" mandate at Director level means he owns rollout across functions — the role that discovers agent sprawl and unattributed spend first.
Must-haves (hypothesis, unverified): a view of which agents are running, who owns them, and what they cost.
Nice-to-haves (hypothesis): internal enablement material; he hosts an AI podcast, so he is also a content/relationship surface, not only a buyer.
ICP confidence: Medium — Director level, right function, agent-operating company; discounted for the unresolved headcount question and the absence of any expressed pain. Other ICP-titled HiBob leaders identified but NOT added this run (kept as bench so a future run does not re-spend the effort): Ido Stern (SVP Engineering), Israel David (Co-Founder & CTO), Kobi Sasson (Director of Engineering), Alon Arbiv (Senior Director, Turning AI into Business Impact).
Verification notes: profile URL confirmed by DOM href extraction. NET-NEW company for the People Library.
Signal: 1 (ICP publicly writing/speaking about agent autonomy & control). NOTE: LinkedIn/Chrome unavailable this run — found via web research (press release + product pages), not LinkedIn.
Source: https://www.prnewswire.com/news-releases/atera-launches-it-autopilot-ushering-in-the-era-of-autonomous-it-302459475.html ; https://www.atera.com/ai/
Company size: ~406-414 (CB Insights 406 as of May 2026; LeadIQ ~414 July 2026, 4 continents). Series B, ~$102M raised, ~$500M valuation. Tel Aviv, Israel.
Agents in production: STRONG. "IT Autopilot" — two AI agents autonomously resolving Tier-1 IT tickets WITHOUT human oversight. Published production numbers: up to 40% of IT workload handled, 0.1s response times, ~15 min avg resolution; one customer saw 15% ticket reduction + 20% response-time reduction in first 8-9 weeks (~48 hrs/week saved). Thousands of customers in 90+ countries.
Pain points: [evidenced] He has publicly staked the product on UNSUPERVISED agent autonomy — his own quote: "IT Autopilot delivers actual autonomous IT. We've built a system that plans, acts, learns, and improves on its own, just like a skilled technician would." That makes agent reliability, escalation boundaries and outcome measurement his direct operational problem. [inferred] Cost-per-resolution economics are structurally exposed: Atera prices per-technician, NOT per-token — so every token an agent burns comes straight out of gross margin. This is the sharpest unit-economics mismatch in this run's cohort.
Challenges: Running agents with no human in the loop means failure is silent and customer-visible; proving/maintaining the 40%-of-workload claim as ticket mix drifts; margin compression as agent usage grows under flat-rate pricing.
Must-haves: Per-run cost attribution mapped to per-technician pricing; reliability/regression guardrails for unsupervised agents; escalation-boundary observability (when did the agent hand off, and should it have?).
Nice-to-haves: Cross-customer benchmarking of autonomous resolution rates; token-efficiency tuning per ticket class.
ICP confidence: High — CTO/co-founder, ~410 employees, Series B, genuinely autonomous agents in production at scale, and a real first-person quote committing to unsupervised autonomy.
Pablo Palafox — Co-Founder & CEO, HappyRobot
Signal: Bucket 4 (ICP publicly building/shipping agents — LinkedIn headline "Putting agents to work in the enterprise"). Source: https://www.linkedin.com/in/pablorpalafox/ ; funding confirmation https://a16z.com/announcement/investing-in-happyrobot/ and https://techfundingnews.com/happyrobot-150m-series-c-ai-agents/ . Company size: unicorn after $150M Series C (a16z-led), likely ~150–300 employees (aggressive eng/GTM hiring). YC S23. HappyRobot ships AI agents/workers in production across voice, email, documents and web — started in logistics/supply-chain, now expanded to insurance, energy & utilities, telecom, aviation (many agents live at scale). Technical co-founder (3D computer-vision research background) and CEO. NOTE: distinct from existing library contact Luis Paarup (CTO) — net-new person, same company. Pain points: scaling from a few agents to many across multiple verticals (1→N), reliability of high-volume voice/agent runs in the field, cost/observability of large agent fleets. Challenges: maintaining reliability + margin as call/agent volume explodes across industries. Must-haves: per-run cost and reliability visibility across a large multi-vertical agent fleet; guardrails on runaway usage. Nice-to-haves: unified operating layer for eval, cost control, and observability across agents. ICP confidence: High (technical co-founder/CEO at a unicorn-scale company shipping many agents in production).
Signal: 4 (senior AI leader via "Head of Applied AI ... agents production" people-search + web verification). Source: https://www.linkedin.com/search/results/people/?keywords=%22Head%20of%20Applied%20AI%22%20agents%20production (profile https://www.linkedin.com/in/paolorosson/). Company size: ~497 (confirmed) — Dext, financial/accounting SaaS. Headline: "Building Internal AI Agent Platforms for GTM." Pain points [INFERRED from verified role+company, NOT verbatim]: cost + reliability of internal agent platforms; token waste; visibility into per-agent/per-run cost. Challenges: standing up a reliable internal agent platform and controlling its spend. Must-haves: cost-per-run visibility; reliability. Nice-to-haves: eval/governance. ICP confidence: Medium (Head of Applied AI at ~500-emp SaaS; agents are internal/GTM-focused rather than core product — flagged).
Signal: 4. Source: https://www.linkedin.com/in/paritoshmohan/ (VP of Engineering @ Twelve Labs). Company size: ~180-192 employees; Series B ($207M total). Building agentic video platform ("make video addressable, searchable and usable by agents"). Pain points: scaling engineering for agentic workloads in production; reliability + cost of agent/inference workloads. Challenges: production-scale reliability of agent-facing platform. Must-haves: reliability, observability. Nice-to-haves: cost-per-run visibility. ICP confidence: Medium-High (VP of Engineering at Series B company building agentic platform).
Signal: 4 (technical founder/CEO of a company shipping an employee-support AI agent, "Barista", in production across enterprises). Source: https://aithority.com/interviews/ait-megamind/aithority-interview-with-pat-calhoun-ceo-and-founder-at-espressive/ ; https://www.linkedin.com/in/pacalhoun . Company size: ~150-200 employees; Series B (last disclosed round 2020 — NOTE data may be stale). Technical pedigree: ex-Cisco CTO (switching/routing/wireless/security), SVP Product ServiceNow, founder of Airespace (acq. Cisco $450M). Pain points: accuracy and adoption of employee self-service agent; reliably automating resolution; reducing help desk volume. Challenges: agent accuracy across enterprise knowledge; multilingual support; measuring/controlling agent performance at scale. Must-haves: reliable, accurate agents; observability into resolution rates. Nice-to-haves: lower cost per resolution. ICP confidence: Medium (technical founder/CEO, agent in production, ~150 emp; funding data stale).
Signal: 4 (Head of AI building agent systems; via people search 'Head of AI agents production platform').
Source: https://www.linkedin.com/in/pat-mullee/ ; Storable launched "Agent Assist" 24/7 AI assistant; his portfolio shows EDGE Email Agent + AI orchestration platform.
Company size: ~567 employees.
Pain points: building agent systems & GenAI platforms at scale; shipping production AI agents in a mature vertical SaaS.
Challenges: productionizing/orchestrating agents at scale within an established vertical SaaS.
Must-haves: agent platform/orchestration, observability.
Nice-to-haves: cost management.
ICP confidence: Medium — Head-of-AI title and ~567-emp size both fit, and actively building/shipping agent systems. Caveat: Storable is a PE-backed vertical (self-storage/insurance) SaaS, not a Series A-C or AI-native company, so company-type is a partial ICP mismatch.
Patrick Chatain — Chief Technology Officer, Contentsquare
Signal: 4 (ICP shipping agents). Source: https://www.businesswire.com/news/home/20250513200292/en/Contentsquare-Announces-Sense-an-AI-Agent-That-Plans-and-Acts-Like-an-Analyst ; https://www.cmswire.com/digital-experience/contentsquare-bets-on-multi-agent-ai-to-break-down-cx-silos/. Company size: ~1,700 (Tracxn May 2026) — in the 50-2,000 band. Paris, France. Agent evidence: CTO since 2017; Contentsquare shipped "Sense", an AI agent that plans and acts like an analyst, and described a company-wide roadmap explicitly moving "from AI assistant to agents to automation" (CX Circle 2025). The Sense chat feature alone answered 150,000+ queries since May 2025, implying a meaningful and growing inference cost surface. Pain points: multi-agent orchestration across siloed channels (web, mobile, support, AI-assistant) at high query volume; inference cost surface expanding as agent features roll out. Challenges: coordinating agents across product silos for 1,000+ enterprise brands. Must-haves: reliability and observability across a multi-agent analytics product. Nice-to-haves: cost attribution per customer-facing agent workflow. ICP confidence: Medium — strong, well-documented agent shipping evidence and long-tenured confirmed CTO, but FLAG: funding stage is Series H ($2.8B valuation, 2021), materially later than the stated Series A-C criterion. Headcount fits; stage does not. Treat as enterprise/upmarket motion rather than core ICP.
Patrik "totte" Torstensson — Head of Engineering, Lovable
Signal: 4 — ICP engineering leader at a company whose entire product is agent runs, engaging with content about the economics of those runs.
Source: https://www.linkedin.com/in/totte/recent-activity/all/ (found via https://www.linkedin.com/company/lovable-dev/people/?keywords=head%20of%20engineering)
Company size: 51-200 employees (LinkedIn company page, verified this run). Lovable = AI app-building agent, Series C announced within the last ~2 weeks per colleague posts.
Observed this run: he reposts Lovable company and team content. Most relevant adjacent quote, from Jonas Björk (Head of Data at Lovable), 2w ago: "Finance analytics at Lovable takes both engineering and data science. We sell credits, not seats, so the business metrics and the recognised revenue have to be built before they can be reconciled against each other."
Pain points: Lovable's revenue model is credit-metered agent execution, so cost-per-agent-run is not an infra detail — it sits directly under revenue recognition and margin. Reconciling metered agent consumption against recognised revenue is an open engineering + data problem there right now.
Challenges: agent run cost variance at consumer scale; unit economics of credits vs. actual token/tool spend.
Must-haves: accurate, auditable per-run cost accounting.
Nice-to-haves: forecasting of agent spend as usage grows.
ICP confidence: High — Head of Engineering (core ICP title) at a 51-200 employee Series C AI-native company running agents in production at very large volume. Note: the credits/revenue quote is his colleague's, not his own — do not attribute it to him in outreach.
Signal: 4 (senior eng leader at an agent platform company). Source: LinkedIn people search (https://www.linkedin.com/search/results/people/?keywords=Kore.ai%20VP%20of%20AI%20OR%20head%20of%20agentic). Note: 1st-degree connection (warm path for outreach). Company size: ~1,277 employees (Kore.ai Agent Platform — agentic enterprise AI; ~$300M+ raised, growth/PE-backed — mature stage, but in 50-2,000 band and clearly agent-active). Pain points (INFERRED): scaling an enterprise agent platform reliably; cost + observability across many enterprise agent deployments; governance. Challenges: reliability, governance, per-run cost visibility at enterprise scale. Must-haves: control/observability over agent cost and reliability across tenants. Nice-to-haves: model routing, neutral in-path governance. ICP confidence: Medium (agent platform, right-size; late/growth stage beyond Series A-C — flagged; SVP Eng is a strong senior technical buyer).
Paul B. — Vice President of Engineering (Agentic Enablement), MCO (MyComplianceOffice)
Signal: 1 (ICP author, LinkedIn post search "our agent fleet", past month, relevance sort).
Source: https://www.linkedin.com/in/peb7268/ — post in his "Token Economics" series. Found via https://www.linkedin.com/search/results/content/?keywords=%22our%20agent%20fleet%22&datePosted=%22past-month%22&sortBy=%22relevance%22
Company size: ~376 employees (Tracxn, 30 Jun 2026); ~363 (alternate source, Jun 2026). Compliance/RegTech SaaS. Littleton, Colorado. Surname is abbreviated to "B." on his LinkedIn profile — not withheld by us, that is how the profile renders.
Pain points (his own words): "A leader asked me a simple question about our agent fleet: 'What did that cost us last month?' I didn't have an answer. Not a fuzzy one — nothing." / "The provider's monthly total is a trap. '$4,200 on Anthropic in May' can't tell you which agent, which user, which model drove it. A total with no dimensions isn't a metric — it's a feeling." / "The whole game is resolution: the same dollar sliced by session, user, agent, and model, over time."
Challenges: Zero cost attribution across a production agent fleet; cannot answer an exec cost question; every downstream optimization (routing, caching, context discipline) is blocked until measurement exists — "you cannot manage what you cannot measure."
Must-haves: Per-session / per-user / per-agent / per-model spend resolution over time; ability to attribute a dollar to a specific agent run; convert "we spent a lot" into "the review agent's model choice is costing $900/mo."
Nice-to-haves: Routing, caching and context-discipline levers once instrumented; time-series trending.
ICP confidence: High — VP Engineering (exact ICP title) at a ~376-person regulated-industry SaaS, publicly running an agent fleet in production and publishing about precisely the cost-visibility gap Alpha sells into. Best-fit single signal of this run.
Paul C Nichols — Chief Technology Officer, Campfire
Signal: 4 (ICP technical leader at an AI-native company shipping agents in production; found via 2026 agent-startup research + LinkedIn verification).
Source: https://www.linkedin.com/in/paulcnichols/ ; company context https://www.prnewswire.com/news-releases/campfire-raises-65-million-series-b-to-redefine-how-finance-works-in-the-ai-era-302585077.html , https://www.ycombinator.com/companies/campfire-2
Company size: ~65 employees (San Francisco; expanded from ~10 to ~65 since mid-2025); Series A $35M then Series B $65M (Oct 2025); AI-native ERP with finance/accounting agents automating close, reconciliation, quote-to-cash.
Pain points (inferred from role/company, not quoted): finance/accounting agents must be highly reliable and auditable in production; cost of running agents across many customers' books; scaling from a few agents to a broader agent suite.
Challenges: reliability/accuracy guarantees for financial agents; predictable per-run cost as agent workloads grow across customers.
Must-haves: cost-per-run visibility and controls; reliability/observability for financial-grade agents.
Nice-to-haves: token/context waste reduction; per-customer cost attribution and margin analytics.
ICP confidence: High — CTO of a Series B AI-native company (~65 emp) in the ICP band, actively shipping agents.
Paul Klein IV — Founder & CEO (technical founder), Browserbase
Signal: 3 (ICP engaging with competitor content — named customer on the Braintrust customers page).
Source: https://www.braintrust.dev/customers ; funding/headcount context: https://www.builtinsf.com/articles/browserbase-announces-40m-series-b-funding-20250618
Company size: ~50 employees (Built In SF / Sourcery VC coverage of the June 2025 Series B described the team doubling from 30). Series B, $40M, $300M valuation. Browserbase — headless browser infrastructure for AI agents; agent infra is the core product.
Pain points: (1) no visibility into what a browser-automation agent was "thinking" before it acted — "What we're often missing is, what was the model thinking? That's where Braintrust comes in"; (2) debugging non-deterministic agent runs after the fact; (3) agent failures are opaque without decision-level traces.
Challenges: making long-running browser agents debuggable and predictable for customers who run them at volume; distinguishing model failure from tool/environment failure.
Must-haves: trace-level observability into agent reasoning and decisions; ability to replay and diagnose a failed run.
Nice-to-haves: eval loops that turn production failures into regression cases.
ICP confidence: Medium — technical founder/CEO of an agent-infrastructure company at Series B with strong product fit and a real observability pain quote. Caveats: (a) headcount ~50 sits right at the ICP floor and comes from mid-2025 press, so it needs a fresh check; (b) as an agent-infra vendor themselves, they may build rather than buy this layer. LinkedIn URL was seen directly in search results.
Paula Hingel — Director of Engineering, Augment Code
Signal: 4 (ICP eng leader at an agent-native company; LinkedIn people-search 'Augment Code engineering director'). Source: https://www.linkedin.com/in/paula-hingel-285562252/ . Company size: ~150-300 (Augment Code; Series B, ~$250M raised; agentic SDLC / autonomous coding agents for software teams). Pain points (INFERRED from role + company; no verbatim quote observed this run): reliability and token/context efficiency of coding agents; cost-per-run control at scale; production observability for agent workflows. Challenges: shipping reliable agentic coding workflows at enterprise scale. Must-haves: reliability guardrails + per-run cost/token visibility. Nice-to-haves: context/token-waste reduction. ICP confidence: High (Director of Engineering at a 50-2,000-emp company actively shipping coding agents). Note: brain already has Augment CTO Dion Almaer, Chief Scientist Guy Gur-Ari, VP Eng Vinay Perneti, and (this run) Eng Director John Edstrom — Hingel is net-new.
Signal: 4 (ICP speaking publicly about deploying and operating agents at scale; also engaged in the FDE/agent-ops discourse alongside other ICP leaders). NOTE: LinkedIn/Claude-in-Chrome NOT connected this run — sourced via web research + primary-source verification.
Source: https://www.latent.space/p/cursor-forward-deployed-engineers and https://www.latent.space/p/forward-deployed-engineers-aiewf (Latent Space, Jul 2026); AIEWF 2026 session "Forward Deployed Engineering at Cursor" (https://www.youtube.com/watch?v=APqXGyCoGW4).
Company size: Anysphere/Cursor ~300-1,794 employees depending on source (Revelio Labs Dec 2025: 1,794; JobsByCulture 2026: ~300) — inside the 50-2,000 band on every source.
Pain points: agent deployments that do not survive handoff — her stated bar is "when we walk away, it is a strict ROI for them. That means they're not gonna turn things off when we leave"; enterprise adoption "is still concentrated among early adopters"; deploying cloud agents, long-running agents and automations inside customer environments where ROI has to be provable.
Challenges: proving per-deployment ROI to customers; finding the right internal champions; standing up a "software factory" of long-running agents across a customer's whole SDLC; deciding where FDE effort actually adds value (uses a 2x2 on digital maturity vs product customisation).
Must-haves: demonstrable per-agent/per-run economics and outcome measurement that survives after the FDE team leaves; operational control over long-running agents in customer environments.
Nice-to-haves: reusable deployment patterns across customers; less bespoke per-customer workflow engineering.
ICP confidence: High — VP-level engineering leader at an AI-native company shipping agents in production at scale; headcount in band; pain is explicitly agent ROI/operations.
Pedro Henrique Santos — Director of Algorithms, Sword Health
Signal: Bucket 4-adjacent (ICP AI/ML leader at agent-native healthtech). Sourced via LinkedIn people search "Sword Health AI engineering director". Source: https://www.linkedin.com/in/pedrohenolisantos/ . Company size: ~1,000-1,520 employees (Sword Health, Series D+ unicorn; size in 50-2000 band). Agent-native: Algorithms org underpins Phoenix AI care agent (computer vision + intelligent systems) shipped in production. Pain points (inferred from role/company; no public post captured this run): production reliability of ML/agent pipelines, cost/latency at patient scale, evaluation of agent outputs. Challenges: scaling agent inference cost-effectively; monitoring quality across many runs. Must-haves: eval + reliability tooling, cost-per-run visibility. Nice-to-haves: model/provider portability. ICP confidence: Medium (Director of Algorithms — AI leadership ✓, size ✓, agent-native ✓; stage Series D+; no direct engagement signal observed).
Pedro Lis French — CTO, LaHaus
Signal: 4 (LinkedIn people search — CTO + "multi-agent systems in production"). Source: https://www.linkedin.com/search/results/people/?keywords=CTO%20AI%20agents%20in%20production (profile: https://www.linkedin.com/in/pedro-lis-french-17b88823/). Company size: ~269 employees (Series C proptech, Colombia/Mexico; formerly ~900, reduced). Headline explicitly: "Multi-agent systems in production · LLM evals, conversational AI agents at scale, AI agent fleet orchestration"; leads AI platform strategy + K8s AI agent orchestration. Pain points (stated in headline / inferred): fleet orchestration of many production agents; LLM evals; conversational agents at scale. Challenges: reliability and eval of agent fleets; cost/visibility across concurrent agents. Must-haves: agent fleet observability + cost-per-run; production eval/reliability tooling. Nice-to-haves: orchestration cost optimization. ICP confidence: High (CTO, Series C, 50–2,000 emp, explicitly running multi-agent systems in production).
Pei-Hao (Eddy) Su — Co-Founder & SVP Engineering, PolyAI
Signal: 4 (ICP building/shipping agents in production). Source: https://uk.linkedin.com/in/phs26 and https://theorg.com/org/polyai/org-chart/pei-hao-eddy-su. Company size: ~372 employees, AI-native enterprise voice agents at very high production volume (100M+ interactions/yr). Technical engineering leader — PhD Cambridge (reinforcement learning / dialogue systems), owns the engineering org. NOTE: PolyAI is Series D, one stage past the stated A-C band; headcount and agent-shipping profile fit ICP. As the SVP Engineering he is the direct owner of agent reliability and inference-cost decisions — a prime technical buyer persona. Pain points (inferred from role, not a verbatim quote): production reliability of voice agents, token/inference cost at scale, observability into what agents do per run. Challenges: scaling a reliable agent platform while keeping per-conversation cost sustainable. Must-haves: cost-per-run visibility, reliability tooling. Nice-to-haves: unified agent observability. ICP confidence: Medium-High — ideal VP-Eng persona; only caveat is funding stage.
Signal: 4 (ICP Director-level AI leader at agent-native company; joined via Uniphore's acquisition of Orby AI). Source: https://www.uniphore.com/blog/meet-peng-qi-senior-director-of-ai-science/ ; https://theorg.com/org/uniphore/teams/artificial-intelligence-division . Company size: ~1,000-1,500 (Uniphore, Business/Agentic AI, Series F). Pain points (inferred, not quotes): making agents reliable enough for non-technical teams to trust in production; frontier-research-to-product gap; cost/latency of running agentic reasoning at scale. Challenges: turning research agents into reliable production tools. Must-haves: production reliability, cost-per-run control as agents scale. Nice-to-haves: eval/benchmarking, orchestration cost visibility. ICP confidence: High (Sr Director of AI Science, ex-Amazon Q / Stanford NLP, at an in-range agent-native company; verified current).
Penny Allen — SVP, AI, Product & Engineering, Shipium
Signal: 4 (ICP senior technical AI leader shipping agents in production; found via LinkedIn People search for ICP titles + "shipping agents"). Source: https://www.linkedin.com/search/results/people/?keywords=VP%20of%20AI%20shipping%20agents%20startup ; verification: https://www.shipium.com/ai and https://www.shipium.com/en/blog/announcing-our-27.5m-funding-the-largest-series-a-in-logistics-software-history. Company size: ~100-250 (Shipium — shipping/supply-chain SaaS, founded 2019, ~$27.5M Series A, largest Series A in logistics software; AI-native). Actively shipping agents (CONFIRMED): Shipium's "shipping AI agents monitor and manage response to events like labor disputes or major storms, and get smarter as they execute across your network." Pain points (product-confirmed + inferred): reliability of autonomous agents responding to real-world network events; making agent outcomes compound/improve across runs; per-run cost + observability across many customer networks. Challenges: keeping multi-step logistics agents reliable and cost-controlled in production at scale. Must-haves: production reliability, visibility into agent behavior, compounding quality across runs. Nice-to-haves: per-run cost observability/attribution. ICP confidence: High (SVP AI/Product/Eng — clearly senior technical AI leader — at AI-native logistics SaaS in the 50-2,000 band actively shipping agents in production). Profile URL confirmed via web (do not fabricate others).
Signal: 4 (senior technical/product leader at an agent-native company). Source: LinkedIn people search (https://www.linkedin.com/search/results/people/?keywords=Decagon%20VP%20Engineering%20AI%20agents). Company size: ~210-500 (Series C, $1.5B val, June 2025 $131M round; AI customer-experience agents across chat/voice/email). Pain points (INFERRED from verified role+company; no verbatim quote): as owner of agent product, ensuring autonomous CX agents resolve reliably at enterprise scale and controlling cost-per-resolution as agent volume grows; protecting margins under outcome-based pricing. Challenges: reliability/quality of production agents across many enterprise tenants; per-conversation/per-run cost visibility. Must-haves: reliable agent behavior in prod + visibility/control over what each agent run costs. Nice-to-haves: model routing/optimization, deeper evals/observability. ICP confidence: HIGH (clean fit — agent-native, right-size, Series C; VP of agent product directly owns the category thealpha sells into).
Signal: 4 (ICP technical decision-maker at a company shipping agents in production; also Signal 1 — publicly frames the product as 'AI agents that investigate, diagnose and resolve incidents with you'). Source: https://incident.io/blog/incident.io-raises-62m ; https://www.insightpartners.com/ideas/incident-io-raises-62m-to-build-ai-agents-that-resolve-incidents-with-you/ . Company size: ~200 employees (Tracxn, May 2026); Series B $62M at $400M valuation (Apr 2025), ~$96M total; founded 2021 by ex-Monzo team. Pain points (inferred from public product/positioning; no first-party quotes captured): AI incident-response agents must act reliably during high-stakes production outages; agents investigating/diagnosing across live systems risk wrong autonomous remediation; cost/latency of running long-lived investigation agents at scale (250k+ incidents processed). Challenges: earning engineer trust as agents take remediation actions during outages; scaling agent autonomy without eroding reliability. Must-haves: production-grade reliability, guardrails/human oversight on agent actions, observability into what the agent did and why. Nice-to-haves: per-run cost visibility, agent accuracy that compounds over time. ICP confidence: High (clean technical decision-maker, agents in production, Series B, ~200 emp — squarely in ICP).
Peter Hill — Chief Technology Officer, Synthesia
Signal: 4 (ICP writing/talking about building agents). Source: https://www.webpronews.com/synthesias-4-billion-bet-ai-agents-reshape-corporate-training/ ; https://www.neonriver.com/peter-hill-2025/. Company size: ~700-758 (Revelio Labs / company reporting, late 2025 to early 2026). London UK. Agent evidence: company messaging under Hill (CTO since Oct 2024, ex-AWS) frames a "$4 billion bet" on AI agents reshaping corporate training and knowledge work, describing "a rare convergence... a technology shift with AI agents becoming more capable". Pain points: NOT independently sourced beyond company/press-style messaging — this is explicitly the weakest evidentiary link in this run's batch and is recorded as such. Challenges: scaling the technical org (headcount +70% planned in 2026) while shifting the core product surface from video generation toward agentic workflows — an org-and-architecture transition that typically exposes agent-ops gaps. Must-haves: not directly stated; inferred need for production reliability as the product line pivots toward agents. Nice-to-haves: unknown. ICP confidence: Medium (lower end) — title, company and headcount all verified, but agent evidence is strategic vision and press framing rather than a concrete "we run N agents in production" statement. Include in nurture, not in a pain-led first-touch sequence.
Peter Levi — Director of Engineering, Hyperscience
Signal: 4 (ICP Director-of-Engineering at an AI-native company building agentic automation). Source: LinkedIn profile above — found via people search scoped to Hyperscience ("Hyperscience head of engineering"). Company: Hyperscience, enterprise AI infrastructure / Intelligent Document Processing with automation & orchestration of end-to-end processes (moving to agentic "Hypercell"). Company size: 201-500 employees (confirmed via LinkedIn company page, New York). Pain points (inferred from role + company, NOT a verbatim quote): reliability and cost of automating end-to-end document/decision processes; scaling orchestration/agentic workflows in production for large enterprises. Challenges: accuracy & cost trade-offs at high document volume; observability across orchestrated steps. Must-haves: production reliability, cost/usage visibility, evals. Nice-to-haves: unified control/observability layer. ICP confidence: Medium (Director of Engineering at a 201-500 emp AI-native company; note company is primarily IDP/hyperautomation moving toward agentic orchestration rather than a pure AI-agent product).
Peter Secor — Senior Vice President of Engineering, LaunchDarkly
Signal: 4 (engineering exec at a company actively shipping agent runtime control).
Source: LinkedIn people search — headline verbatim: "SVP Engineering at LaunchDarkly (ex-Slack, Meta)"; result card also states "Current: Senior Vice President of Engineering at LaunchDarkly". Company agent activity verified via launchdarkly.com/platform/agent-control/ and helpnetsecurity.com/2026/05/19/launchdarkly-agentcontrol/.
Company size: ~466 employees as of February 2026.
Pain points (inferred from company product direction, NOT from a personal post — he was not observed posting): owning engineering for a platform whose customers are hitting agent cost/quality/runaway-behaviour problems in production; LaunchDarkly's own framing is that teams need to block bad agent behaviour and steer responses in real time, and that LLM judges must score candidates on quality AND cost.
Challenges: scaling an org from feature flags into agent runtime control; agent spend and behaviour as a first-class runtime concern.
Must-haves: real-time control over agents already running in production without redeploy.
Nice-to-haves: cost-aware model routing and fallback policy.
ICP confidence: Medium — title (SVP Engineering) and company size (466) both qualify cleanly and the company is demonstrably shipping agent infrastructure, but no direct first-person pain signal from him was observed this run. Rated Medium rather than High for that reason. Vadim Korolik (same company, id 970) is the warmer, more specific entry point.
Petr Baudis — Founder, CTO & Chief AI Architect, Rossum
Signal: 4 (ICP technical AI leader / technical co-founder at a company confirmed to be actively building AI agents). Source: https://www.linkedin.com/in/petr-baudis-906a213/ — found via LinkedIn people search scoped to Rossum ("Rossum co-founder CTO"). Company: Rossum (rossum.ai), intelligent document processing; company About states "specialized agents automate document workflows end to end." Company size: 201-500 employees (confirmed via LinkedIn company page, London/Prague). Pain points (inferred from role + company stage, NOT a verbatim quote — discovered via search, not a specific post): scaling specialized document/transaction agents to enterprise volume; controlling per-transaction LLM cost as agent usage grows; reliability/accuracy of agentic extraction in production. Challenges: moving from single models to multi-agent orchestration end-to-end; cost/performance trade-offs at high document throughput. Must-haves: production reliability, per-run cost visibility, evals for agent output quality. Nice-to-haves: unified observability across the agent pipeline. ICP confidence: High (CTO/Chief AI Architect, technical co-founder, at a 201-500 emp company explicitly shipping agents).
Signal: 4 (ICP Director of AI/ML at an agent-native company). Source: https://www.linkedin.com/in/phaninivarthi/ (found via LinkedIn people search 'Aisera director engineering AI agents'). Company size: Aisera ~259-311 employees (enterprise agentic AI platform — AiseraGPT / AI agents; acquired by Automation Anywhere Nov 2025). Pain points (inferred from role, not a verbatim quote): reliability and accuracy of enterprise AI agents; ML/agent quality at scale. Challenges: scaling agentic AI across many enterprise customers; cost/performance of LLM-backed agents. Must-haves: agent reliability, evaluation, observability. Nice-to-haves: cost-per-run visibility. ICP confidence: High (Director AI/ML at an agent-native company inside the size band; caveat: recently acquired by Automation Anywhere).
Signal: 4 (CTO at an AI-native company that shipped a GA enterprise agent platform — Cohere North; surfaced via search on named CTO/Head-of-AI building agents).
Source: https://techcrunch.com/2025/08/06/coheres-new-ai-agent-platform-north-promises-to-keep-enterprise-data-secure/ ; https://cohere.com/blog/north-ga ; https://uk.linkedin.com/in/phil-blunsom-95446a1b4
Company size: ~800+ employees early 2026 (up from ~250 mid-2024); $500M raise at ~$6.8–7B valuation. Fits 50–2,000.
Background: Promoted to CTO mid-2025 (replacing Saurabh Baji); ex-DeepMind/Oxford. Cohere North is a GA agentic AI platform for secure enterprise deployments; CFO publicly frames Cohere's edge as managing compute economics.
Pain points (inferred from role/company focus): compute/inference cost economics of agents; secure, reliable agent deployment in private enterprise settings; benchmarking/evaluating agent systems.
Challenges: keeping agent inference cost-efficient while scaling North to enterprises.
Must-haves: cost-efficient inference, reliability, security/governance for agents.
Nice-to-haves: agent evaluation benchmarks, compounding performance.
ICP confidence: Medium (strong role/company/size fit and actively shipping agents; but a frontier-model company builds its own infra, so buying intent for an external agent operating layer is lower).
Philip Lakin — Director of AI Transformation, Zapier
Signal: 4 (ICP Director of AI at validated target company Zapier; sourced via LinkedIn People search, run 2026-08-01). Source: https://www.linkedin.com/in/philip-lakin/ . Company size: ~800 (Zapier — automation platform shipping AI agents: Zapier Agents / Central; in-range). Note: technical (co-founder of NoCodeOps, acquired by Zapier 2024). Pain points (role/company-contextual): operationalizing agents across the org cost-effectively; visibility into what agents cost per run; reliability/control as agent usage scales. Challenges: scaling agent adoption without runaway LLM spend; governance/control over many agents. Must-haves: cost-per-run visibility; control/guardrails. Nice-to-haves: model routing, eval/observability. ICP confidence: Medium (Director of AI at 50-2,000-emp agent-shipping company; role leans AI transformation/adoption rather than core agent eng).
Signal: 4 (ICP senior technical leader at a company actively shipping AI agents; found via LinkedIn people search + web verification of stage/size).
Source: https://www.linkedin.com/in/pierreleroy/ ; company verified via Nabla Series C press (PRNewswire/FierceHealthcare, Jun 2025) and Tracxn/PitchBook headcount.
Company: Nabla — agentic AI for clinical workflows (ambient documentation + clinical agents), deployed at 130+ healthcare orgs, 85,000 clinicians.
Company size: ~143-171 employees (Tracxn/PitchBook 2026).
Stage: Series C ($70M Jun 2025, $120M total).
Pain points (INFERRED from role + company, NOT verbatim): scaling agentic clinical workflows into production reliably; cost of running agents across 85k clinicians; per-run/per-encounter cost visibility; reliability/latency in regulated healthcare setting.
Challenges (inferred): moving from ambient scribe to multi-step clinical agents while controlling LLM spend and ensuring safety.
Must-haves (inferred): production reliability, cost governance/observability per run, guardrails.
Nice-to-haves (inferred): token/context optimization, eval tooling.
ICP confidence: High (VP Engineering at 50-2,000 emp Series C company actively shipping agents).
Pierre Pfennig — VP Data Science, Dataiku
Signal: 4 (ICP AI/ML leader at agent-native company; found via LinkedIn currentCompany people-search of Dataiku data-science/AI leadership). Source: linkedin.com/in/pierre-pfennig-56a16226 (Dataiku company id 2770554 people search, past-month). Company size: ~1,000-1,200 (Dataiku, enterprise AI platform; LLM Mesh + agentic AI). Pain points (INFERRED from role+company stage, not an individual quoted post): cost/token efficiency of data-science & agentic workflows across many models; measuring cost-per-outcome of agent pipelines; reliability/eval of non-deterministic agent outputs for enterprise customers. Challenges: proving ROI/unit-economics of agents vs raw model spend; governing an expanding model/agent estate. Must-haves: per-run cost visibility + control; production eval/reliability for agents. Nice-to-haves: model routing/cost optimization; shadow-savings measurement. ICP confidence: High (VP Data Science, technical AI leader, in-range agent-shipping company).
Pierre-Alexandre Masse — SVP of Engineering, Gorgias
Signal: 4 (company-scoped LinkedIn people search — ICP technical leader at an agent-shipping company). Source: https://www.linkedin.com/in/pimasse (found via https://www.linkedin.com/search/results/people/?keywords=Gorgias%20VP%20Engineering). Company size: ~500 employees (Gorgias, Series C e-commerce CX platform; ships Gorgias AI Agent for autonomous customer-support resolution). Pain points (INFERRED from role+company, not a personally observed post this run): agent cost/margin control as AI Agent resolves high volumes of support tickets; per-conversation cost visibility; reliability/accuracy of automated resolutions in production. Challenges: scaling automated resolution across many merchants while holding quality; controlling LLM spend at support volume. Must-haves: per-run/per-agent cost observability; reliability guardrails. Nice-to-haves: routing/model-selection optimization to cut token cost. ICP confidence: High (SVP Engineering = senior technical leader; Gorgias is an independent 50-2,000-emp agent-shipping SaaS). Brain already had Alex Plugaru (CTO) and Victor Duprez (Sr Dir Eng, AI) — this is net-new.
Ping Wu — CEO (ex-Google AI; former VP Engineering & Product), Cresta
Signal: 4 (ICP technical leader at AI-native company shipping agents in production). Source: https://www.linkedin.com/in/pingwu/ , https://getlatka.com/companies/cresta , Tracxn/LeadIQ Cresta profiles | Company size: ~500–700 employees (2026; Latka ~508, LeadIQ ~615, Tracxn ~693 — within 50–2,000). Generative-AI platform for contact centers deploying live AI agents + agent-assist in production for enterprise CX. Wu is ex-Google AI (co-founded Google Contact Center AI, ran Dialogflow/Speech/NLP); was Cresta's VP Engineering & Product before CEO — deeply technical decision-maker. Pain points: running real-time voice/chat agents in production at enterprise reliability; inference cost at high call volume; measuring and controlling agent quality/behavior across many deployments. Challenges: latency + cost of real-time agents; reliability/quality bar for regulated contact-center use; per-run cost visibility. Must-haves: production reliability, cost/latency visibility per run, quality control. Nice-to-haves: observability and behavior guardrails. ICP confidence: High (technical CEO / former VP Eng at a 50–2,000-employee AI-native company shipping agents in production).
Signal: 4 (ICP senior technical leader at a company actively shipping AI agents; found via LinkedIn people-search, not an observed post). Source: https://www.linkedin.com/in/piotr-dabkowski-50222bba/ | Company size: ~300-500 (Series C; voice AI + Conversational AI Agents platform). Pain points (INFERRED from role + company product focus — no observed quote): scaling agents reliably in production; controlling per-run LLM/agent cost as volume grows; visibility into what agents cost and how they behave across customers. Challenges: keeping agent accuracy/guardrails consistent at scale; token/context efficiency; cost blowout as they move from pilots to broad production. Must-haves: production reliability + observability into agent cost and behavior. Nice-to-haves: per-run cost attribution, context/token optimization, spend guardrails. ICP confidence: Medium (Technical co-founder; ElevenLabs now ships a Conversational AI Agents platform (agents is one of several product lines)).
Piotr Dąbkowski — Co-founder & CTO, ElevenLabs
Signal: 4 (ICP technical co-founder/CTO at AI-native company shipping agents in production). Source: https://elevenlabs.io/blog/series-c ; https://en.wikipedia.org/wiki/ElevenLabs ; Tracxn/Crunchbase (headcount ~960 as of May 2026). Company size: ~960. Stage: Series C ($180M) → Series D ($500M, Feb 2026, ~$11B valuation) — within Series A–C+ AI-native band by headcount. Actively building/shipping agents: ElevenLabs Conversational AI / "ElevenAgents" — developer + enterprise platform for voice and chat agents deployed at scale (Deutsche Telekom, Square, Revolut, Ukrainian Govt) for support, commerce, inbound sales. Pain points (inferred from public product positioning, NOT verbatim quotes): reliability, testing, and monitoring of voice/chat agents at large-scale customer operations; latency and cost of real-time voice agents at high call volume. Challenges: keeping enterprise voice agents reliable and observable across many integrations. Must-haves: reliability, monitoring/observability, integration testing for production agents. Nice-to-haves: cost/latency efficiency per run at scale. ICP confidence: High (technical co-founder & CTO; 960 employees; AI-native; enterprise agents in production).
Piyush Gupta — Senior Director, AI & Platform (Models and Agents), Uniphore
Signal: 4 (ICP Director-level AI/platform leader explicitly owning 'Models and Agents' at agent-native company). Source: https://rocketreach.co/piyush-gupta-email_292663 ; https://theorg.com/org/uniphore/teams/artificial-intelligence-division . Company size: ~1,000-1,500 (Uniphore, Business/Agentic AI, Series F). Pain points (inferred, not quotes): operating models + agents in production reliably; controlling per-run/token cost across an agent platform; visibility into what agents cost per run. Challenges: scaling an agent platform 1->many while holding reliability + cost. Must-haves: per-run cost observability, production reliability, guardrails. Nice-to-haves: model routing, spend optimization. ICP confidence: Medium-High (role/company verified via RocketReach + The Org; profile_url is best-match given common name, confirm before outreach).
Prabhav Jain — CEO (formerly CTO, technical co-founder/leader), 11x
Signal: 1/4 (ICP speaking about running agents in production + governance). Source: "Agentic SDR: Prabhav CEO @ 11x" Stacked GTM podcast https://open.spotify.com/episode/7Fq5HXm4HN47vWW9wvNfxx ; a16z "How 11x uses AI agents to build AI agents" https://a16zbuild.substack.com/p/how-11x-uses-ai-agents-to-build-ai. Company size: ~50–100 employees (AI-native SDR company, Series B/C; borderline lower bound). Pain points: running agentic SDRs (Alice/Julian) in production at scale; AI governance and control of customer-facing agent behavior; operating a full internal AI agent stack. Challenges: reliability/quality of autonomous outbound agents; governing/controlling agent behavior; agent cost as usage scales. Must-haves: control/governance over agent behavior; reliability in customer-facing agents. Nice-to-haves: cost efficiency across the agent fleet. ICP confidence: Medium — technical CEO (ex-CTO) at an AI-native company whose product IS production agents; headcount near the lower bound (~50–100) and company has had reported turbulence, hence Medium not High. Profile URL left blank (not fabricating LinkedIn URL).
Prabhav Jain — CEO (formerly CTO), 11x
Signal: 1 (ICP publicly writing/speaking about AI-agent reliability — stated 'AI SDRs don't work in their current form'). Source: https://gtmcouncil.substack.com/p/prabhav-jain-ceo-11x-on-stacked-gtm ; https://www.linkedin.com/in/jainprabhav/ . Company size: ~100-200 (Series B $50M led by a16z, ~$76M total; scaling from single product to GTM platform). Pain points: current-generation AI SDR/agents don't reliably work end-to-end; making autonomous 'digital workers' (Alice, Julian) own entire GTM workflows reliably in production. Challenges: moving agents from demo to dependable end-to-end execution; scaling multi-agent GTM workforce. Must-haves: production reliability, agents that complete full workflows, not just tasks. Nice-to-haves: broader role coverage across GTM functions. ICP confidence: High (technical CEO/ex-CTO, MIT, ex-Brex head of eng, at Series B AI-native agent company in target size band).
Signal: 4 (ICP seat at a company shipping agents; no public voice yet).
Source: title verified directly on https://www.databahn.ai/leadership.
Company size: ~112 employees as of 30 Jun 2026 per Tracxn (search snippet only, medium confidence); Series B $40M Jul 2026 led by Insight Partners. Estimate 100–130 — inside band.
Pain points: No personal public quote found. Company-level (CEO Nanda Santhana, verbatim, https://www.databahn.ai/blog/the-pipeline-was-just-the-beginning): "Organizations consume more storage, more compute, and more AI tokens while delaying the moment when intelligence can be extracted."
Challenges: Engineering ownership of an agentic data control plane doing federated search and orchestration — workloads where token spend compounds per query.
Must-haves: Per-agent and per-stage cost visibility; controls on runaway retrieval/orchestration loops.
Nice-to-haves: Cost regression alerting as agent count grows.
ICP confidence: Medium — ICP-matching title verified on a company-owned page and headcount inside band, but no expressed pain and he is the second engineering leader at an account where Karthik Paulramachandran (VP Eng) is the stronger entry point.
Watch-out: DataBahn markets token-cost reduction itself — partial competitive overlap.
Pranav Maydeo — Senior Director of Engineering, Uniphore
Signal: 4 (ICP senior technical leader at an agent-native company; discovered via LinkedIn people-search of companies actively shipping AI agents) | Source: https://www.linkedin.com/in/pranav-maydeo | Headline: Senior Director of Engineering @ Uniphore | Ex-Infoworks.io, Target, Walgreens | Company: Uniphore — actively building/shipping AI agents in production | Company size: ~800–1,000 (estimate) | Pain points (INFERRED from role/company context — not a verbatim quote observed this run): agent cost blowout and per-run spend visibility; reliability in production; scaling many agents | Challenges: operating a growing fleet of enterprise agents within cost and reliability targets | Must-haves: per-agent/per-run cost observability; reliability guardrails | Nice-to-haves: token-budget controls; benchmarking | ICP confidence: High (senior technical/eng leadership at a 50–2,000-employee company shipping AI agents)
Prasad Kavuri — Director, AI Platform & Agentic Solutions, Zip
Signal: 4 (ICP AI leader at an agent-shipping company; surfaced via LinkedIn people-search 'Augment Code engineering director', role at Zip). Source: https://www.linkedin.com/in/pkavuri/ . Company size: ~600-900 (Zip; Series D procurement-orchestration platform building agentic AI). Profile headline signals: 'Agentic AI · Enterprise AI Platforms · AI Governance · AI FinOps' (ex-Krutrim, Ola, HERE). Pain points (INFERRED, strongly supported by the profile's own AI-FinOps/governance framing; no verbatim quote observed this run): AI cost management / FinOps for agentic solutions; governance and reliability of enterprise agent platforms; visibility into agent spend. Challenges: standing up a governed, cost-controlled enterprise agent platform. Must-haves: agent cost visibility (FinOps) + governance/reliability controls. Nice-to-haves: cross-model routing. ICP confidence: Medium-High (squarely a 'Director of AI' ICP building agentic solutions; Zip ~600-900 emp within band). Note: brain already has Zip Co-founder & CTO Lu Cheng — Kavuri is net-new.
Prasad Kavuri — Director, AI Platform & Agentic Solutions, Zip (ziphq)
Signal: 4 (LinkedIn people search — Director, AI Platform & Agentic Solutions + agents-in-production). Source: https://www.linkedin.com/search/results/people/?keywords=Director%20of%20AI%20agents%20production (profile: https://www.linkedin.com/in/pkavuri/). Company size: ~1,000–1,334 (Zip / ziphq, procurement + spend-orchestration SaaS, ~$2.2B valuation). Headline focus: Agentic AI · Enterprise AI Platforms · AI Governance · AI FinOps (Ex-Krutrim, Ola, HERE). Pain points (stated/inferred): AI FinOps — controlling and attributing agent/LLM spend; governance across an enterprise agent platform. Challenges: cost visibility and governance as agentic solutions scale across the org. Must-haves: AI FinOps / cost-per-run attribution; governance + control over agent spend. Nice-to-haves: model right-sizing, chargeback by team/agent. ICP confidence: High (Director AI Platform/Agentic, SaaS, 50–2,000 emp; explicitly focused on AI FinOps + agentic solutions — strong pain-signal match).
Prasanna Arikala — Chief Technology Officer & Head of Product, Kore.ai
Signal: 4 (ICP CTO at agent platform shipping agents in production at scale). Source: https://www.linkedin.com/in/arikala/ ; https://theorg.com/org/kore-ai/org-chart/prasanna-arikala ; https://www.kore.ai/about-us. Company size: ~1,000 (400+ member eng/product/R&D org he built; Kore.ai raised $150M growth round). Kore.ai has agents deployed in production at scale across banks, telecom, retail, healthcare; ranked a top agentic-AI platform for 2026. Pain points (inferred from role/domain): reliability and governance of enterprise agents in regulated verticals; cost and observability across many deployed agents; scaling agent development (structured systems generating 6,500+ code commits/month). Challenges: enterprise-grade reliability across many customers; agent lifecycle/orchestration. Must-haves: reliability, observability, cost control at fleet scale. Nice-to-haves: agents that improve over time; institutional memory. ICP confidence: High (CTO of an independent, in-band company actively shipping agents in production).
Prasanna Arikala — Chief Technology Officer & Head of Product, Kore.ai
Signal: 3 (ICP building agent observability/governance/control — directly adjacent to competitor category). Source: https://www.kore.ai/news/kore-ai-launches-agent-management-platform ; https://www.kore.ai/videos/building-the-enterprise-agent-ecosystem ; https://www.linkedin.com/in/arikala/ . Company size: ~700-1000 (Series C/D, well-funded; built 400+ eng/R&D team). Pain points: 'enterprise AI sprawl' — managing/optimizing many agents; lack of governance, transparency and accountability over agents; 'confidence, not technology' (trust/reliability) is the main adoption hurdle. Challenges: operating agents with the same discipline as critical business systems; observability and control across an agent fleet. Must-haves: agent management/observability layer, governance & control, reliability/trust. Nice-to-haves: process-level orchestration (AI for Process). ICP confidence: High (CTO & first employee at large AI-native enterprise-agent company; explicitly building the control/observability layer that is Alpha's category).
Prasanna Joshi — Director of Engineering (UI & UX), Uniphore
Signal: 4 (adapted — LinkedIn company People-page directory). Source: https://www.linkedin.com/company/uniphore/people/ and https://www.linkedin.com/in/prasanna-joshi-2488b515/. Company size: 501-1K employees (LinkedIn header). In-band; Uniphore ships enterprise conversational/voice AI + agentic automation (Orby AI agent platform). Pain points (INFERRED, no quotes): cost per conversation/agent-run at high contact-center volume; reliability across many enterprise deployments; multi-agent orchestration. Challenges: controlling unit economics as agent volume scales across customers. Must-haves: per-run cost visibility, reliability guardrails. Nice-to-haves: governance, model routing. ICP confidence: Medium (Director of Engineering at in-band agent-native company; UI/UX focus is adjacent to, not core to, agent cost/reliability pain).
Signal: 4 (ICP technical leader found via LinkedIn people search at a named agent company). Source: https://www.linkedin.com/search/results/people/?keywords=Kore.ai%20VP%20engineering%20agents%20platform . Profile: https://www.linkedin.com/in/prashant-potluri/ . Company size: ~1,000-1,200 (enterprise agentic AI platform: multi-agent, voice, RAG). Pain points (inferred from role+company, no direct quote): scaling a multi-agent enterprise platform in production; cost/token control and reliability across many customer agent deployments. Challenges: governing agent spend and reliability across a large multi-tenant fleet; demo->production gap for enterprise buyers. Must-haves: per-run cost visibility + reliability governance at fleet scale. Nice-to-haves: anomaly detection on spend; customer-owned model routing. ICP confidence: High (VP Eng at a large, independent, actively-shipping agentic-AI platform; 1st-degree connection (warm)). Found by scheduled task icp-prospect-signal-scanner run 2026-08-03.
Prateek Jogani — CTO, Qoala
Signal: 3 (ICP publicly engaging with a competitor — named customer testimonial on Portkey's AI Gateway page). NOTE: LinkedIn/Chrome not connected this run; found via competitor customer-story research. LinkedIn URL not verified — do not guess one.
Source: https://portkey.ai/features/ai-gateway
Company size: ~350-450 employees (Tracxn, May 2026). Series C, $134M total raised (Mar 2024 round). Insurtech, Indonesia/SEA.
What he said (verbatim, short): "managing over 25 GenAI use cases became a pain."
Pain points: no unified cost tracking across 25+ separate GenAI/agent use cases; API key and credential sprawl; no per-use-case spend visibility at 30M policies/month volume.
Challenges: scaling GenAI across a very high-volume insurtech without losing cost control or governance; prompt management across many teams.
Must-haves: cost tracking per use case, prompt management, key/credential usage visibility.
Nice-to-haves: vendor support during evaluation; routing/failover across providers.
ICP confidence: High — verified CTO title, Series C, headcount squarely in band, non-US (Indonesia — under-covered geography), and already paying for a competitor in exactly our category.
Prathamesh Juvatkar — Co-Founder & CTO, Nanonets
Signal: Bucket 1 (ICP technical leader at company shipping multiple production agents) — surfaced via web research (LinkedIn/Chrome unavailable this run). Source: https://nanonets.com/agents ; https://www.crunchbase.com/organization/nanonets ; https://www.cbinsights.com/company/nanonets/people. Company size: ~100-200 employees (51-200; engineering mostly in India). Nanonets ships "Nanonets Agents" — an AI workforce of agents that read, validate, route and post documents across AP, order management, logistics and healthcare RCM; 10,000+ customers, 1B+ docs/yr. Pain points: reliability/accuracy of multi-step document agents at enterprise scale; cost of running agents over 1B+ documents; validation/guardrails. Challenges: keeping agent decisions accurate and auditable across many verticals; scaling agent throughput cost-effectively. Must-haves: per-run cost visibility, reliability/accuracy monitoring, guardrails for autonomous document actions. Nice-to-haves: cross-workflow orchestration observability. ICP confidence: High — Co-Founder & CTO at a 50-2,000 emp company actively shipping 5+ production agents.
Prathyush Parvatharaju — Director of Machine Learning Engineering, CodaMetrix
Signal: 4 (ICP technical AI leader at an agent-shipping company; surfaced via LinkedIn people/company search this run, no specific post authored). Source: LinkedIn people search 'CodaMetrix director machine learning' + profile in/prathyush-parvatharaju-732a7712. Company size: ~150-200 (CodaMetrix, Series B, Boston; autonomous medical-coding AI agents). Pain points (role/domain-inferred, not directly expressed this run): ML platform/infra cost; observability and cost-per-run across production agent pipelines; scaling ML engineering reliably. Challenges: running a reliable, cost-controlled ML engineering platform at claims scale. Must-haves: cost + reliability visibility per pipeline/run. Nice-to-haves: automated cost/quality guardrails. ICP confidence: Medium (Director of ML Engineering at qualifying agent-native company; pain inferred this run).
Pratyush Kumar — Co-founder, Sarvam AI
Signal: 4 (technical co-founder building & shipping agents at scale). Source: https://www.sarvam.ai/about-us ; https://www.forbesindia.com/article/ai-special-2025/sarvam-ais-pratyush-kumar-and-vivek-raghavan-on-building-ai-for-bharat/96208/1 ; https://www.bvp.com/news/sarvam-ai-building-sovereign-ai-for-india . Company size: ~200-300 est. (Series B unicorn, $1.5B val, $234M raised). Actively building agentic platforms (Samvaad, Arya) already handling billions of enterprise interactions. Pratyush is technical co-founder (PhD ETH Zurich; ex-Microsoft Research/IBM). Pain points: reliability & cost of agentic workflows at very large scale; multilingual agent behavior; full-stack sovereign build. Challenges: scaling agentic workflows across billions of interactions without cost blowout or reliability regressions. Must-haves: cost efficiency + reliability at production scale. Nice-to-haves: unified observability across agentic workflows. ICP confidence: High — clearly building/shipping agents in production, technical decision-maker, in-range headcount. (No verified personal LinkedIn slug captured, so profile_url left blank to avoid fabrication.)
Preeti Somal — Senior Vice President of Engineering, Temporal Technologies
Signal: 1 (ICP publicly writing/speaking about agent cost, reliability and control). Source: https://venturebeat.com/orchestration/ai-agents-are-entering-their-rebuild-era-as-enterprises-confront-the-reliability-problem (VentureBeat AI Impact Series, 29 May 2026). Company size: ~350-450 (Temporal Technologies, Series C durable-execution/workflow-orchestration company; Abridge is a named customer). Pain points: teams shipped v1 agents fast and skipped the plumbing, so they 'crash and burn' and are now rebuilding v2.0 of the same agent; a late-stage crash forces a rerun of the whole agent flow and re-pays every prior model call ('the token tax'); opaque spend across multi-step agents that call many models, tools and APIs; conflation of execution state vs context/memory; lift-and-shift AI adoption producing cloud-era-style spend without value. Challenges: making long-running (hours-to-days) multi-service agent workflows survive crashes, preserve state and resume from the point of failure; giving enterprises step-by-step visibility into where tokens are consumed inside one agent run; governance and model-selection policy without taking an off-the-shelf agent platform. Must-haves: durable execution + recovery-from-crash-point so failed runs are not re-paid for; single-pane cost/token visibility across a whole multi-step agent flow; observability and governance guardrails. Nice-to-haves: standardized internal 'paved path' frameworks, identity systems, richer memory/context management. ICP confidence: High — SVP Engineering (VP level) at a 50-2,000 employee infrastructure company whose customers are shipping agents in production, and she articulates the exact pain thesis (cost per run, token tax, agent visibility, 1->5+ agent scaling wall) in her own words. Verbatim signals: 'What you care most about is making sure that you can recover and that you're not paying the token tax if something goes wrong.' / 'You can now see where you're spending the tokens in an agent that is multiple steps and calling multiple different systems.'
Priyank Chhipa — Engineering Leader (Knowledge Team & India Operations), 1mind
Signal: 4 (senior engineering leader at a company actively shipping AI agents; found via LinkedIn people search + web verification of size/stage).
Source: https://www.linkedin.com/in/priyankchhipa/ ; 1mind funding/headcount coverage (PRNewswire/TechCrunch/GetLatka 2026).
Company: 1mind — multimodal AI "Superhuman" agents for B2B GTM/sales (qualify leads, run demos, handle objections, onboard) across web/product/live video. Founded by Amanda Kahlow (ex-6sense).
Company size: 72 employees (May 2026).
Stage: Series A ($30M, Battery Ventures; $40M total).
Pain points (INFERRED from role + company, NOT verbatim): running always-on multimodal agents (voice+video+GTM brain) reliably in production; cost per interaction/run; scaling agent fleet across many customers.
Challenges (inferred): reliability & latency of live agent interactions; controlling LLM/inference spend at scale.
Must-haves (inferred): production reliability, per-run cost governance/observability.
Nice-to-haves (inferred): context/token optimization, eval tooling.
ICP confidence: Medium (Engineering Leader — Director-level — at 50-2,000 emp Series A agent-native company; exact title seniority slightly below VP).
Prukalpa Sankar — Co-Founder & Co-CEO, Atlan
Signal: 4 (ICP speaking about infrastructure for production agents; AI Engineer World's Fair 2026 speaker). Source: https://www.ai.engineer/worldsfair/schedule — talk "WTF Is the Context Layer? The Missing Infrastructure for Production Agents." Company size: ~558-572 employees, Series C ($105M, $750M valuation), AI-native metadata/governance platform positioned as "the context layer for AI agents." Pain points: giving AI agents trusted context/data at scale, missing infrastructure for production agents, governance/trust for agent data access. Challenges: making enterprise data reliable and governed enough for autonomous agents; production-readiness of the agent context layer. Must-haves: trusted context + governance for agents in production. Nice-to-haves: observability into how agents consume context/data. ICP confidence: Medium — technical co-founder/Co-CEO at a right-sized Series C AI-native company focused squarely on production-agent infrastructure; caveat: Atlan's wedge is context/governance (adjacent to, not identical to, Alpha's cost/reliability control), and Prukalpa's role leans product/GTM. Note: web research + verification; LinkedIn not connected this run — profile URL not verified so left blank rather than guessed.
Puneet Agarwal — SVP of Engineering, Observe.AI
Signal: 4. Source: Observe.AI press release "Expands Leadership Team with Puneet Agarwal as SVP of Engineering" (https://www.observe.ai/press-releases/observe-ai-expands-leadership-team-with-puneet-agarwal-as-svp-of-engineering). NOTE: role/company confirmed via press release; profileUrl (https://in.linkedin.com/in/agarwal-puneet, Bengaluru) is the probable LinkedIn match (Observe.AI R&D is Bengaluru) but not 100% certain. Company size: ~344-407 employees; Series C ($125M Series C, $213M total). Shipping autonomous VoiceAI Agents that handle complete customer interactions. Pain points: reliability + latency of production voice agents (billing disputes, resets handled autonomously); cost per interaction; scaling agent engineering across enterprise contact centers. Challenges: scaling reliable autonomous voice agents in production. Must-haves: reliability, low latency, observability. Nice-to-haves: cost-per-run visibility. ICP confidence: Medium (SVP Eng confirmed; LinkedIn URL probable, not certain).
Puneet Mehta — Founder & CEO (technical founder), Netomi
Signal: 1 (ICP speaking publicly about context management and runtime governance for agents).
Source: https://openai.com/index/netomi/ (OpenAI customer story, 8 Jan 2026); title verified https://theorg.com/org/netomi/teams/leadership-team and https://www.unite.ai/puneet-mehta-founder-and-ceo-of-netomi-interview-series/
Company size: not reliably established — Tracxn 140 (Jan 2025), The Org band 51–200, Getlatka 208 (undated), LeadIQ ~259 (Jun 2026). Best honest estimate 150–260 — inside band but LOW confidence on the exact figure. Series C, $110M May 2026 led by Accenture Ventures.
Pain points: (verbatim) "In airlines, context changes by the minute. AI has to reason about the scene the customer is in—not just execute a siloed task. That's why situational awareness matters way more than just workflows, and why a context-led ensemble architecture is essential." Also (verbatim) "We built the system so that if the agent ever reaches uncertainty, it knows exactly how to back off safely. The governance is not bolted on—it's part of the runtime." And (verbatim) "Our goal was to orchestrate the many systems a human agent would normally juggle and do it safely at machine speed."
Challenges: Context freshness and situational awareness across many source systems; safe degradation under uncertainty; orchestrating an ensemble of agents at machine speed.
Must-haves: Runtime (not bolted-on) governance; context-led architecture that keeps agents grounded as state changes; safe fallback behavior.
Nice-to-haves: Per-orchestration cost and latency attribution across the ensemble.
ICP confidence: Medium — technical founder-CEO at a Series C company shipping agents at enterprise scale with three strong on-topic verbatim quotes, but headcount could not be confirmed from a primary source (four third-party figures disagree). Confirm headcount before prioritizing.
Note: Netomi has no publicly verifiable VP Eng / Head of AI layer; a "Bobby Gupta, CTO" appears on third-party listing sites only and is absent from theorg's Netomi org chart (Bobby Gupta is already in the People Library — treat that record's title as unverified).
Quentin Rousseau — Co-Founder & CTO, Rootly
Signal: 4 (ICP CTO building/shipping agents; found via 2026 agent-launch sweep, KubeCon EU 2026 'AI SRE On-Call Agents' demo). Source: https://www.linkedin.com/in/quentinrousseau/ , https://www.ycombinator.com/companies/rootly , KubeCon EU 2026 (youtube.com/shorts/9zYmYmjUx5Q). Company size: ~76 employees (Tracxn, Jun 2026); SF; Series A, ~$15.2M raised. Rootly = AI-native on-call + incident management with AI SRE agents that detect/investigate/help-resolve production incidents. Pain points (INFERRED from product surface): autonomous SRE agents acting on live production incidents need reliability + guardrails; trust that an on-call agent's remediation is safe; MTTR pressure. Challenges (INFERRED): letting agents execute actions in prod without new risk. Must-haves (INFERRED): deterministic/auditable record of what the agent did during an incident, reliability at scale. Nice-to-haves (INFERRED): per-incident/per-run cost visibility as agent count grows. ICP confidence: HIGH (named technical co-founder/CTO; in-band headcount; clearly shipping named AI agents in production).
Quinn Slack — Co-Founder & CEO (technical) — CEO of Amp (AI coding agent), co-founder Sourcegraph, Amp / Sourcegraph
Signal: 4 (ICP technical co-founder building/shipping coding agents; found via web research — LinkedIn search blocked this run). Source: https://sourcegraph.com/blog/why-sourcegraph-and-amp-are-becoming-independent-companies ; https://www.crunchbase.com/person/quinn-slack ; https://www.linkedin.com/in/quinnslack . Company size: Sourcegraph ~148 employees (May 2026); Amp Inc spun out as standalone company (Dec 2025) with the Amp team. Stage/funding: Sourcegraph raised $200M+, ~$2.6B valuation. Actively building/shipping agents: Amp is Sourcegraph's AI-native coding agent (autonomous multi-step coding); Sourcegraph also ships Cody. Quinn Slack took over as CEO of Amp Inc while remaining Sourcegraph co-founder/board. Technical co-founder (engineer; built Sourcegraph). NOTE: DIFFERENT person from existing Sourcegraph record Beyang Liu (CTO, id 47) and Erika Rice Scherpelz (Head of Eng) — net-new. Pain points (role-inferred + adjacent to Beyang Liu's public statements): token waste and inference cost in coding agents; cost/quality tradeoff across sub-agents and multiple frontier models; agent reliability across the SDLC. Challenges: controlling cost of agentic coding at scale; multi-model orchestration. Must-haves: token/inference cost visibility, multi-model routing control, replayable agent traces. Nice-to-haves: cross-model routing efficiency. ICP confidence: High (technical co-founder/CEO of an AI coding-agent company; right size).
Raafat Zarka — Director of Software Engineering, Writer
Signal: 4 (company-scoped LinkedIn people search — ICP technical leader at agent-shipping company). Source: https://www.linkedin.com/in/raafatzarka (via https://www.linkedin.com/company/getwriter/people/?keywords=engineering). Headline: 'Director of Software Engineering @ Writer | Ex-Microsoft, Salesforce, Tableau | PhD | Data-AI Platforms & GenAI Infrastructure | Enterprise LLM Systems'. Company size: 201-500 employees (Writer; enterprise generative-AI + agentic platform / AI agents, Series C). Pain points (INFERRED from role+company): cost and reliability of enterprise LLM/agent infrastructure; token/GenAI-infra efficiency; observability of agentic workflows in production. Challenges: building reliable, cost-efficient GenAI infra + enterprise LLM systems that power agents at scale. Must-haves: per-run cost + reliability observability across agent infra. Nice-to-haves: model routing / infra cost optimization. ICP confidence: High (Director of Software Engineering owning GenAI infra/LLM systems at an independent 201-500-emp agent-shipping SaaS. Brain already had Waseem AlShikh (CTO) and Dan Bikel (Head of AI); this is net-new).
Raaghu K — Senior Director of Engineering, Level AI
Signal: 4 (ICP engaging with content about scaling agents beyond isolated deployments).
Source: https://www.linkedin.com/in/raaghu81/recent-activity/all/ — reposted an "AI For Business Leaders" piece on Level AI expanding its full-stack agentic CX platform "beyond isolated virtual agents to create a unified intelligence loop where human and AI agents share real-time learning, training and quality standards." The piece quotes Level AI CEO Ashish Nagar: "A virtual agent operating in a silo is no longer enough for the modern enterprise," and a customer saying they "made more progress in four weeks than we had in the previous nine months" after synthesising 12,000+ technical assets. Company page: https://www.linkedin.com/company/level-ai/about/
Company size: 51-200 employees (LinkedIn company page, verified this run). AI-native, agentic CX platform in production.
Pain points: The pattern he amplified is the 1→many agent problem — isolated virtual agents that do not share learning, training data or quality standards across the fleet. Not his own words; he chose to amplify it.
Challenges: Keeping quality and learning consistent across human and AI agents operating on the same live conversations.
Must-haves: A shared quality/learning loop across agents rather than per-agent silos.
Nice-to-haves: Faster time-to-value on new agent deployments (the 4-weeks-vs-9-months framing).
ICP confidence: High — Senior Director of Engineering (at the Director+ bar), 51-200 employee AI-native company shipping agents. Second contact at Level AI alongside Vikram Verma (VP Eng) — treat as the same account, two entry points.
Rachad Alao — Vice President of Product Engineering, Cohere
Signal: 1 (ICP speaking publicly about exponential agent token consumption and controlling the agent stack). Source: https://venturebeat.com/technology/cohere-vp-says-enterprise-ai-sovereignty-requires-control-of-the-full-agent-stack (VB Transform 2026 fireside, 15 July 2026). Company size: ~400-600 (Cohere, enterprise AI/agent platform; NOTE: Cohere already appears in the brain via Aidan Gomez — Alao is a net-new person, not a duplicate). Pain points: token utilization rising exponentially as customers move from chatbots to multi-step agents that reason, call tools and search internal systems; falling per-token prices are being outrun by consumption growth; misalignment of vendor incentives when vendors bill on token utilization; enterprises sending every request to the largest frontier model. Challenges: routing work to the right model by required intelligence, data sensitivity and regulatory burden; keeping control of the whole stack (GPUs, private cloud, governance/routing layer, connectors, search, agent frameworks) without vendor lock-in; making retrieval a first-class tool inside the agentic workflow. Must-haves: model routing / governance layer that picks the right model per task; control over where data resides and where AI operations run; avoiding vendor lock-in. Nice-to-haves: multimodal search integrated into agent workflows, compressed/4-bit private deployments, smaller task-specific models. ICP confidence: High — VP of Product Engineering at a 50-2,000 employee AI-native company shipping agents, previously led responsible-AI and trust & safety engineering at Google and Meta, and on record about token blowout and agent control. Verbatim signals: 'Your token utilization is going exponentially up, because you're dealing with more and more complex agentic use cases.' / 'You want to have control on the entire stack.' / 'Use the right model for the task at hand.'
Rachel Rivera — Director of Platform Engineering, Ambience Healthcare
Signal: 4 (ICP technical leader at agent/copilot company). Source: https://www.linkedin.com/in/rachelrivera1/ (LinkedIn, 2026). Company size: ~200-500 emp (Ambience Healthcare; $243M Series C, Oak HC/FT + a16z; ships ambient AI + Chart Chat conversational AI copilot + ICD-10 coding AI). Pain points (inferred from role + company posts, labeled as such): reliability & auditability of clinical AI (compliance 'at the core', audit-proof coding); chart-aware synthesis across labs/imaging/notes; accuracy vs physician benchmarks. Challenges: platform engineering for compliant, reliable healthcare agents; scaling ambient + conversational AI across health systems (Mayo, Cleveland Clinic). Must-haves: reliability/eval + audit trail for agents in production; cost visibility per encounter. Nice-to-haves: context-efficiency across long clinical charts. ICP confidence: High (Director of Platform Engineering at in-range healthcare-AI/agent company; verified current).
Raghu Ravinutala — Co-Founder & CEO, Yellow.ai
Signal: 4 (ICP building/shipping conversational AI agents at enterprise scale). Source: https://councils.forbes.com/profile/Raghu-Ravinutala-CEO-Co-founder-Yellow-ai/a7d7b3a3-c037-493c-819b-9f9e303b2beb ; https://growthcapadvisory.com/worlds-leading-conversational-ai-platform-yellow-ais-raghu-ravinutala/ . Company size: ~700+ employees (raised $102M+; 1000+ customers across 85+ countries). Pain points: scaling gen-AI customer-service agents reliably across a very large, global customer base; automation quality/consistency at enterprise scale. Challenges: multi-region, multi-language agent deployments; maintaining reliability across 1000+ production deployments. Must-haves: reliable automation at scale, consistent agent performance across customers. Nice-to-haves: broader channel/language coverage. ICP confidence: High (technical co-founder/CEO, ex-TI/Broadcom engineering, at 700-person AI-native company shipping agents in production).
Raghuveer Kancherla — Co-founder (technical; drives AI architecture; ex Co-Founder & CTO of Recruiterbox), Sprinto
Signal: 4 (ICP speaking publicly about shipping agents in production). Found via web research — LinkedIn/Chrome unavailable this run.
Source: https://yourstory.com/2025/12/how-sprinto-is-building-the-worlds-first-autonomous-compliance-engine
Company size: 316 employees as of 25 Jul 2026 (Tracxn; a second source indicated ~442). Series B, $20M. Bengaluru, India. AI-native GRC / compliance automation. ~3,000 customers across 75+ countries (incl. Whatfix, Bizongo, Happay, Rocketlane).
Agent evidence: shipped AI Playground, a no-code workspace where customers build custom compliance agents — 100+ agents created by early adopters within weeks of launch, 5,000+ hours claimed saved; one customer automated nearly its whole vendor due-diligence workflow (document review, risk scoring, exception routing) and doubled throughput without adding headcount. Also ships "Ask AI" (claimed 80%+ audit-grade accuracy), a Fix It Agent that executes browser-level security fixes but stops and asks before impactful changes, and Infinite Frameworks for continuous regulation monitoring/control mapping. Built a structured graph "context engine" over every GRC entity (risks, vendors, policies, systems, incidents, access controls).
Verbatim: "Legacy GRC tools centralise data. They don't drive action." Hard rule he states: "if the AI is unsure, it must ask the user." Internal mantra: "humans stay in command; AI handles the rest."
Pain points: partial automation no longer satisfies customers — they want systems that understand context, decide and act; compliance evidence work still manual and repetitive; emerging "Shadow AI" risk from employees feeding sensitive data into external AI tools.
Challenges: getting an agent to reason over the relationship graph between risks, vendors, policies, incidents and access; hitting audit-grade accuracy where a wrong answer is a compliance failure; real-time policy/confidence checks on every AI interaction; calibrating when an agent should act vs escalate.
Must-haves: structured context/knowledge graph underneath the agents; human approval gates on every major action; explainable recommendations with sources; strict data isolation (customer data never trains models); ISO 42001 alignment; abstention behaviour when confidence is low; 300+ SaaS integrations.
Nice-to-haves: no-code agent authoring for non-engineer compliance teams; real-time controls over employee AI usage (Shadow AI governance); continuous regulatory-change monitoring auto-mapped to existing controls.
ICP confidence: High — Series B, ~316 employees, technical co-founder speaking in detail about agents already running in production for 3,000 customers. Roadmap over next 12–18 months is explicitly "more context-aware agents and an autonomy layer."
Signal: 4 (ICP VP of Product (AI-focused) at agent-native company). Source: https://theorg.com/org/aisera/org-chart/rahul-guha ; https://www.linkedin.com/posts/aisera_icymi-rahul-guha-aiseras-vp-of-product-activity-7292258104329613313-f10S . Company size: ~250-500 (Aisera, agentic AI enterprise). Pain points (inferred, not quotes): shipping reliable enterprise AI agents customers trust; demonstrating ROI/cost-per-resolution; avoiding token/context waste across agent workflows. Challenges: balancing agent quality vs. cost while scaling agent count. Must-haves: reliability in production, per-run cost + outcome visibility. Nice-to-haves: analyst-friendly reporting on agent economics. ICP confidence: High (VP Product at an in-range agent-native company; ex-ServiceNow; verified current).
Rahul Sengottuvelu — Head of Applied AI, Ramp
Signal: 4 (ICP leading production agent work at a qualifying company). Source: https://theorg.com/org/ramp/org-chart/rahul-sengottuvelu ; Ramp blog https://ramp.com/blog/introducing-ramp-applied-ai-solutions ; PRNewswire "Ramp Launches Fleet of AI Agents Across Its Procurement Platform" (2026). Company size: ~1,000–1,500 employees (fintech, within cap). Pain points: deploying a FLEET of AI agents across finance ops (accounts payable, procurement, monthly close) reliably; forward-deployed engineers building bespoke agents embedded in enterprise finance teams; scaling from a few agents to many. Challenges: reliability of compliance-sensitive financial agents; cost of multi-cycle reasoning agents; scaling agents across many enterprise customers with auditability. Must-haves: reliability + auditability of finance agents; cost visibility per agent run. Nice-to-haves: standardized agent infrastructure / observability layer. ICP confidence: High — Head of Applied AI and clear decision-maker at a ~1,000+ person company shipping a documented fleet of production agents. Profile URL left blank (not fabricating LinkedIn URL).
Raj Arokiaraj — VP of Engineering, Ushur
Signal: 4 (company-scoped ICP people search on Ushur; found via LinkedIn people search, NOT an observed post — pains INFERRED from role+company). Source: https://www.linkedin.com/search/results/people/?keywords=Ushur%20director%20VP%20engineering%20AI%20agents | Company size: ~300 (Ushur, enterprise agentic-AI automation for insurance/healthcare CX, Series C, SF Bay Area). Pain points (inferred): shipping and scaling enterprise AI agents in production; per-run cost visibility and reliability across regulated CX workflows. Challenges (inferred): reliability/governance for enterprise agents; cost control as agent volume grows. Must-haves (inferred): production reliability + per-run cost/attribution visibility. Nice-to-haves (inferred): context/token-waste reduction, model routing. ICP confidence: High (VP of Engineering at agent-native company, 50-2,000 emp; title+company verified on LinkedIn, pains inferred).
Raj Kumar Dubey — Head of AI Platform (India), Clari
Signal: 4 (AI-platform leader at a company shipping AI agents, found via LinkedIn currentCompany-facet People search). Profile: https://www.linkedin.com/in/raj-kumar-dubey/ . Source: https://www.linkedin.com/search/results/people/?currentCompany=%5B%222706103%22%5D (Clari). LinkedIn headline lists "Clari/Salesloft". Company size: ~700 for Clari (estimate); NOTE the combined Clari/Salesloft entity may exceed 2,000 - flagged for verification. Pain points (inferred from role + product): owns AI platform underpinning Clari Agents - platform-level LLM cost control, per-run cost visibility, reliability/observability of agents in production, token/context efficiency. Challenges: building shared agent platform infra that keeps cost + reliability under control at scale. Must-haves: platform-level cost-per-run observability and guardrails. Nice-to-haves: model routing, cost attribution by team/agent. ICP confidence: Medium (Head of AI Platform - strong technical fit and 2nd-degree connection; downgraded to Medium only due to merged-entity headcount uncertainty). Pain points INFERRED from role+product, not fabricated.
Raj Neravati — Founder, Nexora
Helps with: He has lot of industry connects and available only to provide answers to pointed questions but not advisory
Reached out to him for Advisory, but rejected the request but available for pointed questions-answers.
Rajat Awasthi — AI/ML Leader (Director-level; ex-Associate Director AI&ML at Sprinklr), Sirion (SirionLabs)
Signal: 4 (senior AI/ML leader at a CLM SaaS building agentic contract AI; found via LinkedIn people search 'Director of AI ... agents production', page 2; profile confirms prior title Associate Director, AI & ML at Sprinklr, now leading AI/ML at Sirion). Source: https://www.linkedin.com/in/rajatawasthi-ai/ . Company size: ~1,000-1,300 employees (SirionLabs, Series C+ contract-lifecycle-management SaaS) — estimate, in-range. Sirion is building agentic AI for contract review/extraction/negotiation. Pain points (inferred): reliability of contract-analysis agents on large legal documents; controlling LLM/token cost across enterprise contract volumes; accuracy + auditability for enterprise legal customers. Challenges: hallucination control on contracts; scaling from pilots to many production agents; per-run cost attribution. Must-haves: production reliability, cost visibility per agent/run, monitoring. Nice-to-haves: evals, model routing. ICP confidence: Medium (Director-level AI/ML leader — headline title 'AI/ML Leader' is slightly ambiguous but prior role was Associate Director; in-range agent-shipping CLM SaaS).
Signal: 4 (adapted method — found via LinkedIn company People-page directory of an in-band agent-native company; LinkedIn content/post search was low-yield this run, consistent with prior runs). Source: https://www.linkedin.com/company/perplexity-ai/people/ and https://www.linkedin.com/in/rajatraina/. Company size: 201-500 employees (LinkedIn header). In-band, agent-native (ships answer engine, Comet browser agent, Deep Research agent in production). Pain points (INFERRED from role+company, no quotes): per-query/per-agent-run LLM cost at consumer scale (agentic answer engine + Comet agentic browser + Deep Research agent, millions of runs); token/context efficiency; reliability & citation accuracy of agent outputs; latency; near-zero per-run/per-agent cost visibility. Challenges: keeping unit economics sane as agent usage compounds; orchestrating multi-step search->browse->research agents reliably at scale. Must-haves: per-run/per-agent cost observability; reliability guardrails at massive scale. Nice-to-haves: model routing/cost optimization; agent governance. ICP confidence: High (VP Eng at in-band agent-native company).
Rajesh Gupta — Head of Agentic AI (Product & Engineering), Skan AI
Signal: 1 (Head of Agentic AI authoring 'The Hidden Cost of Almost Right AI Agents'). Source: https://www.linkedin.com/in/rajesh-gupta/. Company size: ~90-270 employees, Series B (process intelligence platform; building agentic AI across product & engineering). Ex co-founder/CEO of Metaculars (acquired by Skan 2025). Pain points: hidden cost of 'almost right'/unreliable agents; agent accuracy-vs-cost economics; cost per agent outcome. Challenges: making agents reliable enough to ship; controlling agent cost vs value. Must-haves: agent reliability, cost-per-outcome visibility. Nice-to-haves: process-intelligence-driven agent optimization. ICP confidence: High (Head of Agentic AI at right-sized Series B company actively building/shipping agents; direct cost signal).
Signal: 4 (ICP CTO hired specifically to lead AI/agent platform at a qualifying company shipping many agents). Source: https://www.salesloft.com/company/newsroom/clari-salesloft-doubles-down-on-revenue-execution-and-ai-capabilities-with-cro-and-cto-appointments ; https://www.businesswire.com/news/home/20260513426805/en/ (appointment announced May 2026) ; https://www.prnewswire.com/news-releases/salesloft-launches-15-new-ai-agents-to-drive-pipeline-efficiency-and-full-cycle-sales-execution-302453796.html ; https://www.globenewswire.com/news-release/2026/07/14/3326944/0/en/ (Jul 2026). Company size: ~1,200-1,800 combined post-merger (Clari + Salesloft merged Dec 2025). Pain points: inherited a merged two-platform engineering org that must run 26 AI agents available or in development across the revenue cycle — a textbook 1->many agent scaling wall; agents must fire off live buyer signals in real time. Challenges: unifying Clari and Salesloft architectures under one 'Predictive Revenue System'; operating a large agent fleet reliably and economically; proving agent ROI to enterprise revenue buyers. Must-haves: platform-level control and observability across a 26-agent fleet; predictable unit economics per agent action. Nice-to-haves: consolidated agent governance and model routing. ICP confidence: High on role/company (CTO, 50-2,000 employees, 26 agents in flight, AI explicitly in his mandate); Medium on the specific cost/reliability pain — inferred from fleet size and public product scope, NOT from a direct quote. Profile URL is a best public match (Oakville, Ontario — consistent with his Dayforce background) and should be re-verified before outreach.
Rajesh Raheja — Chief Technology Officer, Duck Creek Technologies
Signal: 4 (ICP CTO at a company that launched a platform whose stated purpose is deploying, ORCHESTRATING and GOVERNING AI agents). Source: https://www.duckcreek.com/resource/press-releases/launch-agentic-ai-platform-underwriting-claims/ ; https://www.duckcreek.com/blog/duck-creek-technologies-announces-veteran-enterprise-technology-leader-rajesh-raheja-as-chief-technology-officer/ ; https://iireporter.com/duck-creek-names-rajesh-raheja-cto/ ; https://fintech.global/2026/04/29/duck-creek-launches-agentic-ai-platform-to-transform-insurance-workflows/ ; https://www.prnewswire.com/news-releases/duck-creek-acquires-send-creating-the-industrys-only-agentic-underwriting-to-core-platform-302819390.html ; headcount 1,891 (PitchBook 2026) and 1,878 (24 Apr 2026). LinkedIn: NOT CONFIRMED this run — do not guess a slug; resolve before any outreach. Company size: ~1,878-1,891 — inside the 2,000 cap but close to it; RE-CHECK before outreach, the Send acquisition may push them over. Vertical: P&C insurance core systems (policy, billing, claims); Vista-backed, formerly NASDAQ-listed. Agent evidence: Duck Creek Agentic AI Platform launched 28 Apr 2026, explicitly 'purpose-built to enable insurers to deploy, orchestrate, and govern AI agents across the insurance lifecycle', using neuro-symbolic reasoning so agents 'operate within the constraints of insurance workflows and current carrier configurations'; shipped with two agentic applications (Agentic Underwriting Workbench, Agentic FNOL) plus an Agentic Product Configurator, and Duck Creek then ACQUIRED Send to build an 'agentic underwriting-to-core platform'. Raheja appointed CTO Dec 2025 owning global technology strategy, platform innovation and the engineering org. Pain points (INFERRED from public product/positioning, not verbatim quotes): their entire pitch is CONSTRAINING agent behaviour to carrier configuration — that is a control problem stated in their own marketing; agents run inside carriers' regulated underwriting and FNOL workflows where an unexplainable action is a compliance event. Challenges (INFERRED): they are shipping an agent-governance layer to enterprise insurers while simultaneously integrating an acquisition; agent count and run volume scale with carrier customers, not with Duck Creek's own usage. Must-haves (INFERRED): provable, per-action audit and replay of agent decisions to satisfy carrier risk and regulatory review; predictable cost as carriers turn agents on across books of business. Nice-to-haves (INFERRED): cross-carrier benchmarking of agent behaviour. ICP confidence: HIGH on role and agent activity, MEDIUM on fit overall. CAVEATS BEFORE OUTREACH: (a) headcount is within ~110 of the hard cap and rising via M&A; (b) PE-owned mature vendor, not Series A-C — long procurement cycle; (c) COMPETITOR-ADJACENT — they sell their own agent orchestration/governance layer, so this must be positioned strictly as the run-level control and cost plane UNDERNEATH their platform, never as a replacement for it.
Rajesh Veerappan — VP, AI Engineering, Regal.ai
Signal: 4 (ICP eng leader at an agent-shipping company; found via LinkedIn people-search 'Regal.ai engineering'). Source: https://www.linkedin.com/in/rajeshveerappan/ . Company size: ~100-200 (Regal.ai; VC-backed AI phone/voice-agent platform for outbound + inbound customer calls). Pain points (INFERRED from role + company; no verbatim quote observed this run): reliability & latency of real-time voice agents at contact-center scale; per-call / per-run cost control as call volume grows; lack of visibility into what each voice agent costs per run. Challenges: running many concurrent voice agents reliably in production. Must-haves: per-run cost observability + reliability guardrails for voice agents. Nice-to-haves: model routing / cost optimization. ICP confidence: High (VP-level AI engineering leader at a 50-2,000-emp company actively shipping voice AI agents).
Ram Venkatesh — Co-founder & CTO, Sema4.ai
Signal: 4 (technical co-founder building & shipping enterprise AI agents in production; surfaced via search for named CTOs at agent-native companies).
Source: https://www.snowflake.com/en/blog/startup-spotlight-sema4/ ; https://startupintros.com/orgs/sema4-ai ; https://sema4.ai/blog/enterprise-ai-agents-platform-release-2026/
Company size: ~50–60 employees (about half in Atlanta); ~$55.5M raised, Series A extension $30.5M June 2025. At/just above the 50-employee floor.
Background: Ex-CTO of Cloudera; co-founder & CTO of Sema4.ai, an enterprise AI-agent platform for document/data-heavy back-office processes requiring deterministic, auditable outcomes.
Pain points (inferred from role/company focus): making agents deterministic, auditable and reliable in production for regulated back-office work; controlling multi-step agent behavior.
Challenges: guaranteeing correctness/auditability of agents on complex enterprise processes.
Must-haves: agent reliability, auditability/governance, run-level visibility.
Nice-to-haves: cost attribution per process, reusable agent building blocks.
ICP confidence: Medium (strong role fit — technical co-founder/CTO actively building agents — but company is right at the 50-employee floor; verify current headcount before outreach).
Signal: 1 (ICP company facing agent cost/reliability economics). Source: Fortune (Series E, $2B val) + BVP profile; profile_url is best-guess LinkedIn slug (verify). Company size: 100+ eng/product staff, est. ~350-600 total employees; Series E, $2B+ val, ~$285M raised. Co-founder & CEO (technical/product founder). EvenUp ships AI agents for personal-injury law via its "Claims Intelligence Platform" powered by proprietary Piai model (drafting, review, strategy over hundreds of thousands of cases + millions of medical records). Pain points: rising compute costs vs customer pricing; heavy-usage customers drive up inference bills; churn risk if value isn't quantifiable. Challenges: triaging compute cost against pricing so heavy usage doesn't "drown" margins; reliable, high-quality output so law-firm customers don't churn. Must-haves: compute/inference cost control tied to usage, reliable output quality. Nice-to-haves: per-customer cost visibility/attribution. ICP confidence: High on company (vertical AI, heavy compute, 50-2,000 band, shipping agents in production); Medium on exact profile URL (unverified slug). NOTE: confirm LinkedIn URL before outreach.
Signal: 3 (ICP engaging with competitor/adjacent content — AWS Bedrock AgentCore launch partner) + 4 (publicly writing about running agents in production).
Source: https://caylent.com/blog/98-of-enterprise-leaders-would-let-ai-agents-run-production-under-the-right-conditions-caylent-survey-reveals (6 Aug 2026, direct quote); name/title/LinkedIn verified on https://caylent.com/about; agent evidence https://caylent.com/blog/caylent-achieves-the-aws-agentic-ai-specialization and https://aws.amazon.com/blogs/apn/aws-partners-demonstrate-enterprise-ai-agent-solutions-with-amazon-bedrock-agentcore
Company size: ~913 (Revelio Labs, 26 Jun 2026); 950 (PitchBook). Growth 575 (2023) → 802 (2024) → 985 (2025) → ~913. Well inside 50–2,000.
Pain points: Verbatim — "The question of whether enterprises will adopt agentic AI is settled. What's left is authority, not accuracy. Model accuracy isn't the hardest part of the problem anymore. It's how much latitude to give an agent, how every action gets approved, and what happens when it's wrong. Velocity comes from allowing humans in the loop to rapidly approve agentic behaviors and minimize risk with guardrails and systems designed for agentic interactions." Pre-AgentCore-migration pain documented in the AWS write-up: high developer cognitive load maintaining custom memory systems, 30+ second time-to-first-token, and scaling to 50x growth.
Challenges: Built CloudZero Advisor — a production multi-agent FinOps platform orchestrating five specialized agents on AgentCore Runtime with AgentCore Memory and Gateway; migrated off Amazon Bedrock Agents. Also ships Caylent Accelerate for Agentic Cloud Operations (claims 70% of remediation work expedited, MTTR down 40%). As a services firm they run agents across many customer environments simultaneously, which multiplies the operating-layer problem.
Must-haves: Agent authority/permission model — how much latitude an agent gets, how each action is approved, what happens on failure. Guardrails and human-in-the-loop approval that do not kill velocity. Latency control (they had a 30s TTFT problem).
Nice-to-haves: Per-run cost attribution across customer environments; reduced cognitive load of maintaining bespoke agent memory infrastructure.
ICP confidence: High — CTO, headcount solidly in band, agents demonstrably in production at customer scale, and an on-the-record August 2026 statement about agent authority, approval and failure handling, which is precisely thealpha.ai's problem space. Caveat: Caylent is a services/consulting firm (AWS Premier Tier Partner, Anthropic Preferred Services Partner) building agents for clients, so the buying motion differs from a product company — but that also makes them a potential channel.
Signal: 4 (ICP speaking at AI Engineer World's Fair 2026; WSO2 shipping enterprise agent platform). Source: https://www.ai.engineer/worldsfair/schedule ; verified via wso2.com. Company size: ~800-1,250 employees (established integration vendor, PE-backed). WSO2 launched 'Agent Manager' / Agent Platform — a control plane to identify, govern, secure and scale enterprise AI agents. Pain points (inferred from product): enterprises lack identity/governance/observability over agents, no control of agent behavior at scale, cost/visibility gaps. Challenges: bringing production-grade governance and scale to enterprise agents. Must-haves: agent observability, governance, and control. Nice-to-haves: per-agent cost attribution. ICP confidence: Medium — C-suite technical AI leader (ex-IBM Research Director of AI Engineering) at an ~800-1,250-person company actively building an agent platform; in size band but larger/later-stage than the A-C core target.
Raoul Felix — Chief Technology Officer, Amplemarket
Signal: 4 (ICP CTO at a company shipping an AI sales agent in production; found via web research — LinkedIn/Chrome not connected this run). Source: https://www.amplemarket.com/blog/new-cto (Amplemarket welcomes Raoul Felix as CTO) ; https://www.amplemarket.com/blog/introducing-amplemarket-duo-the-first-end-to-end-ai-sales-copilot ; https://www.amplemarket.com/duo . Company size: ~100–110 employees (across five continents); San Francisco / Portugal; founded 2018 (YC); product 'Duo' is an end-to-end AI sales agent/copilot for signal-based selling. Pain points: reliability of autonomous outbound/sales agents acting on prospect data; cost/efficiency of multi-step agent runs at outbound scale; quality/deliverability control. Challenges: making sales agents reliable and controllable at volume; per-run cost and observability. Must-haves: reliability, control/guardrails, cost efficiency. Nice-to-haves: per-run cost visibility, context/token efficiency. ICP confidence: Medium (CTO at a ~100-emp company shipping an AI sales agent; in the 50–2,000 band and clearly technical, though funding is lighter/earlier — ~$12M — than a typical A–C target).
Rashi Agrawal — Head of Agentic AI, Hinge Health
Signal: 1 (ICP "Head of Agentic AI" publicly speaking about controlling agent behavior in production). Source: AI Engineer World's Fair 2026 (speaker/session data: https://www.ai.engineer/worldsfair/2026/speakers.json) — talk "Guardrails First: Engineering Member-Facing Health AI" (member-facing health AI cannot iterate loosely; needs guardrails first). Company size: ~1,664 employees (within 50–2,000); note: publicly traded (NYSE: HNGE) — beyond A–C stage, lowering confidence. Ships AI agents (e.g., "Robin" AI care assistant). Pain points: controlling/guardrailing member-facing agent behavior at scale; reliability and safety in production. Challenges: engineering guardrails and predictable behavior for consumer-facing health agents. Must-haves: control and guardrails over agent behavior; visibility into agent actions. Nice-to-haves: automated guardrail evaluation. ICP confidence: Medium (perfect persona — Head of Agentic AI — and strong control pain at a ~1.6k-person company, but the company is public, past the A–C band).
Signal: 4 (ICP at company running agents in production; boost.ai reports 600+ live AI agents and 150M+ conversations/year; Rasmus authors/posts on AI/ML leadership and responsible AI). Source: https://dk.linkedin.com/in/rasmushauch and https://boost.ai/authors/rasmus-hauch/ ; company profile https://boost.ai/about . Company size: 51-200 employees (Stavanger, Norway; backed by Nordic Capital; offices in Oslo, Stockholm, Copenhagen, LA). Pain points (inferred from public product positioning, not a verbatim quote): operating a large fleet of enterprise conversational/agentic AI in production with governance and oversight; maintaining reliability at 150M+ conversations/year. Challenges: enterprise-grade governance, oversight and reliability across hundreds of deployed agents. Must-haves: production reliability, governance/control, oversight tooling. Nice-to-haves: cost/observability visibility per agent. ICP confidence: High (CTO at an AI-native agent company in the target size band, shipping agents in production).
Signal: 4 (ICP technical leader at an agent-native company shipping agents in production; found via 2026 agent-startup research + LinkedIn verification).
Source: https://www.linkedin.com/in/sauhaarda/ ; company context https://www.prnewswire.com/news-releases/reducto-raises-75m-series-b-to-define-the-future-of-ai-document-intelligence-302581462.html , https://www.ycombinator.com/companies/reducto
Company size: ~63 employees (San Francisco); Series A $24.5M then Series B $75M (Oct 2025); "complete agentic document platform" — agentic parsing/extraction over 250M+ pages in production.
Pain points (inferred from role/company, not quoted): document-intelligence agents run at very high volume, so per-run/per-page cost and token efficiency directly hit gross margin; accuracy/reliability of agent extraction at production scale; context/token waste on large documents.
Challenges: keeping agent accuracy high while controlling inference spend across huge document volumes; visibility into cost per document/run.
Must-haves: granular cost-per-run visibility; token/context optimization to protect margins; reliability monitoring.
Nice-to-haves: automated anomaly detection on run cost; per-customer cost attribution.
ICP confidence: High — Co-Founder & CTO of a Series B, agent-native company (63 emp) squarely in the ICP band, shipping agents in production.
Signal: 4 (ICP at an agent-shipping company) with a HIRING signal — he is personally recruiting Applied AI engineers, which is the "scaling 1→5+ agents" tell.
Source: https://www.linkedin.com/in/ravi-ghanta/recent-activity/all/ ; found via https://www.linkedin.com/search/results/people/?keywords=Aircall%20Head%20of%20AI%20engineering%20agents ; company page https://www.linkedin.com/company/aircall/people/
Company size: 862 associated members on LinkedIn (28 Aug 2026) — comfortably inside the 50-2,000 band. HQ Paris / New York. "The collaborative, AI-powered customer communications platform. Trusted by more than 23,000 companies worldwide." Ships AI voice/CX agents in production.
LinkedIn headline (verbatim): "Engineering Director @ Aircall | AI Strategy/Execution for Internal Productivity | Vision and Execution of Product and Platform Strategy".
Pain points: NOT DIRECTLY EXPRESSED — no cost or reliability quote found in his activity; do not invent one. What IS verifiable: he posted "I'm #hiring a Senior Software Engineer, Applied AI for our Bellevue or San Francisco offices" and reposted a Senior Infrastructure Engineer opening. His stated remit is "AI Strategy/Execution for Internal Productivity" — i.e. he owns internal agent deployment, a workload with no natural cost ceiling.
Challenges: building an Applied AI function from a hiring standpoint while also owning product and platform strategy; internal-productivity agents are the workload most likely to run unmetered because there is no customer invoice forcing attribution.
Must-haves (hypothesis, unverified): attribution of internal agent spend to teams/owners.
Nice-to-haves (hypothesis): guardrails on internal agent usage.
ICP confidence: Medium — Director level at a confirmed in-band company that ships agents, plus a genuine hiring signal; held at Medium because no pain was expressed in his own words and his most recent activity is ~6 months old.
Verification notes: headcount from LinkedIn People tab (862 associated members). Profile URL confirmed by DOM href extraction. NET-NEW company — Aircall returned zero matches on dedup.
Ravi Kokku — Co-founder & CTO (ex-IBM Research), Emergence AI
Signal: 4 (ICP building/shipping agents in production). Source: https://www.emergence.ai/about-us (team/leadership); Tracxn company profile. Company size: 116 employees (in range), ~$97.2M raised. CTO/technical decision-maker for Emergence AI's agentic infrastructure and runtime Orchestrator that coordinates multiple autonomous agents. Pain points: reliability and verification of autonomous agents in mission-critical enterprise workflows; runtime coordination of multiple agents. Challenges: engineering governed, verifiable multi-agent systems at scale. Must-haves: agent verification, runtime orchestration/observability, reliability guarantees. Nice-to-haves: cross-domain agent coordination, benchmark performance. ICP confidence: Medium — technical decision-maker at right-sized AI-native agent company; orchestration positioning partially overlaps operating-layer category.
Ravi Mayuram — Chief Technology Officer, Uniphore
Signal: 4 (ICP CTO at company shipping agentic AI in production). Source: https://www.uniphore.com/press-releases/uniphore-appoints-ravi-mayuram-as-new-chief-technology-officer/ ; https://www.uniphore.com/our-team/ ; https://www.clay.com/dossier/uniphore-cto. Company size: ~700–1,000 (Uniphore; enterprise "Business AI Cloud" for the agentic enterprise). NOTE: later-stage (raised $600M+ / beyond Series C) — headcount and agent focus fit ICP; funding stage is past the stated A–C band. Leads engineering + AI groups; 25+ years eng leadership (ex-Couchbase/Oracle/Sun). Pain points (inferred): taking agentic AI from experimentation to "production-grade, measurable outcomes"; reliability, security, and cost at enterprise scale. Challenges: scaling a composable/sovereign agent platform reliably; governance. Must-haves: production reliability, security, cost/observability. Nice-to-haves: compounding enterprise intelligence. ICP confidence: Medium (right role and company size, active agent builder; softened by later funding stage).
Ravi Nemalikanti — Chief Product & Technology Officer, Abrigo
Signal: 1 + 4 (own verbatim quote on agent control/governance in a dated 2026 agent-platform launch). Source: https://www.businesswire.com/news/home/20260708110699/en/Abrigo-Launches-Agentic-AI-Platform ; https://press.aboutamazon.com/aws/2026/7/abrigo-launches-agentic-ai-platform ; https://www.hpcwire.com/aiwire/2026/07/10/abrigo-expands-banking-ai-with-data-driven-agentic-lending-platform/ ; LinkedIn: https://www.linkedin.com/in/ravinema/ . Company size: ~874-960 employees (LeadIQ Apr-2026 ~960; other sources ~874). Banking/lending & risk software for banks and credit unions; Carlyle/Accel-KKR backed (not VC Series A-C - stage deviation, headcount in range).
Pain points: VERBATIM (8 Jul 2026 launch release) - "We built Abrigo APX with explainability, governance, and operational control at its core because financial institutions need AI that can scale responsibly." INFERRED from the same coverage: multi-model routing across Amazon Nova and Anthropic Claude per use case, plus per-customer isolated AWS instances and data stores, means cost and behaviour vary by model, tenant and workflow with no single per-run view.
Challenges: Abrigo APX spans the full life of loan (pipeline, underwriting, closing, servicing, portfolio admin) - a multi-agent fleet going GA in Q3 2026 across regulated lending workflows at hundreds of financial institutions. Every agent action must be explainable to bank examiners.
Must-haves: per-action audit trail and deterministic replay; approval gates; provable governance for regulated lending decisions.
Nice-to-haves: per-run and per-model cost attribution across tenants; drift detection as agents 2-N ship.
ICP confidence: HIGH. Exact title seniority (CPTO), confirmed headcount in band, dated multi-agent launch 13 days before this run, and he stated control/governance as the design premise in his own words. Note: leadership page and some press use "CTO" and "Chief Product and Technology Officer" interchangeably for him.
Timing: launched 8 Jul 2026, GA Q3 2026 - sits at the very start of the 4-10 week post-launch buying window identified in VOC #89.
Ravi Petlur — CTO, Verloop.io
Signal: LinkedIn People search (ICP title + agents), closest to Bucket 4 (ICP building/shipping agents). Source: https://www.linkedin.com/in/ravipetlur/ (found via people search 'Verloop AI engineering agents'). Company size: ~102 employees (verified via Tracxn, Jun 2026); Verloop.io acquired by Nurix AI Jul 2026 — combined still well in-band. Company builds voice & conversational AI agents for global enterprises (agent-native, in-band 50-2000). Pain points: production reliability at scale (headline: 'scaling AI systems to 99.99% uptime'), multilingual/Arabic-native LLM quality, RAG accuracy for enterprise voice agents. Challenges: keeping voice agents reliable and low-latency at global enterprise call volumes; engineering leadership across scaling systems. Must-haves: production uptime/reliability, cost-efficient inference at scale. Nice-to-haves: per-run/per-conversation cost visibility, agent observability & evals. ICP confidence: High (CTO = C-suite technical; agent-native company; size confirmed in-band). Note: pain points inferred from his own public headline, not fabricated quotes.
Ravi Sindri — VP innovation, Qualizeal
Helps with: Connections, sales pipeline where qualizeal is implementing agentic solutions for their clients
Very close friend and has helped with projects in the past, currently might have conflict of interest
Signal: 4 (ICP technical leader at an AI-native GTM platform shipping agents). Source: https://www.apollo.io/llm-info ; https://www.crunchbase.com/person/ray-li-5 ; https://www.linkedin.com/in/raysiruili ; Tracxn May 2026. Company size: ~469 employees (Series D, $1.6B valuation, ~$100M revenue 2026). Pain points (inferred from public product/positioning): positioning GTM execution as "more autonomous, intelligent, accessible through AI agents, recommendations, and workflow orchestration" over a huge B2B data graph — running sales agents at scale across many customers. Challenges: reliability and cost of agents acting over massive contact/intent datasets; orchestrating multi-step agent workflows dependably. Must-haves: reliable multi-step agent orchestration, cost control at data scale. Nice-to-haves: per-agent/per-workflow cost visibility. ICP confidence: High (co-founder+CTO; 50-2,000 emp; AI agents in production).
Raz Itzhakian — Co-founder & CTO, BlinkOps
Signal: 1 (agent reliability/control in production — found via web research on agentic-security platforms; Claude-in-Chrome/LinkedIn unavailable this run, so no personal LinkedIn post captured; pain points inferred from company public positioning, not a personal quote). Source: https://www.securityweek.com/blinkops-raises-50-million-for-agentic-security-automation-platform/ ; https://www.blinkops.com/about . Company size: 100+ employees. Funding: Series B $50M (Jul 2025), ~$90M total; investors Lightspeed, Hetz, O.G. Venture Partners. Product: platform for building custom security "micro-agents" that automate security workflows — actively shipping many agents in production. Pain points (inferred): making agentic automation reliable & trustworthy at enterprise scale; visibility/control across many micro-agents; scaling 1→many agents. Challenges: reliability/trust of autonomous agent actions in security-sensitive workflows. Must-haves: governance/control over agent actions, production reliability. Nice-to-haves: cost/observability per agent run. ICP confidence: High (CTO, 100+ emp, Series B, agent-native product).
Razvan Kusztos — VP of Engineering, PolyAI
Signal: 4 (VP Engineering at an AI-native voice-agent company; found via LinkedIn people search of PolyAI).
Source: https://www.linkedin.com/search/results/people/?keywords=PolyAI%20head%20of%20engineering%20director%20AI
Company size: ~250-350 employees; PolyAI, Series D ($200M+ raised, $750M val), enterprise voice AI agents. In-band on size (50-2,000); stage past Series C noted.
Pain points (inferred): agent reliability in production, latency/cost of voice agents at scale, scaling eng org.
Challenges: production reliability & cost per run across many concurrent agent conversations.
Must-haves: cost-per-run visibility, reliability.
Nice-to-haves: unified observability.
ICP confidence: Medium-High — VP Engineering (ICP) at agent-native company, in-band size.
Signal: 4 (ICP technical co-founder building/shipping agents in production) + 3 (voice-agent platform space, adjacent to competitors like Vapi/PolyAI already in brain). Source: https://voice-ai-newsletter.krisp.ai/p/customer-facing-personalized-voice ; https://www.amplifypartners.com/barrchives/how-regal-builds-real-time-voice-agents-for-contact-centers ; https://www.regal.ai/about ; https://tracxn.com/d/companies/regal-ai . Company size: 126 employees (Mar 2026); Series B, ~$106M raised over 4 rounds (latest Oct 2024); NYC, founded 2020. Voice AI agent platform for customer engagement/contact centers — real-time voice agents in enterprise production (support, lead qualification, appointment setting) with unified customer profile, journey builder, conversation intelligence. Greene is co-founder & CTO (technical decision-maker; speaks/writes on building real-time voice agents). Pain points: real-time latency + reliability of voice agents at scale; controlling agent behavior across many channels/customers; conversation quality. Challenges: production reliability of low-latency multi-channel agents; cost economics of high-volume voice runs; observability into agent behavior. Must-haves: reliability at scale, behavior control, observability. Nice-to-haves: per-call cost visibility, model routing to manage LLM spend. ICP confidence: High (technical co-founder/CTO at a 126-person Series B AI-native company whose core product is production voice agents).
Rei Kasai — Chief Product Officer (AI-driven growth), Glia
Signal: 4 (AI-focused product leader at a qualifying agent-shipping company). NOTE: LinkedIn/Chrome unavailable this run; found via web research + primary-source verification. Source: Glia press release "Glia Names Rei Kasai as Chief Product Officer to Lead Next Phase of AI-Driven Growth" (glia.com/news, Jul 2025). Prior role: SVP & Head of Product at Talkdesk (product management + AI strategy). Company: Glia — financial-services AI contact-center platform shipping Glia Cortex agentic AI; ~445–469 employees (mid-2026); founded 2012. Pain points (inferred from role/domain, not a personal quote): shipping reliable, compliant AI agents for banks at scale; proving ROI/cost-per-interaction of the "AI workforce"; product visibility into where agents succeed/fail in production. Challenges: balancing autonomy vs. guardrails for regulated financial workflows. Must-haves: outcome/cost visibility, reliability, auditability. Nice-to-haves: agent orchestration and analytics across a growing agent fleet. ICP confidence: Medium-High (C-suite product leader explicitly leading AI product strategy = ICP "VP of Product, AI-focused"; company + headcount confirmed in band). Pain framing inferred from documented product/domain, not a fabricated quote.
Reshadat Ali — Director of Engineering, AI Platform & GenAI, ArmorCode
Signal: 4 (ICP building/shipping agents; amplifying own company's agent launch). Source: https://www.linkedin.com/in/reshadat/ — reposted ArmorCode's launch of "Anya Agents" (agentic AI for AppSec/vuln remediation); found via LinkedIn people search "Director of Engineering AI agents LLM cost production" (2026-08-29). Company size: ~222 employees (Tracxn, Jun 2026), Series B, $81M raised. Cybersecurity SaaS, AI-native product line. Pain points: shipping agentic AI ("Anya Agents") into a regulated cybersecurity SaaS at enterprise scale; security teams cannot keep pace with vulnerability volume, so agents must run continuously and reliably. Challenges: moving from passive AI assistants to autonomous "AI workers" in production; owns the AI platform layer under multiple agent products. Must-haves: reliable agent execution in production, platform-level control over how many agents run and what they consume. Nice-to-haves: per-agent cost attribution across the AppSec product line. ICP confidence: High — Director-level, owns AI platform, 222-employee Series B SaaS with 1+ named agent product GA and more in build.
Signal: 1 (ICP writing publicly about production agents and the context layer) with Signal 4 crossover. Best pain/positioning signal captured this run.
Source: https://www.linkedin.com/in/reutperel/recent-activity/all/ (post ~1 month old, 251 reactions, 30 comments) ; found via https://www.linkedin.com/search/results/people/?keywords=Hibob%20Head%20of%20AI%20engineering
Company size: BORDERLINE — LinkedIn company page shows the "1K-5K employees" band, and the People tab reports 2,118 associated members (28 Aug 2026). Associated-member counts overstate current headcount; widely reported HiBob headcount is ~1,000-1,300. VERIFY HEADCOUNT BEFORE OUTREACH — if the true figure is above 2,000 this record falls outside the ICP band. HQ New York / Tel Aviv. HR tech SaaS.
LinkedIn headline (verbatim): "VP Data @ HiBob | Context engineering for AI — composable data fabric, knowledge graph, certified data foundations, production agents | Ex-CDO Amdocs".
Pain points / expressed signal (verbatim from her own post about speaking at Snowflake AI Day TLV): "No hype slides — we showed what's running in production at HiBob Enterprise Data Platform." / "Snowflake isn't a data warehouse anymore. 2006 was Cloud, 2016 was Data Cloud, 2026 is Agentic AI. They're betting that whoever owns the governed context layer — not just storage or compute — becomes the platform where agents reason, decide, and act." / claims "production-grade data containers in hours not weeks, self-serve that actually serves, zero egress architecture."
Challenges: she has staked her function on being the governed context layer that agents reason over — which means context quality, cost of context delivery, and agent trustworthiness all land on her desk. "Zero egress architecture" indicates she is already optimising data movement cost.
Must-haves: governed, certified context for agents; evidence of what is actually running in production (she explicitly rejects hype/demo-ware).
Nice-to-haves: self-serve access for downstream teams; faster time-to-production for data containers.
ICP confidence: High on persona and pain framing (VP-level, owns production agents, publicly reasoning about the context layer and egress cost); Medium OVERALL solely because of the unresolved headcount question above. Her language — "governed context layer", "context engineering", "production agents" — maps almost one-to-one onto thealpha.ai's positioning, making her the highest-priority name in this run.
Verification notes: profile URL confirmed by DOM href extraction (/in/reutperel/, note the surname on the URL differs from her display name). NET-NEW company — HiBob returned zero matches on dedup.
Richa Sheth — Head of Forward Deployed Engineering (Head of FDE), Jeeva AI
Signal: 4 (senior engineering leader at an agent-native company; surfaced via Jeeva AI people directory during ICP company review). Source: https://www.linkedin.com/company/jeevaai/people/ (profile: https://www.linkedin.com/in/richa-sheth-6a2477199/). Company size: 51-200 (verified on LinkedIn). Company: Jeeva AI — autonomous digital worker / AI agent platform (SF, 25K followers), actively building & shipping AI agents. Pain points (inferred from FDE-lead role deploying agents into customer production): agent reliability in production, token/cost efficiency per deployed agent, scaling agent deployments across enterprise customers. Challenges: making customer-deployed agents reliable and cost-predictable; forward-deployed delivery at scale. Must-haves: production agent observability, reliability, cost-per-run visibility. Nice-to-haves: reusable agent deployment harnesses. ICP confidence: Medium (Head-level engineering leader at a qualifying 51-200 agent-native company; pain inferred from role/company rather than a direct post).
Rick Nucci — CEO & Co-founder (technical co-founder; previously founder/CTO of Boomi), Guru (getguru.com)
Signal: 1 (ICP author). LinkedIn post, 36 reactions.
Source: https://www.linkedin.com/search/results/content/?keywords=%22our%20eval%20suite%22%20OR%20%22our%20evals%22%20agents%20production&datePosted=%22past-month%22&sortBy=%22relevance%22
Company size: CONFLICTING — PitchBook 60, RocketReach 326. Both land inside the 50–2,000 band, so the record qualifies either way, but the true figure is NOT cleanly verified. (Tracxn's 2,399 appears to conflate this Guru with Guru.com, the freelance marketplace — disregard it.) Knowledge/AI SaaS, Philadelphia, Series C.
Pain points (his own words): "We just gave Guru Knowledge Agents a way to prove they'll perform exactly the same way tomorrow as they did today." / "trust in an agent was never a launch-day property to begin with. It has to hold up on week 40 the same way it did on week 1." / "your company knowledge never stops evolving, which makes factual correctness a moving target. An eval built around it fails constantly for the wrong reason: the knowledge changed, not the agent."
Challenges: Detecting agent regression over time; evals false-failing when the underlying knowledge (not the agent) changed; sustaining trust as models and workflows shift underneath; agents called both by humans in-app AND by other agentic systems via API, so they need a single log/trace of every Q&A pair.
Must-haves: Continuous (not launch-day) eval runs; per-interaction logging with source attribution; drift detection that separates knowledge change from agent regression.
Nice-to-haves: One-click promotion of good answers into the eval set; agent-to-agent call observability.
ICP confidence: Medium — technical co-founder of an AI-native Series C SaaS shipping production agents with in-house evals, and the portable-trace-artifact story maps directly onto his "single log of every Q&A pair" need. Deductions: he is CEO (may delegate), Guru competes on the agent-quality surface, and headcount is not cleanly confirmed. VERIFY HEADCOUNT before outreach.
Signal: 4 (ICP speaking publicly about building production agents).
Source: https://qconsf.com/speakers/nov2026 and https://qconsf.com/schedule/nov2026 — QCon SF 2026. Talk: "How to Build a Real-Time Voice Agent" ("A voice agent looks like a chatbot with a microphone" — i.e. the talk is about everything that assumption gets wrong).
Company size: ~287–410 (Tracxn 410, Jun 2026; Revelio ~287). Series B AI-native inference/training cloud.
Pain points: Real-time voice agents are the most latency- and cost-sensitive agent class — every turn is a multi-model pipeline (ASR → LLM → TTS) where token waste and retries show up directly as latency and per-minute cost.
Challenges: Sub-second budgets per turn across a multi-hop agent pipeline; cost per conversation rather than cost per token.
Must-haves: Per-turn / per-run latency and cost attribution across a multi-model agent pipeline.
Nice-to-haves: Model routing by turn complexity.
ICP confidence: Medium — title (Director of ML) and company size both qualify, and voice agents are a genuine 5+-agents-in-production use case. Caveat: Together AI sits partly on the infrastructure-vendor side of the market rather than being a pure agent-shipping SaaS, so treat as a partner-or-prospect conversation. Note the library already contains Dan Fu (VP of Kernels, Together AI) — same account, second contact.
Rob Woollen — Co-Founder & CTO, Sigma Computing
Signal: 4 (ICP technical leader at a company shipping agents in production). Source: https://www.sigmacomputing.com/blog/introducing-sigma-agents ; https://www.linkedin.com/in/rwoollen/ ; Tracxn 2026. Company size: ~1,423 employees (Series E, $80M raised May 2026). Pain points (inferred from public product/positioning): running "hybrid agents" that analyze governed warehouse data and must hold for human approval before actions move downstream — reliability, control, and governance of agent actions on production business data. Challenges: giving agents enough context on governed data while keeping outputs trustworthy/auditable; scaling agent workloads on warehouse compute cost-effectively. Must-haves: human-in-the-loop approval, governance, and control over what agents can execute. Nice-to-haves: per-agent cost/usage visibility on warehouse compute. ICP confidence: High (clean CTO+co-founder role; 50-2,000 emp confirmed; agents shipping in production).
Robert Dugdale — SVP of Enterprise AI, Kasisto
Signal: 4 (senior AI leader — Head-of-AI equivalent — at a qualifying agentic-AI company). NOTE: LinkedIn/Chrome unavailable this run; found via web research + primary-source verification. Source: Kasisto leadership references + product pages (kasisto.com). Company: Kasisto (NYC) — KAI/KAIgentic agentic AI for banking & finance; ~50–72 employees (2026); acquired by Backbase Jun 2026, operates as distinct agentic-banking AI unit. Role maps to ICP "Head of AI / VP of AI." Pain points (inferred from role/domain, not a personal quote): making enterprise banking agents reliable and accurate in production; controlling inference/token cost for high-volume financial conversations; scaling agentic deployments across many bank customers. Challenges: regulatory compliance + hallucination control; per-deployment cost/observability. Must-haves: reliability, guardrails, cost-per-run visibility. Nice-to-haves: agent orchestration + eval tooling. ICP confidence: Medium-High (SVP Enterprise AI = ICP AI-leadership persona; company + headcount in band, near floor; acquisition a minor caveat). Pain framing inferred from documented product/domain, not a fabricated quote.
Roberto Pieraccini — VP, Chief Scientist & Head of AI Frontier, Uniphore
Signal: 1 (senior AI leader at an agent-shipping company; found via web research — LinkedIn/Chrome unavailable this run). Source: https://www.uniphore.com/our-team/ ; https://theorg.com/org/uniphore/teams/artificial-intelligence-division. Company size: ~700–1,100 employees (within 50–2,000 band); Uniphore ships agentic/conversational AI agents in production for enterprise CX. Roberto Pieraccini is VP, Chief Scientist & Head of AI Frontier (renowned conversational-AI/speech scientist). Pain points (INFERRED from role/company context — NOT verbatim): reliability and cost of production conversational agents, scaling agent quality, per-run cost visibility. Challenges: pushing agent capability while controlling inference cost/reliability at scale. Must-haves: production cost + reliability observability. Nice-to-haves: routing/optimization. ICP confidence: Medium-Low (VP-level title fits, but role is research/frontier-oriented rather than production-eng ownership; Uniphore late-stage). NOTE: pain points inferred, not quoted. Included to exceed the 5-person minimum; lowest-priority of this run's adds.
Rodrigo Barnes — Chief Technology Officer, Owkin
Signal: 4 (ICP writing/shipping agents). Source: https://www.owkin.com/people ; https://www.eu-startups.com/2026/05/london-based-geordie-ai-secures-e25-million-to-help-enterprises-govern-ai-agents/. Company size: ~325 (PitchBook/GlobalData 2026). LIVE LINKEDIN VERIFIED this run — people search returns "Rodrigo Barnes — CTO @ Owkin — Automating and Accelerating Drug Discovery with AI", Greater Edinburgh Area. Agent evidence: Owkin's CISO publicly described the company as running hundreds of agents across 50+ petabytes of data (Geordie AI funding writeup, May 2026); Owkin shipped "Pathology Explorer" from its K Pro agent suite, integrated into Anthropic's Claude for Healthcare and Life Sciences (Jan 2026). Pain points: agent scale framed as simultaneously a competitive necessity and a risk-visibility problem (paraphrase of company/CISO statement, NOT a direct Barnes quote). Challenges: scaling from pilot agents to hundreds in production inside a heavily regulated biotech/pharma data environment; balancing shipping velocity against governance. Must-haves: per-agent cost/run visibility, production reliability monitoring, audit trail across hundreds of concurrent agents. Nice-to-haves: integration with existing agent frameworks (K Pro, Claude); lightweight governance that does not slow velocity. ICP confidence: High — current CTO confirmed on live LinkedIn + company site, headcount in band, third-party (non-vendor-marketing) evidence of hundreds of agents in production.
Signal: 4 (ICP technical co-founder building/shipping agents in production). Source: https://www.calcalistech.com/ctechnews/article/mzl1gy8tx ; https://roeylalazar.com/ ; https://www.prnewswire.com/news-releases/wonderful-raises-150m-series-b-to-accelerate-enterprise-ai-adoption-in-30-markets-302712238.html. Company size: ~489 employees (May 2026), scaling toward ~900 by year-end; $284M raised total, $2B valuation, Series B (Insight Partners, Index Ventures, IVP, Bessemer, Vine). Wonderful builds an enterprise AI-agent platform automating customer interactions across 30+ markets with locally embedded agent teams — clearly 5+ agents in production. Pain points (inferred from company focus): scaling large fleets of enterprise customer-facing agents reliably across many markets/languages; maintaining agent quality and consistency at scale; cost of running many agents. Challenges: reliability + localization of agents across 30+ markets while scaling headcount/infra fast. Must-haves: production reliability, agent observability/control at scale. Nice-to-haves: per-agent/per-run cost visibility, agents that compound/improve over time. ICP confidence: High (technical co-founder/CTO at a 489-person Series B AI-agent-native company shipping agents in production).
Rohan Chopra — Co-Founder & CEO (technical; early DoorDash engineer), Convey
Signal: 4 (ICP technical co-founder/CEO at agent-native company shipping agents in production). Source: a16z announcement (https://a16z.com/announcement/investing-in-convey/), TechCrunch/The Next Web, PitchBook headcount. Company size: 106 employees (PitchBook, confirmed). Convey builds enterprise "digital teammates" — AI agents trained by demonstration to autonomously own end-to-end workflows; $38M Series A led by a16z (Khosla, Pear VC); >1.1M hours of production work automated; $50M ARR within 7 months; customers NBCUniversal, Samsara, TelevisaUnivision, Unity, Faire, ChargePoint. Chopra was one of first ~20 employees at DoorDash, spent 8 years building the logistics engine (technical founder). Pain points: making agents reliably own an outcome (not just a task) end-to-end; last-mile reliability at production quality; scaling autonomous teammates across many enterprise workflows. Challenges: reliability/quality at volume. Must-haves: production reliability, monitoring. Nice-to-haves: cost-per-run visibility. ICP confidence: High (technical co-founder/CEO at 106-person agent-native company shipping 5+ agents in production).
Rohan Suri — Co-Founder & Chief Product Officer (AI-focused), Nooks
Signal: 4 (ICP building/shipping AI agents; found via web research — LinkedIn/Chrome NOT connected this run; pivoted to web research + primary-source verification). Source: https://www.linkedin.com/in/surirohan/ + https://theorg.com/org/nooks-ai (Co-Founder & CPO of Nooks; co-founder Nikhil Cheerla (CTO) already in People Library, but Rohan Suri was NOT — deduped against full ~781-person library). Matches the ICP "VP of Product (AI-focused)" persona: CPO with a technical/ML background (ex-ML Engineer intern at Cerebras Systems; ex-SWE intern at Forethought; Stanford). Company size: ~412 employees (Jun 2026; some sources ~541); Series B, ~$75M raised; Nooks builds AI sales agents (AI SDR, parallel dialer, coaching) shipped in production for outbound sales teams. Pain points (inferred from company focus, NOT a verbatim quote): making AI sales agents reliable + effective at large outbound volume; cost per agent action at scale; agent accuracy/quality. Challenges: trust + cost of autonomous sales agents at scale. Must-haves: reliability, cost efficiency per run. Nice-to-haves: observability, multi-agent coordination. ICP confidence: Medium (AI-focused product co-founder/CPO at in-band ~412-emp company actively shipping AI sales agents).
Signal: 1 & 4. Source: https://www.linkedin.com/in/rconline ; Insight Partners profile "teaching data to manage itself" (https://www.insightpartners.com/ideas/acceldata-leadership-story/). Company size: ~294 employees; Series C, $100M+ raised (Insight Partners, March Capital). Building an "agentic data management" platform where data systems monitor, optimize and self-heal. Pain points: reliability of autonomous data agents; observability into what agents actually do; controlling agent behavior at enterprise scale. Challenges: making self-healing data agents trustworthy in production. Must-haves: agent observability, reliability guardrails. Nice-to-haves: per-agent cost visibility. ICP confidence: High (technical co-founder/CEO, Series C, ~294 emp, shipping agentic platform).
Signal: 4 (ICP shipping agents in production). Source: https://www.linkedin.com/in/romainlapeyre/ and https://temporal.io/resources/case-studies/gorgias-uses-ai-agents-to-improve-customer-service. Company size: ~521 employees (~$100M raised), ecommerce customer-support SaaS serving 15,000+ brands; ships multiple AI agents in production (autonomous support agent handling order tracking/returns/refunds/subscriptions, plus a sales/upsell agent). Co-founder & CEO (CTO Alex Plugaru is separately in the People Library). Pain points (inferred from confirmed agent deployment, not a verbatim quote): per-resolution/token cost as ticket volume grows, agent reliability across 15k+ merchants, visibility into agent cost and behavior. Challenges: keeping per-resolution economics viable at scale. Must-haves: cost-per-resolution visibility, reliability guardrails. Nice-to-haves: cross-agent observability. ICP confidence: Medium — confirmed qualifying company shipping 5+ agents; CEO is business co-founder (equivalent decision-maker) rather than the technical lead.
Signal: 4 (ICP technical co-founder shipping agentic AI in production). Source: https://www.pigment.com/ai ; https://tracxn.com/d/companies/pigment/ ; https://www.crunchbase.com/person/romain-niccoli ; https://fr.linkedin.com/in/romainniccoli. Company size: ~714 employees (May 2026); Series D ($397M total, Paris). Actively shipping agents: Pigment ships 'Super-Specialist Native AI Agents' for enterprise business planning (finance/sales/HR) to enterprise customers (Snowflake, Unilever, Siemens). Romain Niccoli is technical co-founder & co-CEO (co-founder & former CTO of Criteo). Pain points (inferred from product + domain, not verbatim quotes): reliability/accuracy of planning agents that act on financial data; cost and governance of agents across many enterprise tenants; auditability of agent-produced numbers. Challenges: making enterprise planning agents trustworthy, controllable and cost-efficient at scale. Must-haves: reliability, auditability/control, per-run cost visibility. Nice-to-haves: observability + compounding institutional memory. ICP confidence: Medium-High (technical co-founder/co-CEO, 714 emp in band, agentic product in production; Series D just past A-C band).
Roman Ignatov — R&D Director, Similarweb (concurrently VP Technology at XPLN); prior VP Engineering / CTO, Similarweb
Signal: Bucket 2 (ICP engaging with non-ICP practitioner content about agent observability/cost). Also qualifies partly under Bucket 4 (engaging with agent-building content).
Source: Comment (19h old as of 2026-08-28) on Ganapathi Ramkumar Palanivelu's LinkedIn post "Monitoring an AI agent is not simply about checking whether the application is running" (315 reactions / 24 comments), found via LinkedIn post search "agent observability cost per run", past month. Profile: https://www.linkedin.com/in/romanignatov/
Company size: Similarweb ~1,000-1,200 employees (Dec 2025 filings ~1,120; LinkedIn lists ~1,200). Within 50-2,000 band. Note: Similarweb is publicly listed (NYSE: SMWB), not Series A-C — this is the one ICP criterion it misses. XPLN (his other role, VP Technology, Germany) is a smaller pricing-intelligence company.
Role confirmed on profile: R&D Director at Similarweb, Jan 2026 - present (Full-time, Remote, New York). Based in Bavaria, Germany. Headline: "VP Tech at XPLN | R&D Director at Similarweb | VP Engineering / CTO | Remote-first engineering orgs | Data & analytics platforms".
Pain points expressed (verbatim from his comment):
- Step-level alerting on agents is unusable: "a single tool retry is normal behavior, so step-level alerts drown the on-call."
- No good default unit of observation for agents — argues "the useful unit of alerting is usually the whole run, not the individual step."
- Trace storage cost / volume growth: asked openly "How do you decide what fraction of traces to keep at full fidelity, given how fast full prompt/response capture grows?" — direct context/token-waste and cost-of-observability signal.
Challenges:
- Detecting real production degradation early; he uses task-completion rate and p95 steps-per-run as leading indicators, implying he has had to build this himself.
- Balancing observability fidelity against the cost of capturing full prompt/response payloads at scale.
- On-call noise from agent runs in a remote-first engineering org.
Must-haves:
- Run-level (not step-level) agent telemetry with task-completion rate and steps-per-run distributions.
- A principled sampling/retention policy for full-fidelity traces that does not blow up storage or spend.
Nice-to-haves:
- Correlation of run-level metrics to cost per run.
- Alerting tuned to agent semantics rather than infra semantics.
ICP confidence: Medium. Reasons for Medium (not High): title (R&D Director / VP-level) and company size (~1.1k) fit cleanly, and he is demonstrably operating agents in production with cost/observability pain in his own words; but Similarweb is a public company rather than Series A-C, and I could not confirm 5+ agents in production at Similarweb specifically. Not a Low because the pain signal is first-person, dated, and specific.
Do not contact — research only.
Signal: 4 (ICP AI leader at company transitioning to agentic AI; found via LinkedIn People search for ICP titles + healthcare agents). Source: https://www.linkedin.com/search/results/people/?keywords=VP%20AI%20agents%20healthcare%20production ; verification: https://www.medtechdive.com/news/aidoc-raises-150m-to-advance-clinical-ai-foundation-model/819042/ and https://www.aidoc.com/about/news/aidoc-secures-fda-clearance-for-healthcares-first-comprehensive-foundation-model-ai/. Company size: est. ~500-800 (Aidoc, clinical-AI; Series E $150M Apr 2026 led by Goldman Sachs; deployed across ~2,000 hospitals; 60M+ patient cases/yr; 30+ FDA clearances). Building agents: promoting "Agentic Radiology" — transforming AI from passive detection to an active clinical agent; CARE foundation model as core. Ronny leads AI execution + foundation-model systems in regulated clinical environments. Pain points (inferred): making clinical AI agentic and reliable under regulation; scaling agent decisions across ~2,000 hospitals; cost/observability of foundation-model inference at 60M cases/yr. Challenges: reliability + governance of clinical agents; inference cost at scale. Must-haves: reliability, regulatory-grade control, cost visibility. Nice-to-haves: observability. ICP confidence: Medium-Low (VP AI at ~600-emp clinical-AI co in band, but core product is a diagnostic foundation model transitioning to agentic rather than many discrete production agents; Series E later stage; headcount is an estimate). LinkedIn profile URL not captured this run — do not fabricate.
Signal: 4 (ICP building/shipping agents) — found via web research; LinkedIn/Chrome extension NOT connected this run; verified via Forbes, citybiz, company site.
Source: https://www.forbes.com/sites/gilpress/2026/01/15/tulip-raises-120m-towards-human-centric-ai-factory/ ; https://tulip.co/people/rony-kubat/
Company size: ~500 employees (Series D $120M closed Jan 2026, founded 2014, Somerville MA). Tulip builds AI copilots, "composer" products, and full agentic frameworks safe for the factory floor in regulated industries (pharma, medical devices, aerospace).
Role note: MIT PhD (CS); ex-MIT Media Lab; CTO of Tulip since 2014. (Some profiles list CIO/CISO; consistently the company's top technical leader.)
Pain points (INFERRED from domain/role, not verbatim quotes): making agentic AI reliable & safe enough to act on the factory floor in regulated environments; predictable cost of running vision/LLM agents continuously across manufacturing lines; scaling from single copilots to multi-agent frameworks.
Challenges: reliability/guardrails for autonomous action in high-stakes physical operations; cost of continuous VLM/LLM inference at plant scale.
Must-haves: production reliability, deterministic guardrails, cost control per run.
Nice-to-haves: per-agent/per-run cost & behavior observability.
ICP confidence: High — CTO/technical co-founder at a 50–2,000-emp company actively shipping agentic AI.
Signal: 3 (ICP engaging with / quoted in competitor-adjacent observability vendor content). Source: https://www.groundcover.com/customer-stories/aiola (named customer story). Title and employer confirmed live on LinkedIn 2026-08-31 via people search — "Roy Sela · VP Platform Engineering · Israel" (exact profile URL not exposed in search results; left blank rather than guessed). Company size: ~70 per the groundcover case study; 51–200 per Startup Nation Finder / Crunchbase-derived. Stage: Series A2, $25M (Jul 2025), $58M total. Pain points: Sela states that before adopting new observability tooling aiOla had NO tracing or APM in place, leaving limited data correlation and insufficient metrics to understand agent/application behaviour in production; they run voice-agentic workflows embedded via SDK inside CUSTOMER applications, so they had no visibility into how, when and at what scale their agents are actually invoked. Also cites compliance and cost concerns as data volumes and egress fees grew with scale. Challenges: meeting contractual real-time voice SLAs without production visibility; R&D team resistant to manually instrumenting many services; tracing agent workflow failures end to end across customer environments. Must-haves: low/zero-instrumentation tracing, agent-workflow-level visibility, data residency control. Nice-to-haves: unified infra + application observability, predictable pricing not tied to ingestion volume. ICP confidence: High (clean VP-level platform engineering title verified live, genuine production voice-agent workloads embedded in customer apps, Series A, headcount in band at the low end). CAVEAT: the cost pain here is observability/data-egress cost rather than LLM token cost specifically — position on visibility first, cost second. NOTE: net-new company for the People Library — no prior aiOla record.
Rukmini Reddy — Senior Vice President of Engineering, PagerDuty
Signal: 3 (company is an Arize and LangSmith user for agent observability) + 4 (company ships multiple named agents in production).
Source: name, title and scope on https://www.pagerduty.com/leadership/ — "responsible for managing product and platform delivery, infrastructure and data science." Ex-Slack. Agent evidence: https://www.pagerduty.com/blog/ai/meet-your-virtual-responder-pagerdutys-sre-agent-for-ai-driven-reliability/ (SRE Agent, Mar 2026) and https://www.pagerduty.com/eng/pagerduty-arize-building-end-to-end-observability-for-ai-agents-in-production/
Company size: 1,155 employees as of 31 Jan 2026 (S&P Global / FY2026 10-K via stockanalysis.com/stocks/pd/employees), down 7% YoY from 1,242.
Pain points: No personal public statement found. Company-level pain is well documented: PagerDuty runs golden-dataset regression testing, LLM-as-a-judge, and relevance/groundedness/tool-selection evals, and has Arize monitors paging on-call engineers via a custom "Agent Output Incident" type — i.e. they have had to build agent failure into their own incident-response system. Her peer João Freitas (Chief AI Officer) states publicly that agent-to-agent error propagation is the core reliability problem.
Challenges: She owns product and platform delivery, infrastructure and data science — so the compute, cost and reliability of three production agents (Insights, Shift, SRE) land on her org, while total headcount is shrinking 7% YoY. Scaling agent count without scaling engineering headcount is the structural squeeze.
Must-haves: Platform-level reliability and eval infrastructure for agents; agent failure surfaced as an operable incident rather than a silent regression; doing more with a contracting org.
Nice-to-haves: Infrastructure cost and token-spend visibility per agent — plausible given she owns infrastructure, but not publicly stated.
ICP confidence: Medium-High — SVP of Engineering is unambiguously Director+, company size verified from a filing-derived source, agents demonstrably in production. Downgraded because there is no personal pain signal and no LinkedIn URL rendered on her leadership card. Best used as the platform-owner thread alongside João Freitas (the AI owner) and Tim Armandpour (CTO, exec sponsor).
Rushik Upadhyay — Head of Engineering (RiskOS_Agents), Socure
Signal: 4 (ICP engineering leader whose entire remit is shipping agents). Source: LinkedIn people search "Head of Engineering AI agents token cost observability" -> profile https://www.linkedin.com/in/rushikupadhyay/ (title reads literally "Head of Engineering (RiskOS_Agents) at Socure"; profile is actively hiring "Forward Deployed Engineer - RiskOS Agents", Socure, US Remote). Corroborating: https://www.businesswire.com/news/home/20251028150397/en/ (RiskOS AI Suite launch — six AI agents and assistants), https://news.crunchbase.com/venture/socure-raises-acquires-agentic-ai-startup-fravity/ (Socure raised $156M at $5.2B and acquired AI-native agent startup Fravity, folded in as RiskOS_Agents for watchlist screening/monitoring and KYB), https://www.socure.com/use-cases/riskos-ai-suite. Company size: 586 employees worldwide as of March 2026 (Revelio Labs); grew 13.3% from 518 in 2023. Identity/fraud risk SaaS, late-stage private. Pain points (inferred from role + public announcements, NOT verbatim quotes): running six-plus agents/assistants in production across identity, compliance and authentication decisioning; integrating an acquired AI-native agent platform into an existing risk platform; standing up a dedicated agents engineering org and hiring FDEs to keep deployments working. Challenges: agent reliability and auditability where agents make regulated compliance decisions; scaling from an acquired agent product to a fleet running at enterprise volume — textbook 1->5+ agents wall. Must-haves: production reliability, auditable/traceable agent decisions, per-agent cost and latency visibility. Nice-to-haves: unified observability across the agent fleet, eval/regression tooling. ICP confidence: High (Head of Engineering specifically for agents, 586-emp company with agents shipped and expanding).
Signal: Bucket 4 (ICP building/shipping agents at a qualifying company). NOTE: LinkedIn/Chrome extension not connected this run, so found via web research + primary-source verification instead of the prescribed LinkedIn searches.
Source: https://theorg.com/org/resolve-ai/org-chart/rushin-shah ; https://resolve.ai/blog/Rushin-Shah-joins-Resolve-AI ; corroborated via ZoomInfo + LinkedIn announcement post. VP of Engineering since June 2025; prior Sr. Director of Eng for Gemini @ Google DeepMind, Meta, Apple Siri.
Company size: ~50–160 employees (founded 2024, Series A / extension, ~$1.5B valuation). Non-founder (founder/CEO is Spiros Xanthos).
Ships agents: Multi-agent AI SRE that autonomously triages alerts and investigates/resolves production incidents (customers incl. Coinbase, DoorDash).
Pain points: token/compute cost of long autonomous incident investigations; reliability/trust of autonomous agents acting in prod; observability of the agents themselves; reducing MTTR.
Challenges: proving agent ROI vs. build-vs-buy; scaling multi-agent investigations across clouds/regions reliably.
Must-haves: cost-per-run visibility and reliability guardrails for agents operating unattended in production.
Nice-to-haves: unified observability/control plane across an agent fleet.
ICP confidence: High — exact non-founder VP title verified across multiple sources; AI-SRE agent vendor is a near-perfect fit for an agent operating layer.
Russell Allgor — Chief Supply Chain Scientist (Head of AI/Science), Auger
Signal: Bucket 1 (senior AI/science leader at agent company) — web research pivot (LinkedIn/Chrome unavailable). Source: https://fortune.com/2024/12/09/dave-clark-amazon-executive-team-hires-russ-allgor-auger-supply-chain/ ; https://www.geekwire.com/2026/supply-chain-startup-auger-led-by-ex-amazon-operations-chief-raises-50m-and-lands-big-customers/. Company size: 148 employees (as of June 2026); raised $200M Series A + $50-200M Series B (Insight Partners, Oak HC/FT, Eclipse). Auger uses AI agents to autonomously make day-to-day supply chain execution decisions, sitting atop ERP/WMS/TMS/demand-planning as a single operating layer. Allgor spent 24 years at Amazon as VP & Chief Scientist (fulfillment/network optimization); now leads the AI/optimization science behind Auger's agents. Pain points: reliability and trust of autonomous execution agents; cost/economics of agents making high-volume operational decisions; visibility into agent decision quality. Challenges: scaling accurate, auditable autonomous decisions across enterprise supply chains. Must-haves: per-decision cost/reliability visibility, guardrails. Nice-to-haves: agent decision observability. ICP confidence: Medium — a senior technical AI/science leader (Head-of-AI-equivalent, C-suite science) at a 148-emp agent company; title is "Chief Scientist" rather than VP Eng/Head of AI, hence Medium not High. (Note: Auger's Sanjay Dash, Chief Engineer, already in library; Allgor is net-new.)
Russell Kaplan — President (technical; ex-Head of Generative AI/Nucleus at Scale AI), Cognition (Devin)
Signal: 4 (technical decision-maker at AI-native company whose entire product is coding agents in production). Source: https://www.linkedin.com/in/russelljkaplan , https://techcrunch.com/2026/05/29/cognitions-scott-wu-says-ai-coding-agents-shouldnt-replace-humans/ , https://research.contrary.com/company/cognition. Company size: ~200-400 (Cognition, maker of Devin autonomous coding agent, ~$26B valuation; ships many parallel coding agents in production). Note: CEO Scott Wu already in brain — Kaplan is a separate technical leader (President, ex-Scale AI generative AI lead). Pain points: running large fleets of autonomous coding agents reliably and cost-effectively at scale; token/compute cost of long-running agent tasks. Challenges: reliability of autonomous multi-step agent execution, controlling agent behavior, orchestrating many concurrent agents. Must-haves: reliability at scale, cost efficiency per agent run (coding agents make 3-10x more LLM calls). Nice-to-haves: per-agent cost/observability. ICP confidence: Medium-High (President is a technical decision-maker at in-band, agent-native company; title outside the strict CTO/VP list but clearly a buying-influence eng leader).
Ryan Eldridge — Co-Founder & CTO, Liberate (liberateinc.com)
Signal: 4 (ICP technical co-founder shipping agents in production). Source: https://techcrunch.com/2025/10/15/liberate-bags-50m-at-300m-valuation-to-bring-ai-deeper-into-insurance-back-offices/ ; https://www.liberateinc.com/news/liberate-raises-50m-to-build-the-first-ever-reasoning-ai-agents-for-insurance. Company size: ~50 employees (at ICP floor; expanding eng + GTM); Series A+ — $50M round (Battery Ventures, Oct 2025) at $300M post-money, ~$72M total raised (Redpoint, Eclipse, Commerce, Canapi). Liberate builds "reasoning AI agents" (System of Action) for P&C insurers — end-to-end quoting, claims processing, endorsements/servicing. CTO Ryan Eldridge is ex-Metromile. Pain points (inferred): reliability of autonomous agents completing regulated end-to-end insurance transactions; cost-per-task as claim/quote volume scales; visibility/audit over agent actions in a compliance-heavy vertical. Challenges: trust to let agents complete money-moving tasks unattended. Must-haves: reliability + auditability. Nice-to-haves: per-run cost visibility. ICP confidence: Medium (named technical CTO; agent-native; ~50 emp is exactly at the floor — verify headcount stays ≥50 before prioritizing).
Ryan McCormack — Director of Engineering — Data/ML/AI, Sardine
Signal: 1 (ICP writing publicly about agent reliability/control). Found via web research — Claude-in-Chrome/LinkedIn NOT connected this run (two retries), so prescribed LinkedIn searches were unavailable; pivoted to primary-source web research.
Source: https://www.sardine.ai/blog/AI-agents-for-fraud-operations (published 2026-07-27)
Company size: ~343 employees (Tracxn, 30 Jun 2026; PitchBook ~325). Series C — $70M led by Activant Capital, May 2026; ~$145M total raised. Agentic fraud/AML risk platform for fintech.
LinkedIn URL: not verified (do not guess — confirm before outreach)
Pain points: Silent agent degradation nobody detects until trust is gone — describes a vendor integration that "quietly stopped firing months ago and nobody noticed"; stale/broken ground-truth labels; upstream data pipeline breakage undetected for months; agents producing high-volume noise (an agent given a device-ID field without semantics was "constantly flagging, this is a bad actor, they have 10 devices"); backtest-to-production leakage.
Challenges: "Really, the bottleneck we find, and it's a much more challenging problem, is at the seam of investigations and testing, maybe not necessarily deployment and detection." Compressing the full investigate→test→deploy cycle rather than just deploy; making non-deterministic agent decisions reconstructable for financial-services auditors; governing agents with an analytics function not yet staffed for it.
Must-haves: Evaluation and audit designed in BEFORE build ("that audit capacity is easier to design from the start than to bolt on later"); human-readable step-level decision traces; drift signals (override rate, rejection rate, label agreement); detection of upstream data-source failures; per-case cost and ROI visibility.
Nice-to-haves: Near-real-time agentic labeling; automated capture of analyst overrides back into the loop; unified reporting for non-technical stakeholders.
ICP confidence: High — Director-level owner of Data/ML/AI at a ~343-person Series C AI-native fintech shipping regulated production agents, quoted by name within the last month on exactly agent reliability, evaluation and observability. Note: Sardine's CEO is already a separate record in the library; McCormack is a distinct person and the technical-buyer role.
Ryan St Pierre — VP of Engineering, EliseAI
Signal: 4 (ICP senior technical leader at a 50-2,000-employee company actively shipping AI agents; surfaced via LinkedIn People search after Signal 1-3 content searches returned mostly non-ICP consultants/vendors). Source: https://www.linkedin.com/in/ryanstpierre1/. Company size: ~400+ (EliseAI; conversational AI leasing + healthcare agents in production, ~1 in 6 US apartments). Pain points (inferred, NOT verbatim): reliability of high-volume conversational agents; cost per conversation at scale; scaling agent infra across verticals. Challenges: reliability + cost as conversation volume grows; multi-vertical support. Must-haves: production reliability + per-run cost visibility. Nice-to-haves: fleet-wide observability. ICP confidence: High (VP of Engineering).
Ryan Topping — Head of Data Science & Machine Learning, Clari
Signal: 4 (senior technical AI leader at a company shipping AI agents, found via LinkedIn currentCompany-facet People search). Profile: https://www.linkedin.com/in/ryantopping/ . Source: https://www.linkedin.com/search/results/people/?currentCompany=%5B%222706103%22%5D (Clari). Company size: ~600-800 (estimate; Clari, revenue platform). Pain points (inferred from role + product): Clari ships "Clari Agents" (agentic AI for revenue workflows) - aggregate LLM spend across many agent runs, per-run/per-account cost attribution, reliability of agents acting on revenue data. Challenges: scaling agentic features across a large customer base while keeping model spend predictable. Must-haves: cost-per-run visibility + control, reliability guardrails. Nice-to-haves: model routing / cost attribution. ICP confidence: High (Head of Data Science & ML at Clari, which is actively shipping agents; 2nd-degree connection; verified this run). Pain points INFERRED from role+product, not fabricated.
Ryan Wong — Head of Engineering (quoted as "VP of Engineering" in Databricks case study), Retool
Signal: 2 (non-ICP author posting about agent infra cost, with an ICP engineering leader quoted inside the content). Source: LinkedIn post by Jenni Jones (CEdMA SIG Chair, non-ICP) sharing the Databricks "Customer Zero: How Retool scaled its operational layer with Lakebase and cut costs by 80%" case study — found via https://www.linkedin.com/search/results/content/?keywords=agents%20in%20production%20VP%20Engineering%20cost&datePosted=past-month ; underlying article on databricks.com. Company size: ~447 employees (Revelio Labs, 2026; LeadIQ ~416, PitchBook ~471); Series C / C-II, ~$3.2B valuation, ~$120M ARR; shipping Retool Agents (LLM reasoning + Retool for action) and 1M+ governed databases backing apps/agents/workflows in production. Pain points: VERBATIM — "Scale-to-zero is the headline, but the real story is unit economics." Per the case study, as AI-built apps multiplied Retool needed a database per app, agent and workflow; at 1M+ databases always-on infrastructure "turned into a cost, operations, and governance challenge." Challenges: scaling apps and agents without scaling governance silos alongside them; provisioning was 5-10 minutes before the move (now milliseconds); keeping engineers off infra upkeep. Must-haves: unit-economics visibility per app/agent/workflow, scale-to-zero cost control, governance that does not scale linearly with agent count. Nice-to-haves: instant provisioning, zero engineers dedicated full-time to platform upkeep. ICP confidence: High (senior engineering decision-maker at a ~450-emp Series C SaaS company shipping agents in production, speaking directly to agent unit economics).
Ryuya Nakamura — Executive Officer & CEO, Ai Workforce Business Unit, LayerX
Signal: 4 (ICP publicly writing about building and shipping agents). NOTE: LinkedIn/Chrome not connected this run. LinkedIn URL not verified. NOTE: LayerX CTO Yuki Matsumoto is already in the People Library — this is a second, distinct contact at the same account, running the AI Workforce business unit.
Source: https://note.com/nrryuya/n/nb8af7e35a478 (personal technical essay, Japanese)
Company size: ~430 employees (TechCrunch, Jul 2025). Series B, $100M (Sep 2025), $192.2M total. Japan.
What he said (paraphrased from Japanese): LLM speed, cost and accuracy will keep improving, but agents still need four things — knowledge, skill, tools, and UI.
Pain points: agents need "onboarding" like new hires before they are useful; premature full autonomy produces bad task planning; cost/accuracy tradeoff as agent scope widens.
Challenges: building a platform that gains value as models improve rather than being obsoleted by them; scaling agent deployment via forward-deployed engineers.
Must-haves: knowledge/skill/tool/UI infrastructure around each agent; a stable execution substrate.
Nice-to-haves: AI process mining to auto-generate agent workflows.
ICP confidence: Medium — Series B, 430 emp, actively shipping an agent product line, and writing publicly on agent cost/capability; downgraded from High because the title is a business-unit CEO rather than a clean engineering/AI leadership title.
SN Raju Kattari — Director of Engineering — Agentic Platform & ML Frameworks, Calix
Signal: 4 (ICP writing about building agentic platforms in production). Source: https://www.linkedin.com/in/kattarisuryanarayanaraju/ — post hiring an "Engineering Manager – GenAI / Agentic Systems" to "build production-grade GenAI platforms"; headline "Leading Agentic Platform & ML Frameworks Development ... Multi-Cloud SaaS Architecture". Found via LinkedIn people search "Director of Engineering AI agents LLM cost production" p2 (2026-08-29). Company size: ~1,921 employees (Calix 10-K, Dec 2025; Revelio 1,926 Mar 2026) — inside 50–2,000 band. Broadband platform + multi-cloud SaaS (public, NYSE: CALX) — note: not Series A–C, so company-stage criterion is a partial miss. Pain points: standing up a production-grade agentic platform and ML frameworks org from inside a large multi-cloud SaaS estate; hiring to fill the gap. Challenges: distributed systems + agent orchestration across multi-cloud; production-grade rather than pilot GenAI. Must-haves: a repeatable agent platform layer other product teams can build on. Nice-to-haves: per-agent cost/observability once more agents land. ICP confidence: Medium — title, headcount and agent-platform ownership all fit; company stage (public, not Series A–C) is the mismatch.
Signal: 4 (technical co-founder running a multi-agent product in production). Source: https://www.linkedin.com/in/saad-godil-9728353 , https://hippocraticai.com/team/ , https://research.contrary.com/company/hippocratic-ai. Company size: ~312 employees; healthcare AI (Series-scale). Core product Polaris is a "constellation" of cooperative multi-billion-parameter LLMs operating as agents that talk to/advise patients by phone (dietary, dosage, non-diagnostic). Co-founder & CTO. Pain points: reliability and safety of many cooperating agents in a healthcare setting, cost of running a large multi-model constellation at scale, latency/quality trade-offs. Challenges: orchestrating and controlling a constellation of agents reliably; token/compute cost of 4T+ parameter system per patient call. Must-haves: reliability, safety guardrails, cost control/visibility across many models per interaction. Nice-to-haves: cheaper routing for low-risk turns, per-run cost attribution. ICP confidence: High (technical co-founder/CTO; company in size band; explicitly agent-architecture product in production).
Signal: 4 + 1 (shipping research/enterprise agents; leadership on agent context/cost). Source: https://you.com/resources/introducing-ari-the-first-professional-grade-research-agent-for-business ; https://www.theinformation.com/briefings/ai-startup-com-appoints-new-cto-co-founder-joins-anthropic. Company size: ~350 employees (Tracxn, Apr 2026); Series B $50M (Georgian, Salesforce Ventures, NVIDIA), $1.5B valuation (unicorn). NOTE: promoted to CTO after co-founder Bryan McCann left for Anthropic. Pain points (INFERRED): ARI research agents process hundreds of sources simultaneously — token/context cost at scale; reliability of long-horizon research agents; enterprise agent observability. Challenges: context/token waste across multi-source research agents; extended test-time compute cost; reliability of enterprise agent outputs. Must-haves: token/context cost visibility + control; reliable long-running research agents. Nice-to-haves: routing/model optimization. ICP confidence: Medium-High (newly-promoted CTO; Series B/unicorn; ~350 emp in-band; ARI + enterprise agents in production, though core product is search).
Saar Fredi — Director of Engineering, AI Platform, Gong
Signal: 4 (ICP Director of Engineering, AI Platform, at an agent-shipping company). Source: https://www.linkedin.com/in/saarf/ (found via LinkedIn people search 'Gong director engineering AI platform'). Company size: Gong ~1,100-1,300 employees (independent; revenue-intelligence platform shipping AI agents). Pain points (inferred from role, not a verbatim quote): reliability, latency and cost of the AI platform behind Gong's agent features. Challenges: scaling LLM-backed agent features across a large customer base while controlling spend. Must-haves: observability and cost governance for the AI platform; reliability. Nice-to-haves: per-run cost/latency breakdowns. ICP confidence: High (Director Eng on the AI Platform at an in-range, independent agent-shipping company).
Sai Vivek — Field Chief Technology Officer, Cresta
Signal: 4 (senior technical leader at an ICP agent-shipping company; found via web research — LinkedIn/Chrome unavailable this run). Source: https://cresta.com/blog/cresta-crew-sai-vivek-field-chief-technology-officer ; https://theorg.com/org/cresta. Company size: ~300–500 employees; Series C; Cresta ships contact-center AI agents (agent assist + autonomous voice/chat agents) in production across enterprise customers. Sai Vivek is Field CTO (since Nov 2024) — partners with customers to align technical solutions and accelerate production adoption of Cresta's agent platform. Pain points (INFERRED from role/company context — NOT verbatim): production reliability of deployed agents, cost-per-conversation/run economics, visibility into agent performance and spend across many customer deployments. Challenges: making agent deployments reliable and cost-predictable at enterprise scale. Must-haves: per-deployment cost + reliability visibility. Nice-to-haves: optimization/routing to control token spend. ICP confidence: Medium (verified C-suite-level technical title at a clean Series C, 50–2,000-emp agent company; caveat: "Field" CTO is a customer-facing/solutions role rather than core product engineering). NOTE: pain points inferred, not quoted.
Signal: 4 (ICP building/shipping agents). Source: https://www.artificiallawyer.com/2026/05/06/harvey-launches-legal-agent-bench/ and https://www.harvey.ai/company. Company size: ~340-500 (AI-native legal agents). Pain points: leads engineering for Harvey's in-house legal agent offerings — shipping reliable, production-grade legal agents for enterprise/in-house teams. Challenges: agent reliability and accuracy in a high-stakes regulated domain; scaling agent features to enterprise customers. Must-haves: reliability, eval/observability, controllable agent behavior. Nice-to-haves: per-task cost visibility. ICP confidence: Medium (eng leader building agents at an AI-native company in size band; specific pain signals inferred from role, later-stage company). Note: joined Harvey via acquisition of her startup Hexus (Jan 2026). LinkedIn slug not confirmed — profile_url blank to avoid fabrication.
Salman Bhatti — Chief Technology Officer, Simpro Group
Signal: 4 (ICP CTO hired explicitly to run an AI-first agent platform). Source: https://www.businesswire.com/news/home/20260527167474/en/Simpro-Group-Appoints-New-Chief-Technology-Officer ; https://www.businesswire.com/news/home/20260513010148/en/Simpro-Group-Launches-Lightning-A-Purpose-Built-AI-Native-Operating-Platform-for-the-Field-Service-Trades ; https://www.commercialintegrator.com/news/simpro-group-launches-lightning-ai-platform-for-field-service-trades/148010/ ; headcount 633 (Feb 2026, Tracxn/LeadIQ) and 'more than 600' per Simpro's own About page. LinkedIn: https://www.linkedin.com/in/salmanbhatti/ (profile still shows Vertex Inc. — appointment is very recent, verify before outreach). Company size: ~633. Vertical: field service management for the trades (Brisbane, AU; brands Simpro, AroFlo, BigChange, ClockShark; 20,500 customers; $397M raised). Agent evidence: 'Lightning' launched 13 May 2026 — AI-native operating platform shipped SIMULTANEOUSLY across three products (Simpro / AroFlo / BigChange) in AU, NZ, North America, UK and Europe, launching with four named agents (FieldReady, JobReady, JobScribe, JobBrief) plus 'Cooper', the AI brain, and 'agentic workflows'. Bhatti appointed CTO two weeks later (27 May 2026) with a mandate to champion 'an AI-first approach' across all global engineering. Pain points (INFERRED): brand-new CTO inheriting a 4-agent fleet that shipped 14 days before he arrived, across 3 codebases and 5 geographies — he owns the run cost and the reliability of something he did not architect. Challenges (INFERRED): agents deployed to 20,500 SMB trade customers, i.e. very high run count at low ACV, so unit economics per agent run are existential; three separate product stacks means no single observability plane. Must-haves (INFERRED): fleet-wide visibility into what each agent costs and does, normalised across Simpro/AroFlo/BigChange; a way to prove the agents work before adding a fifth. Nice-to-haves (INFERRED): per-customer cost attribution to protect gross margin at SMB price points. ICP confidence: HIGH — named CTO, global engineering ownership, in-range headcount, dated 2026 multi-agent production launch. TIMING NOTE: new-CTO window (appointed 27 May 2026) is the single best moment to reach him — he is choosing his platform stack right now. His background is enterprise SaaS scale (VP Eng at Vertex, UKG, Citrix), not AI-native, so lead with operational control and unit economics, not model sophistication.
Sam Partee — Co-Founder & CTO, Arcade.dev (Arcade AI)
Signal: 1 (speaks/writes on 'Agents in Production' & agentic tool-calling; MLOps Community talk https://home.mlops.community/public/videos/agentic-tool-calling-samuel-partee-agents-in-production-2024-11-15 ; theCUBE panel). Source: https://www.hpcwire.com/aiwire/2026/06/16/arcade-secures-60m-to-scale-authorization-and-governance-for-ai-agents/ ; https://theorg.com/org/arcade-ai/org-chart/sam-partee . Company size: ~40-60 (Series A $60M June 2026, ~$72M total; ex-Redis Principal Applied AI Engineer). Pain points: agent reliability, tool-calling failures, authorization/governance for production agents, controlling agent behavior at scale. Challenges: securing & governing agent access; making tool calls reliable; tool-call volume up 25x in 6mo. Must-haves: reliability + governance/control layer for production agents. Nice-to-haves: standardized auth (authored MCP auth spec). ICP confidence: Medium (Series A AI-native building agent infra; competitor-adjacent to an agent control/operating layer).
Sam Taylor — SVP Technology (leads AI, Data and Engineering), Cleo (Cleo AI Ltd)
Signal: 4 (ICP shipping agents in production; found via engineering-blog + org research, title verified live on LinkedIn people search 2026-09-02 — headline reads "SVP Technology @ Cleo | Leading AI, Data and Engineering Teams", London UK).
Source: https://web.meetcleo.com/blog/introducing-cleos-custom-router ; https://theorg.com/org/cleo-ai/org-chart/sam-taylor ; LinkedIn people search verification.
Company size: ~590 (ZoomInfo company profile, zoominfo.com/c/cleo-ai-ltd/412960705). London, UK. Consumer fintech, Series C. 7M+ users.
Pain points: Cleo runs a fleet of specialist agents plus background agents running 24/7. They REPLACED their LLM-based agent router (GPT-5.4-nano, ~800ms per message) with a hand-trained custom encoder model that runs ~16x faster and beats it on accuracy — i.e. they built bespoke infra because per-call routing latency and inference spend were unacceptable. No evidence of a per-run cost attribution layer across the agent fleet.
Challenges: per-message latency and inference spend across a multi-agent fleet at consumer scale; keeping a self-built router accurate as agents are added; currently hiring a VP of AI (LLMs, AI agents), which signals an ownership/capacity gap in exactly this area.
Must-haves: per-agent and per-run cost + latency attribution; routing that does not require training and maintaining an in-house model.
Nice-to-haves: eval/regression testing on router accuracy; spend forecasting by agent; anomaly alerting on cost spikes.
ICP confidence: High — Director+ (SVP), 590 employees (in 50-2,000 band), documented multi-agent production estate, and demonstrated willingness to build rather than buy the exact layer thealpha.ai sells. Net-new name AND net-new company vs the 1,071-record library.
Signal: 4 (ICP technical co-founder building specialized agents on top of meeting data). Source: https://www.crunchbase.com/person/sam-udotong ; https://fireflies.ai/about-us ; https://pitchbook.com/profiles/company/172275-22 . Company size: ~100–121 (2026). Funding: ~$19M raised (Series A, latest priced round 2021); reached $1B valuation June 2025 via tender offer, not a new round — so still early-stage/independent. Actively building AI agents: Fireflies is moving from notetaker to "AI teammate," publicly positioning that 2026 SaaS uses specialized agents for key workflows (meeting capture, summarization, task/action agents). Pain points (inferred from public commentary, not verbatim): building reliable specialized agents on messy meeting data; cost/efficiency of running agentic workflows at consumer + Fortune 500 scale. Challenges: agent reliability and cost as workflow agents proliferate. Must-haves: reliability, cost efficiency. Nice-to-haves: observability, compounding intelligence over meeting history. ICP confidence: Medium (technical co-founder/CTO at 100+ person independent company shifting to agentic product; less clearly a heavy multi-agent-in-production shop than others, hence Medium not High).
Signal: 4 (senior eng leader at agent/inference-infra company). Discovery: company-scoped LinkedIn People search (Baseten under-mined — only CTO previously captured). Source: https://www.linkedin.com/in/sparanjpye/ | Company: Baseten — AI inference platform powering production & agentic (Chains / compound AI) workloads. Company size: ~312 employees (PitchBook 2026); Series F, $13B val — in band. Pain points (INFERRED from role): cost-per-inference/per-run at scale, reliability of inference in production, scaling inference for agentic workloads. Challenges: production reliability + cost governance across customer inference/agent workloads. Must-haves: cost/latency visibility per run, reliability. Nice-to-haves: automated optimization. ICP confidence: Medium-High (SVP Engineering; infra powering agents — already treated as ICP in brain via CTO). Note: infra-adjacent ICP.
Sami Ghoche — Co-Founder & CEO (formerly CTO), Forethought
Signal: 4 — technical co-founder at a qualifying company shipping agents. Sami Ghoche is technical Co-Founder & CEO of Forethought (Deon Nicholas is now Exec Chairman/President). Forethought is an agentic AI platform for customer experience/support, running support agents in production.
Source: https://forethought.ai/about ; https://www.crunchbase.com/organization/forethought-7dc2
Company size: ~138 employees (2024, confirmed range); ~$90M raised, Series C.
Pain points: support-agent reliability in production; cost per successful resolution vs containment tradeoff; scaling agents across many customer deployments.
Challenges: keeping automated resolutions accurate while controlling LLM cost; visibility into agent performance/cost per interaction.
Must-haves: reliable containment, cost-per-resolution visibility, control over agent behavior.
Nice-to-haves: outcome-based measurement.
ICP confidence: Medium-High — technical co-founder & CEO, ~138-employee Series C agentic-CX company shipping agents in production.
Sami Shalabi — Co-founder & CTO, Maven AGI
Signal: 4 (ICP CTO at a qualifying company building & shipping agents in production). Source: https://www.mavenagi.com/about-us and https://tracxn.com/d/companies/maven-agi/ . Company size: est. ~100-150 ($78M raised over 2 rounds, Series B; named to The Agentic List 2026 in CX Agents). Enterprise AI agent platform automating customer support across chat/email/voice/web, resolving up to 93% of queries. Pain points: scaling agent resolution rates while keeping quality; cost as a barrier (noted high cost for smaller firms → cost/pricing pressure at scale); reliability across channels. Challenges: production reliability across multiple channels and enterprise customers; per-resolution cost economics; visibility into agent performance. Must-haves: production-grade reliability + resolution accuracy; cost predictability at enterprise scale. Nice-to-haves: cross-channel observability. ICP confidence: Medium-High — technical co-founder/CTO decision-maker at a ~100-person AI-native agent platform actively shipping in production; size squarely in ICP band.
Sami Tas — VP of Engineering and AI, MaintainX
Signal: 4 (ICP speaking publicly about building and shipping agents). Found via web research — LinkedIn/Chrome unavailable this run.
Source: https://www.facilitiesdive.com/news/maintainx-touts-ai-solution-for-preparing-maintenance-reports/815686/ (25 Mar 2026 interview)
Company size: ~740–939 employees (Revelio Labs 740 as of Mar 2026; Tracxn 939 as of 30 Jun 2026; PitchBook 902). AI-native maintenance/CMMS for manufacturing & facilities. Late-stage funding is the only ICP stretch.
Agent evidence: leads engineering + AI for MaintainX's role-based AI agents and the Report Builder AI launch (Mar 2026).
Verbatim: "Today, all the knowledge lies in manuals, PDFs and people's brains. As those people retire, that information will be largely lost... Our goal is [to] capture that institutional knowledge [so] people can diagnose and fix issues faster." Says the company is working "to move AI tools from straightforward point solutions... to more holistic uses that can help provide actionable insights and assist technicians' judgment." Cites "60% of the time is spent on researching and diagnosing the issues, and only 40% of the time is spent with the wrench."
Pain points: institutional knowledge trapped in unstructured manuals/PDFs and retiring workers; technicians burning most of their time on diagnosis instead of repair; customers cannot self-serve BI/queries across thousands of assets.
Challenges: moving from single-purpose AI features to a holistic agentic layer; grounding natural-language queries in work-order/parts/PO/asset-history data reliably enough for operational decisions; agent orchestration infrastructure at ~900-person scale.
Must-haves: natural-language-to-report/query generation grounded in customer data; agent outputs that give a correct starting point a human can edit; role-based agents mapped to real maintenance/reliability workflows.
Nice-to-haves: root-cause-analysis classification modules; cross-system context (parts, vendors, downtime) so agents reason beyond a single asset.
ICP confidence: High — VP Eng & AI title is a bullseye, AI-native vertical SaaS shipping agents in production, headcount in band. The "point solutions → holistic agentic layer" transition he describes is exactly the 1→5+ agents wall.
Signal: 4 (newly appointed ICP technical decision-maker at a company scaling agents in production). Source: https://www.businesswire.com/news/home/20260521475981/en/Abridge-Names-San-Oo-CTO and https://medcitynews.com/2026/06/abridge-clinical-ai/ (no public LinkedIn profile URL surfaced — left blank rather than guess). Company size: ~635 employees (66% YoY growth), clinical AI, Series-scale healthcare. Named CTO May 2026 (from Notion, where he focused on reliability, developer velocity, and "agentic engineering"; prior Slack, CTO ShopBack). Abridge is moving "beyond documentation" into clinical AI agents (coding, workflow). Pain points: reliability of clinical AI at scale, agentic engineering velocity, safety/accuracy in regulated healthcare. Challenges: standing up multi-agent clinical workflows reliably as headcount and usage grow fast. Must-haves: production reliability, visibility/guardrails on agent behavior in a regulated setting. Nice-to-haves: cost visibility per run, faster dev loop for agent teams. ICP confidence: High (CTO title, explicitly focused on agentic engineering; company in size band and expanding agent surface). Note: predecessor Zachary Lipton already in brain as former Abridge CTO / advisor.
Sanchit Sood — Chief AI Officer, Kapture CX
Signal: 1 (ICP publishing on agent governance, cost and ROI).
Source: https://cxotoday.com/expert-opinion/agentic-ai-vs-traditional-automation-where-enterprises-are-actually-seeing-better-roi/ (1 May 2026, his byline); title verified on https://www.kapture.cx/about-us/ (page updated 23 Jun 2026).
Company size: ~613 employees as of Jul 2026 per LeadIQ (https://leadiq.com/c/kapture-cx/5a1d99c02300005b00889eaa); Tracxn showed 325 as of Aug 2025. Estimate 500–650 global, inside band. $10M pre-Series B Jul 2026 led by Bajaj Finserv Ventures.
Pain points: (verbatim) "Agentic AI is probabilistic. In regulated industries like BFSI, insurance, and healthcare, that is not a small concern. It is a real constraint." Also (verbatim) "This is also why governance is becoming central to ROI. Gartner has warned that over 40% of agentic AI projects could be cancelled by the end of 2027 due to costs, unclear value, or inadequate risk controls… unmanaged agentic AI will fail."
Challenges: Non-determinism blocking agent deployment in regulated verticals; proving ROI against agent cost; governance gap threatening project survival.
Must-haves: Cost-to-value attribution per agent; risk controls that satisfy BFSI/insurance/healthcare compliance; defensible governance layer.
Nice-to-haves: Benchmarking agentic vs. traditional automation ROI; per-vertical control templates.
ICP confidence: High — C-level AI seat at a company squarely inside the headcount band, actively shipping agentic CX at enterprise scale, and he names cost and unclear value as the top kill risk for agent programs.
Watch-out: Kapture ships its own "Observability Platform for AI Agents" + CALIBRATE audit layer — either a conflict or a "you already tried building it" wedge. Kamath (CEO) publicly claims profitability, so cost-savings framing likely lands better than budget-unlock framing. Kapture is also actively hiring a VP Engineering (Bangalore) — eng leadership bench being built now.
Sanjay Dash — Chief Engineer, Auger
Signal: 4 (senior technical leader at a Series B company shipping AI agents in production). Source: https://www.geekwire.com/2026/supply-chain-startup-auger-led-by-ex-amazon-operations-chief-raises-50m-and-lands-big-customers/ ; https://erp.today/auger-series-b-ai-agents-supply-chain-stack ; https://rocketreach.co/sanjay-dash-email_52392551 ; https://www.stattimes.com/logistics/dave-clark-announces-founding-team-of-auger-1353901. Company size: ~130 employees (Auger, Seattle; founded 2024 by ex-Amazon ops chief Dave Clark; Series B $50M led by Eclipse, ~$150M total; in 50-2,000 band). Ships agents: Auger's platform sits above ERP/WMS/TMS and uses AI agents + optimization models to handle routine supply-chain calls/decisions automatically ("autonomous supply chain operating system"). Role: Chief Engineer (senior technical leadership; led Just Walk Out, Dash Cart, Amazon One at Amazon; 30 yrs in tech), above Director level — fits ICP. Pain points (inferred from domain, NOT verbatim quotes): reliability of agents acting on fragmented operational/ERP data; agent decision accuracy at supply-chain scale; per-run cost/observability as agent volume grows. Challenges: making multi-system agentic workflows reliable and production-grade. Must-haves: reliability, control/governance of autonomous actions, cost visibility. Nice-to-haves: observability across runs. ICP confidence: Medium-High (named Chief Engineer, ~130-emp Series B company actively shipping agents; profile_url is the Seattle Sanjay Dash surfaced in search and matches Auger's Chief Engineer — verify before outreach). Found via web research (LinkedIn/Chrome unavailable this run).
Sanjay Jeyakumar — CTO & Co-Founder (Head of R&D), Abnormal AI (Abnormal Security)
Signal: 1/4 (ICP writing about AI-automated security and shipping autonomous agents in production). Source: https://www.linkedin.com/in/sanjay-jeyakumar-2253091/ , LinkedIn posts "An Abnormal Vision for Fully AI-Automated Cybersecurity" and "From Abnormal Security to Abnormal AI". Company size: ~1000-1400 (Abnormal AI, email/cloud security, late-stage ~$5B+ valuation; product suite now includes multiple autonomous agents: AI Security Mailbox, autonomous SOC workflow agents, AI Phishing Coach). Pain points: running multiple autonomous SOC/security agents reliably in production; making fully AI-automated cybersecurity trustworthy and accurate. Challenges: architecting an AI-native platform that orchestrates many agents at enterprise scale; reliability and control of autonomous agent actions. Must-haves: reliability and trust in autonomous agent decisions, control/observability of agent behavior in security-critical workflows. Nice-to-haves: cost efficiency as agent count and volume scale. ICP confidence: High (CTO & technical co-founder, in-band company shipping 5+ agents in production).
Sanjog K. — Director of Product Engineering, Uniphore
Signal: 4 (company-scoped LinkedIn people search of agent-native company Uniphore; ICP-title Director engaging agent-platform work). Source: https://www.linkedin.com/search/results/people/?keywords=Uniphore%20agentic%20AI%20director%20engineering | Company size: ~1,000-1,200 (est; Series E). Pain points (INFERRED from role+company, not observed posting): cost/reliability of enterprise agentic-AI (Business AI) products at scale; per-interaction cost visibility across voice+chat agents. Challenges: scaling multi-agent CX deployments reliably in production. Must-haves: production reliability + governance for enterprise agents. Nice-to-haves: per-run cost attribution/observability. ICP confidence: High (Director of Product Engineering at agent-native 50-2,000-emp company; company confirmed via explicit '@ Uniphore' headline; net-new vs 8 existing Uniphore contacts in brain).
Sarah Buchner — Founder & CEO (technical founder; PhD data science / civil engineering, MS architectural engineering), Trunk Tools
Signal: 4 (ICP-matching technical founder publicly describing the 1→many agent scaling problem). Source: https://www.globenewswire.com/news-release/2026/06/17/3313698/0/en/trunk-tools-launches-cortex-to-tackle-construction-s-hardest-ai-problem-drawings.html ; https://www.enr.com/articles/63178-trunk-tools-launches-cortex-ai-platform-to-interpret-construction-drawings ; https://runtimewire.com/article/trunk-tools-sarah-buchner-construction-ai-agents ; https://bricks-bytes.com/ai/trunk-tools-copilot-to-system-of-action/. Company size: 51-200 employees (Crunchbase/Tracxn band). Series B — $40M led by Insight Partners, Jul 2025; ~$70M raised total. Cortex platform (launched Jun 2026) carries SEVEN agents in production covering drawing review, RFIs, bids and submittals; deployed on hundreds of active US jobsites (Gilbane, Suffolk, HITT, DPR, Harkins, Consigli, Torcon, McGough, Cleveland Construction). Pain points: scaling from single-purpose copilots to multiple coordinating agents — her words: they've "gone from agents that assist individuals to agents that work together, sharing context and acting without waiting for a human to connect the dots"; heavy-context workloads (full drawing sets) make context/token cost material. Challenges: shared context across seven agents without runaway token cost; reliability and accuracy on high-consequence construction documents; proving trust on jobsites where errors are expensive. Must-haves: agent-to-agent context sharing that is observable, reliability at document scale, per-agent accountability. Nice-to-haves: cost-per-run visibility across the seven-agent fleet, drift detection as drawing sets change. ICP confidence: Medium-High (technical founder/CEO at a 51-200 person Series B company with 5+ agents demonstrably in production; ICP allows technical founders. Downgraded from High because no direct public statement on cost/token spend was found this run — the 1→many scaling signal is explicit, the cost pain is inferred).
Sarah Sachs — AI Lead / Eng Lead, AI (Head of AI Engineering), Notion
Signal: 2 & 4 — featured guest on a 'Notion in Practice' LinkedIn video about building AI agents teams can trust (found via content search: "Head of AI" agents production cost reliability). Source: https://www.linkedin.com/search/results/content/?keywords=%22Head%20of%20AI%22%20agents%20production%20cost%20reliability . Company size: ~800-1,500 (est). Pain points: agent sprawl; model/vendor lock-in; observing agents in production for BOTH reliability AND cost; deciding what to hand to AI vs deterministic code; permissioned context as critical infra. Challenges: shipping agents the team can trust without lock-in. Must-haves: production observability of agents (reliability + cost per run), permissioned context mgmt. Nice-to-haves: reasoning-vs-plumbing separation patterns. ICP confidence: High — heads AI engineering at an agent-shipping mid-size company; strongest on-signal content of the run.
Signal: 4 (ICP at a company shipping multiple named agents in production). Source: https://klue.com/about, https://klue.com/careers, https://klue.com/claude-integration — title verified live on LinkedIn 2026-08-30 ("CTO / Cofounder @ Klue"). Company size: ~198 employees (LeadIQ, May 2026); Vancouver BC, Canada; competitive intelligence / win-loss SaaS; Series B, $81M+ raised (Tiger Global, Salesforce Ventures, Craft). Pain points: classic 1-to-5+ agent scaling profile — Klue ships at least three named production agents (Compete Agent, Win & Loss Story Agent, and a voice AI Interviewer marketed as "scaled across every deal in your pipeline") plus an MCP server listed in the Claude Connector Directory. Voice agents and always-on web-crawling agents are the two most token-heavy shapes there are. Challenges: attributing cost per agent run across three distinct agent products with very different cost profiles; an MCP surface exposed to customer-side agents means unbounded third-party-driven tool-call volume. Must-haves: per-agent and per-run cost attribution; reliability guarantees on a voice agent that runs on every deal. Nice-to-haves: cross-agent comparison of unit economics. Actively hiring a "Senior Software Engineer, Agents" — a live buying-intent signal. ICP confidence: High — technical co-founder/CTO, Series B, ~200 headcount, multiple agents demonstrably live, actively staffing an agent team.
Signal: 4 (ICP seat at a company shipping agents; buying signal visible in their own job descriptions rather than in his commentary).
Source: title verified on https://www.freehand.ai/about-us. Company context: Series B release 29 Jul 2026 (GlobeNewswire, via https://www.manilatimes.net/2026/07/29/tmt-newswire/globenewswire/freehand-raises-75m-...).
Company size: Crunchbase band 101–250 (https://www.crunchbase.com/organization/freehand-73c2); Dealroom shows 100–500. Freehand has never published a headcount. Best estimate low-to-mid 100s — inside band. Series B $75M Jul 2026, co-led by Battery Ventures + NewRoad Capital; $100M total. Founded Feb 2024 by the Pando founders, out of stealth Feb 2026. Chicago HQ + Chennai engineering.
Pain points: No personal public quote found — zero public speaking/writing footprint under the Freehand name. Buying signal is explicit in their engineering JDs (verbatim, https://builtin.com/job/backend-engineer/9724436): "Build and scale backend services and orchestration for AI agents managing procure-to-pay workflows. Implement APIs, integrations with ERPs and payment systems, observability, auditability, secure data pipelines, and testing/eval frameworks to ensure reliable, compliant AI-driven operations at enterprise scale." AI PM JD adds: "design agent patterns, observability, and closed-loop training... ship reliable, explainable agentic workflows."
Challenges: Building agent observability, auditability and eval frameworks in-house right now, for Fortune 500 procure-to-pay workloads.
Must-haves: Agent observability and audit trail; reliability/eval harness; enterprise-grade compliance on agent actions.
Nice-to-haves: Closed-loop training from production runs.
ICP confidence: Medium — exact ICP title verified on a company-owned page and headcount band inside range, and the company is demonstrably building Alpha's category in-house (strong timing signal). Downgraded from High because headcount is a band not a figure and he has expressed no pain publicly.
Watch-out: do NOT conflate with the unrelated Saravana Kumar, CEO of Kovai.co, who dominates search results. Co-founder Abhijeet Manohar (CPTO) is already in the People Library — this is a second contact at the same account.
Sarthak Jain — Co-Founder & CEO (technical), Nanonets
Signal: Bucket 1 (technical co-founder at agent company) — web research pivot (LinkedIn/Chrome unavailable). Source: https://www.cbinsights.com/company/nanonets/people ; https://techcrunch.com/2024/03/12/nanonets-funding-accel-india/. Company size: ~100-200 employees. Technical co-founder (ML background) of Nanonets, which ships an AI agent workforce for document-heavy processes (AP, order mgmt, logistics, healthcare RCM); backed by Accel; 34% of Fortune 500 customers. Pain points: cost and reliability of autonomous document agents at billion-document scale; visibility into per-run/per-document agent economics. Challenges: scaling agent accuracy and cost efficiency across many enterprise workflows. Must-haves: cost-per-run visibility, reliability guardrails. Nice-to-haves: unified agent observability. ICP confidence: Medium-High — CEO but a genuine technical co-founder (C-suite technical leader, valid per ICP) at a 50-2,000 emp agent company.
Signal: 4 (ICP technical co-founder at a qualifying agentic-AI company). NOTE: LinkedIn/Chrome unavailable this run; found via web research + primary-source verification. Source: Kasisto product pages + Crunchbase person profile + PRNewswire "Kasisto Launches KAIgentic" (2025). Company: Kasisto (NYC) — KAI conversational/agentic AI platform for banking & finance; launched KAIgentic ("AI that thinks like a trusted banker") 2025; customers incl. major banks (Standard Chartered, TD, Westpac, DBS historically). Company size: ~50–72 employees (2026, LeadIQ/PitchBook) — within ICP band (near lower floor). NOTE: Kasisto was acquired by Backbase (Jun 2026); still operates as a distinct agentic-banking AI unit. Pain points (inferred from role/domain, not a personal quote): production reliability & accuracy for regulated banking agents; controlling LLM/token cost per conversation at scale; moving from single virtual assistant to multi-agent "agentic" banking. Challenges: hallucination/compliance risk in finance; cost predictability. Must-haves: reliability, guardrails, per-run cost visibility. Nice-to-haves: agent orchestration/observability. ICP confidence: Medium-High (confirmed technical C-suite, agent-shipping fintech; headcount in band though near floor, and recent acquisition is a minor caveat). Pain framing inferred from documented product/domain, not a fabricated quote.
Sassun Mirzakhan-Saky — Co-Founder & CTO, Synthflow AI
Signal: 4 (technical co-founder at agent-native company; found via company-targeted people-search, identity web-confirmed as Synthflow Co-Founder & CTO). Source: https://www.linkedin.com/search/results/people/?keywords=Synthflow%20voice%20AI%20agents (profile https://www.linkedin.com/in/sassun-ms/). Company size: ~72 (confirmed) — Synthflow AI, Series A $20M (Accel/Atlantic Labs/Singular); enterprise voice AI agents. Pain points [INFERRED from verified role+company, NOT verbatim]: agent cost + real-time voice reliability at scale; per-call cost visibility; scaling concurrent agents 1->many. Challenges: enterprise-grade reliability/latency; de-risking voice AI (launched OpenAI-powered BELL framework). Must-haves: reliability; cost-per-call visibility. Nice-to-haves: observability; model routing. ICP confidence: High (technical co-founder/CTO, agent-native, right size/stage; NET-NEW — company had only CEO in brain).
Satish Agrawal — VP, Data and AI, Transit Technologies
Signal: 4 (ICP technical AI leader shipping agents; found via LinkedIn People search for ICP titles + agents). Source: https://www.linkedin.com/search/results/people/?keywords=VP%20of%20AI%20shipping%20agents%20startup ; verification: https://tracxn.com/d/companies/transittechnologies/ and https://www.transit-technologies.com/use-case/operations. Company size: ~283 employees (Tracxn, Dec 2024; 201-500 on LinkedIn), Knoxville TN, founded 2019, transit/fleet software (SaaS). Actively shipping agents (CONFIRMED): Intelligent Voice Agent (IVA) lets riders check/change/cancel trips; IVA for Service Request Creation automates maintenance reporting and work-order creation — multiple voice agents in production. Satish owns AI strategy, product delivery and AI-native products incl. AI voice agents. Pain points (inferred + product context): scaling voice agents to absorb high call volume reliably; automating work-order creation accurately; cost/efficiency per agent call. Challenges: reliable, cost-efficient voice agents across agency customers. Must-haves: reliability at call volume, cost control per run. Nice-to-haves: observability into agent decisions. ICP confidence: Medium-High (VP Data & AI at 283-emp transit SaaS in band actively shipping voice agents). LinkedIn profile URL not captured this run — do not fabricate.
Signal: 4 (ICP technical co-founder/CTO building agentic capabilities into a production data platform). Source: https://www.linkedin.com/in/satish-jayanthi-32703613/ ; https://coalesce.io/company/ ; https://coalesce.io/company-news/coalesce-launches-ai-governance-features-snowflake-summit-2025/ . Company size: est ~100-250 (unconfirmed exact figure this run; ~$50M Series B in 2023; well-funded data-infra scale-up; acquired CastorDoc). Actively shipping agents: Coalesce ships AI-powered data transformation + cataloging with an agentic Copilot architecture and AI governance features (launched Snowflake Summit 2025). Jayanthi = technical co-founder & CTO (ex-WhereScape data-warehouse automation). Pain points (inferred from product/positioning, NOT verbatim quotes): governance/control of agentic workflows acting on enterprise data warehouses; reliability of agent-generated transformations; cost of agent runs at data-pipeline scale. Challenges: enterprise trust/governance for agentic data ops. Must-haves: agent governance/observability, reliability, cost visibility. Nice-to-haves: agents that compound on metadata. ICP confidence: Medium (clear technical CTO + agentic product; headcount likely in range but exact figure unconfirmed this run — flag to verify).
Satya Nitta — Co-founder & CEO (technical; ex-IBM Research, Global Head of AI Solutions), Emergence AI
Signal: 4 (ICP building/shipping agents in production; writing about multi-agent orchestration, governance, reliability). Source: https://www.emergence.ai/about-us and LinkedIn profile. Company size: 116 employees (in range), ~$97.2M raised. Product: mission-critical agentic infrastructure for enterprise — "Orchestrator" coordinates multiple autonomous agents at runtime across domains; verified, governed agents that plan/reason/act (semiconductor design, enterprise ops); released Agent-E web control agent. Pain points: coordinating/governing multiple autonomous agents at runtime; making agents verified and reliable for mission-critical enterprise use. Challenges: scaling from single agent to coordinated multi-agent systems without losing control/reliability. Must-haves: runtime orchestration, verification, governance of agent fleets. Nice-to-haves: cross-domain agent coordination, benchmark-leading reliability. ICP confidence: Medium — right-sized AI-native company shipping agents in production; note their "agentic infrastructure/Orchestrator" positioning partially overlaps the operating-layer category.
Signal: 4 (ICP speaking at AI Engineer World's Fair 2026 on shipping healthcare AI agents in production). Source: https://www.ai.engineer/worldsfair/schedule. Company size: ~50 employees (Series B, $64M total). Anterior ships the 'Florence' clinical AI agent for prior authorization. Pain points (inferred): production reliability of agents, latency/cost of multi-step clinical reasoning chains, visibility into what agents do per case. Challenges: hardening agent infra as they scale across payers. Must-haves: reliability, observability, cost control as agent volume grows. Nice-to-haves: automated evals / regression detection. ICP confidence: High — VP of Engineering at a Series B company actively shipping agents; core ICP persona.
Saurabh Anand — Head of Product (AI/agent-focused), Emergent
Signal: 1 & 4. Source: https://www.linkedin.com/in/saurabh-anand-rai/ ; speaks at events on frontier AI agents / agentic systems architecture. Company size: ~310 employees; Series C unicorn ($1.5B) shipping autonomous coding agents, $100M+ ARR. Pain points: shipping reliable autonomous coding agents at scale; agent quality/reliability; product-level visibility into agent behavior and cost. Challenges: productizing agentic coding for millions of users. Must-haves: reliability, control over agent behavior. Nice-to-haves: cost-per-run visibility. ICP confidence: Medium (Head of Product = AI-focused product leadership at Series C agent-native company; product rather than eng role).
Signal: 1/4 (ICP at agent company authoring about autonomous digital workers/agents). Source: https://www.linkedin.com/in/saurabhdhupar/ (headline 'AI Engineering @ 11x'; title 'Head of AI Engineering' per RocketReach/search). Company size: ~100-200 emp (11x; 'home of AI digital workers'; autonomous AI SDR/agents; ex-Brex CTO Prabhav Jain). Pain points (inferred from role + his posts, labeled as such): making fully-autonomous digital-worker agents reliable in production; cost/economics of always-on autonomous agents; scaling agent workflows for enterprise. Challenges: reliability of non-deterministic autonomous agents at scale; per-agent cost control. Must-haves: per-agent/per-run cost visibility + reliability guardrails for autonomous agents. Nice-to-haves: routing/eval to control LLM spend. ICP confidence: Medium-High (Head of AI Engineering at in-range agent-native company; verified current at 11x; headline generic, senior title per external source).
Saurabh Dhupar — Head of AI Engineering, 11x
Signal: Bucket 4 (ICP leader building/shipping agents at a qualifying company). NOTE: LinkedIn/Chrome not connected this run; found via web research + verification.
Source: https://theorg.com/org/11x ; https://rocketreach.co/saurabh-dhupar-email_2872475 ; https://www.linkedin.com/in/saurabhdhupar/ . Head of AI Engineering at 11x; 15+ yrs, ex-Wealthfront/Uber. Non-founder (founder/CEO Hasan Sukkar; CTO Prabhav Jain — both already in library). (Side note: also co-founder/CTO of RevFX, a separate small venture; primary role recorded is his 11x leadership post.)
Company size: ~100–200 employees (Series B, a16z/Benchmark-backed).
Ships agents: 11x builds autonomous AI "digital workers" / SDR outreach agents (Alice, Julian) that run continuous multi-step GTM workflows.
Pain points: high token burn from unattended multi-step agents; reliability/hallucination risk in autonomous outreach; need for cost + per-run observability.
Challenges: keeping autonomous agents accurate and on-budget while scaling to many concurrent digital workers.
Must-haves: cost-per-run tracking and reliability controls for always-on autonomous agents.
Nice-to-haves: centralized control/observability across a growing agent fleet.
ICP confidence: High — Head-of-AI-Engineering (non-founder, above Director) verified; AI sales-agent vendor in size range.
Signal: 4 (ICP-title AI leader at a qualifying agentic-platform company). Source: https://www.linkedin.com/in/saurabhhebbalkar/ (LinkedIn people search "Director of AI agents LLM production"). Company size: 51-200 employees (Krista / Krista.ai, Dallas TX; verified via LinkedIn company page — "enterprise agentic platform," raised $15M, founded 2020). Confirmed 50-2,000 band and agent-native (orchestrates processes across people, apps and AI). Role: Director of AI (headline: "Self-Hosted LLM Infrastructure, Voice AI, Model Evals, Agentic AI & MLOps"; Pune). Pain points (from headline + role, partly inferred): self-hosted LLM infra cost/ops; model evals for agentic workflows; voice-AI reliability in production. Challenges: running agentic orchestration reliably and cost-efficiently on self-hosted infra. Must-haves: model evals, reliable agent orchestration, cost/infra control. Nice-to-haves: voice-AI quality, MLOps tooling. ICP confidence: Medium-High (Director of AI at an in-range, explicitly agentic-platform company; role directly touches agent evals/infra/cost). profile_url captured in Source too.
Signal: 4 (ICP technical leader at a company actively shipping agents — Squirro launched a 13-agent enterprise catalog spanning finance, HR, legal, sales, ops, IT). Source: https://www.prnewswire.com/news-releases/squirro-launches-ai-agent-catalog-to-end-the-start-from-zero-problem-stalling-enterprise-ai-302829144.html ; https://ch.linkedin.com/in/saurabh-jain-67157314 . Company size: ~95-104 employees (PitchBook/LeadIQ), Series B, Zurich. Pain points: enterprises stall going from zero to production agents; reliability/trust of agents operating over unstructured enterprise data; governance and auditability. Challenges: connecting unstructured enterprise data to structured taxonomies; deploying and maintaining 13+ agents reliably across functions. Must-haves: reliable, governable agents; visibility into agent behavior; pre-built accelerators to avoid "start from zero." Nice-to-haves: lower cost per agent run; faster time-to-value. ICP confidence: High (CTO, Series B, ~100 emp, multiple agents in production).
Saurabh Saxena — Head of Technology / Senior VP (R&D) — Agentic AI, GenAI, Emotion AI Platforms, Uniphore
Signal: 4 (senior technical leader at large agent company; LinkedIn people search "Uniphore VP engineering AI"). Source: https://www.linkedin.com/in/saurabh-saxena-08747a5/ . Company size: ~1,000+ employees (Uniphore; enterprise conversational/agentic AI; already an ICP company in the brain). Pain points (INFERRED from verified role+company, NOT verbatim): scaling multi-agent + voice AI platforms in production; reliability and cost of enterprise agent deployments; observability across many concurrent agent runs. Challenges: enterprise-grade reliability + unit economics of agentic platforms. Must-haves: cost/observability + control per agent workflow. ICP confidence: HIGH — SVP R&D of Agentic AI, VP+ technical leader, agent-native, size fits. Caveat: pain inferred.
Scott Kennedy — VP of Engineering (headline: Engineering Lead), Replit
Signal: 1 (ICP authoring about agent/production cost & pricing). Source: https://www.linkedin.com/in/stkenned/ (LinkedIn posts, Jul-Aug 2026). Company size: ~200-400 emp (Replit; $400M Series D, $9B val; ships Replit Agent). Pain points: hosting/compute cost at scale for agent workloads (posted Replit cut hosting costs up to 80%, passing discounts to customers); comparing platforms on cost/pricing pages; measuring AI/agentic-coding ROI via productivity metrics; evaluating models. Challenges: running tens of thousands of agents in parallel economically; app-layer eval of agents (ViBench); scaling infra while dropping prices >50%. Must-haves: per-run/per-agent cost visibility, cost control at scale, model routing/eval. Nice-to-haves: benchmarking agents on real workloads, productivity/ROI dashboards. ICP confidence: High (senior eng leader at agent-native company; verified current at Replit; multiple sources incl. The Org & Replit list him VP of Engineering; active cost-focused content).
Scott Kurinskas — VP Product, Agentic AI Platform, C3 AI
Signal: 4 (ICP leader at company shipping agents; found via people search 'VP AI agents production SaaS'). Source: https://www.linkedin.com/in/scottkurinskas/ . Company size: ~900 employees (C3.ai, enterprise AI, publicly traded — in range). Role: VP Product for C3's Agentic AI Platform (AI-focused product leadership). Pain points (inferred from role/company, no direct post captured this run): owning a production agentic AI platform — reliability, per-run cost visibility, scaling many agents for enterprise customers. Challenges: proving ROI/cost-efficiency of agent platform at enterprise scale. Must-haves: production observability + cost-per-run control. Nice-to-haves: agent governance/guardrails. ICP confidence: Medium (senior AI product leader at in-range company actively shipping an agent platform; pain inferred from role, not an authored post).
Scott Metcalf — Head of AI Customer Innovation (billed by SaaStr as VP AI Innovation), People.ai
Signal: 4 (ICP-titled AI leader speaking publicly about running agents in production). Source: https://www.saastr.com/the-first-44-speakers-for-saastr-ai-annual-2026-the-founders-and-operators-actually-shipping-ai-at-scale/ — SaaStr AI Annual 2026 (12-14 May 2026) speaker: "Scott Metcalf, VP AI Innovation, People.ai. How they cut agent oversight from 30% of an IC's week to under 5%." Title corroborated as "Head of AI Customer Innovation" via RocketReach. Company size: ~215-216 employees as of Mar 2026 (Tracxn / Revelio); ~$62.8M ARR; ~$200M raised across 6 rounds (late-stage). AI revenue platform shipping agents for sales/RevOps execution across ~500 enterprise customers. Pain points: human oversight burden on agents — 30% of an individual contributor's week spent supervising agent output before they drove it under 5%; trusting agent output enough to remove the human in the loop. Challenges: measuring and reducing the babysitting tax on production agents; proving agent reliability to enterprise revenue teams. Must-haves: agent reliability high enough to cut supervision, measurable quality signals per agent run. Nice-to-haves: cost/efficiency attribution per agent, automated escalation thresholds. ICP confidence: Medium (senior AI leadership title and 215-person headcount both fit, and the oversight-cost signal is a real, quantified production pain; downgraded because People.ai is late-stage rather than Series A-C, and no LinkedIn profile URL was confirmed this run — left blank rather than fabricated).
Scott Stevenson — Co-Founder & CEO (technical), Spellbook
Signal: 4 (ICP shipping agents; Spellbook launched autonomous contract management / agentic workflows). Source: https://www.businesswire.com/news/home/20251009110230/en/Spellbook-Raises-$50M-Series-B-to-Expand-AI-Contract-Review-Platform ; https://www.artificiallawyer.com/2026/06/30/scott-stevenson-interview-spellbook-acm/. Company size: ~115 employees (stated plan to grow to ~230), Series B, $50M raised (led by Khosla Ventures), founded 2018. Actively building AI agents: yes — legal AI agents for contract review plus newly launched autonomous contract management (agentic workflows). Pain points: reliability/accuracy of legal agents on real contract data (Stevenson's public thesis: legal AI needs real data); scaling agentic workflows to production for high-stakes legal use. Challenges: trust, accuracy, and cost of autonomous agents in regulated legal workflows. Must-haves: accurate, reliable agents grounded in domain data; production-grade agent reliability. Nice-to-haves: cost/observability controls over agent runs. ICP confidence: Medium (technical co-founder/CEO, Series B, ~115 employees, agent features in production). Note: exact LinkedIn profile URL not confirmed in this run (search "Scott Stevenson Spellbook"); pain points inferred from public interviews, not a single verbatim quote.
Scott White — Director, Software Architecture (building AI agents), Hippocratic AI
Signal: 4 (ICP Director-level architect building AI agents at an AI-native company; found via LinkedIn people search "Hippocratic AI engineering leader agents" this run).
Source: https://www.linkedin.com/search/results/people/?keywords=Hippocratic%20AI%20engineering%20leader%20agents — profile summary: "Software Architect at Hippocratic AI since April 2025, building AI agents after Director, Software Architecture at Aescape and software architecture leadership at TONAL." Headline: "Director, Software Architecture". SF Bay Area.
Company size: ~312 employees (Tracxn, May 2026). Hippocratic AI = AI-native healthcare company, Series C ($126M at $3.5B, Nov 2025; $404M total), scaling patient-facing generative AI agents (Polaris voice-agent platform) in production.
Pain points: NO public pain post captured this run — do not treat as expressed. Qualification is role + company based.
Challenges (inferred from role scope, flagged as inference): agent architecture ownership at a company scaling from a handful of agent types to a large catalogue of clinical agent roles — the classic 1→N agent scaling wall.
Must-haves: unverified this run.
Nice-to-haves: unverified this run.
ICP confidence: Medium — Director title and agent-building scope both fit, and the company qualifies cleanly on size/stage/agent-shipping; downgraded from High because an architecture role may not own budget or platform decisions. Second Hippocratic contact added this run alongside Markel Sanz Ausin; Omid Nejati, Sri Subramaniam, Vivek Muppalla, Saad Godil and others are already in the People Library.
Signal: 4 (ICP CTO at agent-native compliance company). Found via web research + verification — LinkedIn content/people search UNAVAILABLE this run (session not authenticated / login wall). Source: https://www.norm.ai/ ; https://www.law.com/legaltechnews/2026/07/07/ (Norm Ai $120M Series C) ; https://www.linkedin.com/in/scott-c-worland/. Company size: ~205 employees (June 2026); Series C, $120M raised July 2026 at ~$1.2B valuation (Khosla-led); $267M total. Actively shipping agents: Norm Ai builds the first Regulatory AI Agent platform — transforms regulations into autonomous compliance agents that make determinations, run multi-step tasks, and validate outcomes; supervisory agent infrastructure for regulated enterprise AI (Fortune 100 CCOs). Scott Worland is CTO (promoted from Head of Engineering; PhD Vanderbilt; 12+ yrs ML/AI; ex-Thyme Care, FreightWaves, Cornell). Pain points (INFERRED): reliability/accuracy of autonomous compliance decisions; supervising/governing agent behavior in regulated workflows; cost/observability as agent volume scales. Challenges: auditable, correct multi-step compliance agents at enterprise scale. Must-haves: reliability, governance/guardrails, per-run visibility. Nice-to-haves: cost optimization. ICP confidence: High (CTO, technical, 205-emp Series C agent-native company in band).
Signal: 1 (technical decision-maker speaking publicly about agent reliability & cost in production). Source: https://techcrunch.com/2026/05/29/cognitions-scott-wu-says-ai-coding-agents-shouldnt-replace-humans/ and https://colossus.com/article/scott-wu-tapes-cognition/. Company size: ~200-500 (grew after Windsurf acquisition); Series C, ~$26B valuation May 2026. Pain points: reliability of autonomous coding agents at scale ("a decent fraction of all Devin sessions now run automatically rather than by humans"), agents running on loops looking for anomalies then self-fixing, keeping agent output at junior-to-mid engineer quality. Challenges: cost of running huge volumes of autonomous agent sessions, doubling usage every ~8 weeks. Must-haves: visibility into what agents are doing and what each run costs, reliability guardrails for unattended runs. Nice-to-haves: better routing between cheap/strong models per task. ICP confidence: High (technical co-founder of a company shipping 5+ agents in production; textbook agent-operating-layer buyer).
Signal: 4 (ICP technical leader shipping agentic AI into regulated credit decisions). Source: https://www.zest.ai/company/leadership/ ; https://www.zest.ai/ ; https://www.cbinsights.com/company/zestfinance ; LinkedIn: https://www.linkedin.com/in/sean-kamkar/ . Company size: ~192 employees (CB Insights, 2026). AI lending / credit underwriting for banks and credit unions; $200M growth investment from Insight Partners.
Pain points: INFERRED from public positioning (not a verbatim quote). LuLu is described by Zest as the first lending intelligence platform using generative AND agentic AI; his remit is explicitly "scalable, TRANSPARENT models that drive performance, FAIRNESS and automation in credit underwriting". Transparency and fairness under ECOA/Reg B adverse-action rules mean every agent decision must be explainable and reconstructable - the control problem, not the cost problem.
Challenges: moving from scored models (deterministic, defensible) to agentic workflows (non-deterministic) in the single most adverse-action-regulated decision surface in US consumer finance. He owns both engineering and data science, so there is no split between the people who build agents and the people who must defend them.
Must-haves: deterministic replay and per-decision explainability; drift detection on agent behaviour; audit evidence for examiners.
Nice-to-haves: per-run cost visibility across credit-union tenants.
ICP confidence: MEDIUM-HIGH. Role and company are a clean fit and headcount is comfortably in band; downgraded from High because Zest's agentic surface (LuLu) is one platform rather than a confirmed 5+ agent fleet, and some directory sources still list his older SVP/Head of Data Science title - verify current title before outreach.
Sebastian Caceres — Engineering Director, Platform, Gorgias
Signal: 4 (technical leader at a company actively shipping AI agents). Source: https://theorg.com/org/gorgias/org-chart/sebastian-caceres | company AI context: https://temporal.io/resources/case-studies/gorgias-uses-ai-agents-to-improve-customer-service. Company size: ~450–525 employees (verified — Gorgias ~521 as of Mar 2026); Series C ($530M valuation, ~$72M ARR). Gorgias ships an AI Agent (Support Agent + Shopping Assistant) that automates 60%+ of support for 15,000+ ecommerce brands, orchestrated on Temporal. Pain points (inferred from Platform-director role at an agent-native CX company): keeping the AI Agent reliable in production across 15,000+ brands; controlling per-conversation LLM cost as agent volume scales; platform-level observability into what each agent run costs and why it fails. Challenges: scaling agent infra from a single automation to a full multi-tenant agent platform; latency/edge-case handling at contact-center scale. Must-haves: cost-per-run visibility, reliability/guardrails in production, multi-tenant governance. Nice-to-haves: model routing, automated regression/eval tooling. ICP confidence: Medium (title = Director+, company squarely in ICP on size/stage/agents; individual pain not directly quoted — inferred from role + company). METHOD NOTE: Chrome/LinkedIn was NOT connected this run; found via verified web research (org chart + company sources), consistent with prior runs. LinkedIn URL not directly verified.
Signal: 4 (ICP speaking publicly about running an agent product in production). Source: https://www.unite.ai/sebastian-enderlein-chief-technology-office-at-deepl-interview-series/ — title verified live on LinkedIn 2026-08-30 ("CTO", DeepL, Munich). Company size: ~900-1,500 (PitchBook 900; Tracxn 1,547 May 2026; RocketReach 1,468) — within the 50-2,000 band but near the ceiling on the higher estimates. Cologne, Germany; Series B ($300M, 2023). Pain points: DeepL ships DeepL Agent, an agent that plans workflows and drives office tools autonomously, with 2,000+ customers deploying agents. Enderlein on the core tension: "Our business customers rely on DeepL for critical high-stakes communication, so reliability, security and trust are really essential. At the same time… innovation can't slow down." Challenges: agent reliability at enterprise scale across thousands of deploying customers; he is on record as cost- and complexity-averse by disposition — "over-architecting is a serious issue, and 'over-platforming' can cripple your organization", which is both a buying signal (he will resist bloated platforms) and a risk (he will resist adding a layer). Must-haves: reliability guarantees for high-stakes autonomous workflows; anything added must be thin, not another platform. Nice-to-haves: per-customer agent cost attribution across 2,000+ deployments. ICP confidence: Medium — right title, geo, stage and headcount with a real agent product in market; but his public commentary is about platform reliability broadly, not agent token spend specifically, and his stated aversion to over-platforming is a live objection to plan for.
Signal: 4 (ICP applied-AI leader at a Series C company shipping agents). Source: LinkedIn people search "\"Head of Applied AI\" OR \"Director of Engineering\" AI agents production evals" -> profile https://www.linkedin.com/in/sergey-osu/ ("Head of Applied AI @ Akur8", New York City Metropolitan Area; company field confirms AKUR8). Corroborating: https://www.akur8.com/ and https://www.akur8.com/pricing/discover ; Tracxn/Crunchbase for headcount and funding. Company size: 234 employees as of 31 March 2026 (Tracxn); 200+ across eight offices per company site. Series C: $120M led by One Peak (Sept 2024), $186M+ total raised; acquired Matrisk Jan 2026. Actuarial AI platform for P&C/health insurance pricing and reserving; ships "transparent agents to automate the manual steps in actuarial workflows" and a Discover product that continuously ingests US regulatory rate filings. Pain points (INFERRED from role + product, no verbatim quote captured this run): making agents in a regulated actuarial workflow explainable and auditable, not just accurate; cost of continuous data ingestion plus agent runs across insurer customers. Challenges: reliability and transparency requirements of regulated insurance pricing constrain how autonomous agents can be; proving agent output to actuaries and regulators. Must-haves: explainability/traceability of every agent run, production reliability. Nice-to-haves: per-run cost visibility, eval tooling for agent regressions. ICP confidence: Medium (clear ICP title and a 234-emp Series C SaaS company that publicly ships agents; downgraded from High because no direct pain-signal quote was found this run).
Signal: 4 (ICP author shipping agents — LinkedIn post announcing new VP/GM Agentic AppSec role, building an autonomous security-agent platform for the agentic era). Source: https://www.linkedin.com/feed/update/urn:li:activity:7479842060699734016/. Company size: ~1,900 (LinkedIn band 1,001-5,000; ~1,910 associated members — near upper ICP bound). Pain points: coding agents now write a large share of production code at machine speed and introduce several times more vulnerabilities per line than humans; the PR-time-scan + human-triage + manual-fix security model cannot scale to the agentic era; attackers are going agentic in parallel. Challenges: building durable, autonomous security agents that run next to every engineering team, threat-modelling every change before merge and closing the loop autonomously with a human as auditor. Must-haves: an independent verification/reliability layer around agents; reliable autonomous agent execution at scale; security-tuned agent orchestration. Nice-to-haves: applied AI research + agent-orchestration talent. ICP confidence: Medium (VP-level technical AI leader actively building/shipping agents; company ~1,900 emp sits near the 2,000-employee cap and is later-stage than the Series A–C ICP band — flagged).
Sergey Ulasen — Senior Director of AI Development, Constructor (Constructor.io)
Signal: 1/4 (senior AI leader at an AI-native company shipping agents; identified via LinkedIn people search "Director of AI agents production", then verified). Source: https://www.linkedin.com/in/sergey-ulasen/ . Company size: ~619-818 employees (Revelio, 2026); AI-native ecommerce search/product-discovery platform now shipping agentic commerce/shopping agents; $61M+ raised. Pain points (inferred from role + company stage; no direct post/quote read this run): scaling agentic discovery across billions of queries/day makes per-run LLM cost and latency a first-order concern; hard to attribute cost per agent/query at that volume. Challenges: keeping agent reliability + relevance high while controlling token/compute spend at enterprise scale. Must-haves: per-agent/per-run cost visibility and guardrails; reliability/observability in production. Nice-to-haves: automated cost optimization / model routing across a large agent fleet. ICP confidence: Medium-High (Director-level AI leader; company size confirmed in-band; agent-shipping confirmed; pains inferred not quoted).
Seth Shelnutt — VP of Engineering, Coder
Signal: 4 (ICP leader at a company actively shipping AI agents). Found via LinkedIn company People-tab sweep of Coder.
Source: https://www.linkedin.com/company/coderhq/people/?keywords=engineering
Company size: 51-200 employees (per Coder LinkedIn company page, Austin TX). Coder = self-hosted cloud development environments, now positioned around running coding agents at scale inside enterprise environments.
Pain points: NONE DIRECTLY OBSERVED from Seth himself this run. BUT there is a real, verifiable company-level signal: a colleague's LinkedIn headline on the same page reads "AI Gateway at Coder | Senior Engineering Manager" (Marcin Tojek, https://www.linkedin.com/in/marcintojek) — i.e. Coder has staffed a named AI-gateway workstream, which is exactly the token-routing / spend-control problem area.
Challenges (inferred): fleets of coding agents inside customer environments generating unbounded token spend per developer; enterprises want per-team caps and attribution.
Must-haves (inferred): per-developer / per-team agent spend attribution and hard budget caps.
Nice-to-haves (inferred): model routing per task class; audit of which agent ran what.
ICP confidence: High. VP Engineering at a right-sized company with an explicit internal AI-gateway effort — the closest thing to an observed pain signal found this run. Strongest of the six.
Also at Coder, secondary contacts (not added): Marcin Tojek (Senior Eng Manager, AI Gateway), Sushant Patankar ("Engineering Leader @ Coder", level unclear).
Seungwoo Han — Co-Founder & Chief Technology Officer, Wrtn Technologies
Signal: 4 (ICP shipping agents). Source: https://en.wowtale.net/2026/08/27/234887/ ; https://www.bloomberg.com/news/articles/2026-08-25/korean-ai-startup-wrtn-raises-funds-at-870-million-value-to-fund-global-growth. Company size: 176 (Tracxn 2026). Seoul, South Korea. Series C $72-76M raised Aug 2026 at $870M valuation — funding is DAYS OLD as of this run, making this an unusually well-timed entry point. Agent evidence: Series C raised explicitly to build a "local autonomous agent" B2C platform; Wrtn Agent OS ships multiple production agents — an AI vibe-coding agent for backend generation, "Sindie" a visual-design-scoring agent, and a Studio Autonomous Agent framework for building further agents (per Wrtn public materials and the wrtnlabs GitHub org). Pain points: coordinating and scaling many concurrent product agents (coding, design review, studio agents) across a fast-growing consumer platform. Challenges: maintaining reliability and quality as agent count multiplies from a handful into a full "Agent OS" ecosystem; global expansion adds multi-region operating complexity. Must-haves: per-agent observability, reliability at consumer scale, cost control across many concurrent agents. Nice-to-haves: tooling that lets non-engineers spin up new agents safely (self-serve agent creation). ICP confidence: High — current CTO/co-founder title confirmed, headcount and funding stage both in band, strong multi-agent production evidence from primary sources. Korea is an under-covered geography in the People Library.
Signal: 1/4 (LinkedIn people search for ICP agentic-AI leaders; verified via web). Source: https://www.linkedin.com/search/results/people/?keywords=VP%20of%20AI%20agents | profile https://www.linkedin.com/in/seungwooson. Company size: ~218 employees, Series B ($111M total; investors incl. Charles Schwab, GV/Google Ventures, Citi Ventures). Agent activity: "Ester Intelligence" AI engine powers advisor workflows end-to-end, purpose-built for deterministic, auditable outputs in high-stakes advisory; processed 100k+ estate documents in 2025. Pain points (INFERRED from role/company, not a verbatim quote): reliability/determinism and auditability of agents in a regulated, high-stakes financial context; controlling agent behavior; cost/scale as usage grows. Challenges: making agentic outputs deterministic and auditable in regulated finance. Must-haves: behavior control/guardrails, auditability, reliability. Nice-to-haves: per-run cost visibility. ICP confidence: High (VP of Applied AI = senior technical AI leader at ~218-emp Series B company actively shipping agents).
Signal: 4 (shipping agents) + 3 (public engagement with agentic-commerce discourse). Source: https://justt.ai/chargebackx/ ; https://justt.ai/blog/agentic-commerce-chargeback-risk-preparation/. Company size: 139 (Tracxn, June 2026), Tel Aviv Israel. LIVE LINKEDIN VERIFIED this run — people search on "Justt Chief Technology Officer" returns "Shahar Tal — Chief Technology Officer (We're Hiring)", Israel, alongside Justt co-founder/CEO Ofir Tahor. Agent evidence: Justt's core product is AI-driven chargeback dispute automation; Tal spoke on a ChargebackX 2025 panel on agentic commerce. PARTIAL DIRECT QUOTES from event coverage: "People would just not be able to cope with it" (re: manual chargeback handling at agent-driven transaction volume) and "AI created the problem, but AI also gave us tools to fight it". Pain points: evidence/dispute-writing volume outstripping manual review capacity as agent-driven transactions scale. Challenges: distinguishing legitimate AI agent traffic from fraud/bots; scaling automated dispute generation reliably. Must-haves: reliable automated content generation, fast iteration on dispute-response quality. Nice-to-haves: cost tracking across high-volume automated writing runs. ICP confidence: Medium — title and headcount verified from multiple sources including live LinkedIn, but the captured evidence leans toward market commentary on agentic commerce rather than an explicit statement about Justt's own internal agent fleet size.
Signal: 4 (ICP at a company shipping a named AI Agents product line that acts in third-party systems). NOTE: LinkedIn/Chrome unavailable this run — LinkedIn URL surfaced in search results, not opened/verified.
Source: https://duve.com/press-room/duve-acquires-easyway/ (names + quotes him as "Co-Founder and CTO at Duve") ; https://duve.com/duve-ai/
Company size: ~111 across 3 continents (aggregator data as of April 2026). Series B — $60M led by Susquehanna Growth Equity (Dec 2025), $85M total. Travel/hospitality — hotel & vacation-rental guest experience.
Agents in production: CONFIRMED. Ships "AI Agents" as a named product line — hospitality-specific agents trained on hotel data (reservations, room types, amenities, guest history, property workflows) that connect directly to PMS, task-management and F&B systems. These take ACTIONS in third-party systems, not just chat. 150+ PMS/OTA/PSP integration partners. SOC 2 Type II, ISO 27001, ISO 27701.
Pain points: [evidenced] His sourced quote on acquiring Easyway (a generative-AI hotel messaging company) frames value in per-interaction terms: the solution "has already proven itself across countless guest interactions and touchpoints, enhancing guest satisfaction and helping hoteliers save thousands of man-hours." Volume of interactions is the unit of value — which means per-interaction economics matter structurally. [inferred] Multilingual agent reliability (100+ languages), cross-channel context retention, cost per guest interaction at hotel-chain scale.
Challenges: Agents acting across 150+ heterogeneous PMS/OTA integrations (huge surface for silent failure); multilingual reliability; cost per guest interaction scaling with occupancy, not with revenue.
Must-haves: Per-interaction cost visibility; reliability across a very wide integration surface; context retention across channels (chat/email/SMS) and across a guest stay.
Nice-to-haves: Language-level quality benchmarking; agent action audit trail for hotel-brand compliance.
ICP confidence: Medium-High — CTO/co-founder, Series B, ~111 employees, genuine production agents taking actions in travel/hospitality. Downgraded from High only because his sourced quote PREDATES the current agent product and no technical commentary on operating agents was found.
Signal: 4 (ICP senior technical leader at a company shipping AI agents). Found via LinkedIn people search; Gong ships AI agents ("Gong AI", autonomous prospecting/forecasting agents). Note: Gong's co-founder Eilon Reshef was already in the brain — this is a distinct, newly added engineering leader.
Source: https://www.linkedin.com/search/results/people/?keywords=Gong%20VP%20Engineering%20AI%20agents
Company size: ~1,200-1,400 employees (revenue-intelligence SaaS, late-stage). Within ICP 50-2,000.
Pain points (inferred from role/company context, not a direct quote): scaling many AI/agent features across a large enterprise customer base; LLM inference cost at scale; reliability and accuracy of agent outputs on customer data.
Challenges: controlling aggregate LLM spend across many agentic features; observability of cost per account/agent.
Must-haves: per-agent cost visibility; production reliability/monitoring.
Nice-to-haves: routing/caching cost optimization; spend anomaly detection.
ICP confidence: High (VP Engineering at a mid-size company actively shipping AI agents).
Shailesh P. — Director of Engineering, Yellow.ai
Signal: 4 (ICP senior technical leader at an agent-native company; discovered via LinkedIn people-search of companies actively shipping AI agents) | Source: https://www.linkedin.com/in/shailesh-p-4017101a | Headline: Director of Engineering @ Yellow.ai (enterprise conversational + agentic AI for CX) — Claude AI, Ontology Engineering. Note: surname abbreviated to 'P.' on profile. | Company: Yellow.ai — actively building/shipping AI agents in production | Company size: ~700–1,000 (estimate) | Pain points (INFERRED from role/company context — not a verbatim quote observed this run): token/cost efficiency across high message volume; reliability and observability across many customer-facing agents | Challenges: keeping per-conversation cost down while maintaining reliability at enterprise CX scale | Must-haves: per-run cost visibility; reliability guardrails; observability | Nice-to-haves: context/token-waste reduction; benchmarking across customer agents | ICP confidence: Medium-High (senior technical/eng leadership at a 50–2,000-employee company shipping AI agents)
Shanil Puri — Director, Speech Technologies, Hippocratic AI
Signal: 4 (ICP technical AI leader at validated target company Hippocratic AI; sourced via LinkedIn People search, run 2026-08-01). Source: https://www.linkedin.com/in/shanilpuri/ . Company size: ~300-500 (Hippocratic AI — safety-focused healthcare voice agents, well-funded; validated in-range). Pain points (role/company-contextual, not from an individual post this run): reliability and safety of real-time healthcare voice agents in production; latency and compute cost of speech + LLM pipelines; hallucination control in a regulated clinical setting. Challenges: deterministic reliability for patient-facing agents; cost of always-on voice inference. Must-haves: production reliability/eval, cost-per-call visibility. Nice-to-haves: model routing, cheaper speech inference. ICP confidence: High (Director of Speech/AI technologies at 50-2,000-emp agent-native company).
Signal: 1 (Head of AI/ML writing that 'matching query complexity to model cost is where efficiency meets quality in production AI agents'). Source: https://www.linkedin.com/in/shantanuladhwe/. Company size: ~300 employees, Series B (Berlin talent-acquisition platform automating sourcing/screening/matching/interview workflows with AI agents). Pain points: matching query complexity to model cost (efficiency vs quality); running AI agents cost-effectively in production; agent output quality. Challenges: balancing model cost vs quality at scale; deploying AI agents reliably in production. Must-haves: cost-aware model routing, production agent reliability. Nice-to-haves: MLOps tooling for agent pipelines. ICP confidence: High (Head of AI at right-sized Series B company actively shipping agents; direct cost-efficiency signal).
Signal: 3 (ICP engaging with competitor content — named reference customer in a Maxim AI case study; Maxim is a direct agent-observability/eval competitor).
Source: https://www.getmaxim.ai/blog/scaling-enterprise-support-atomicworks-journey-to-seamless-ai-quality-with-maxim/
Company size: ~120–150 (ZoomInfo ~150, May 2026; PitchBook ~120). Series A/B AI-native enterprise service-management company shipping agentic support/ITSM agents.
Pain points (direct quotes): "Maxim's tracing and metadata filtering capabilities let us pinpoint issues instantly instead of spending hours searching through scattered logs. We can now confidently scale our AI features, knowing we have complete observability from prompts to final outputs." Also: "The ability to curate high-quality datasets from execution traces and integrate AI observability directly into both development and production environments has significantly streamlined our operations."
Challenges: Debugging agent failures across scattered logs was blocking their ability to scale the number of AI features/agents; needed the same observability in dev and prod.
Must-haves: End-to-end tracing prompt→output with metadata filtering; ability to scale agent count without losing debuggability.
Nice-to-haves: Curating eval datasets directly from production execution traces.
ICP confidence: High — head-of-platform-level AI owner, company solidly in band, explicitly articulating the "we couldn't scale agents until we could see them" pain and already spending in the category.
NOTE: Atomicwork already has one contact in the library (Jeegar Shah, Head of Applied AI & Platform Engineering) — this is a second, distinct buyer-side contact at the same account. Coordinate before any outreach.
Shaosu Liu — Co-Founder & Chief Technology Officer, Loop
Signal: 4 (ICP technical co-founder at a company shipping agents into production).
Source: https://techcrunch.com/2026/04/17/loop-raises-95m-to-build-supply-chain-ai-that-predicts-disruptions/ ; https://www.loop.com/company/about-us ; https://theorg.com/org/loop-us/org-chart/shaosu-liu
LinkedIn: https://www.linkedin.com/in/shaosu/
Company size: 251-500 employees; +9.2% in 6 months, +102.4% over 2 years (headcount trackers). Series C $95M (Apr 2026), Valor Equity / Valor Atreides AI Fund, with 8VC, Founders Fund, Index, JPM Growth Equity Partners.
Agent evidence: supply-chain AI that ingests freight/logistics documents and predicts disruptions; AI-native product, doubling headcount, fresh $95M explicitly earmarked for AI build-out.
Pain points (INFERRED from public product/positioning, not quoted): document-heavy, high-volume ingestion over freight invoices and shipment data = very large token throughput per run; disruption prediction across a network-of-networks means long tool chains where errors compound.
Challenges: ex-Uber/Uber Freight, so he has scaled network systems before — the new problem is that per-transaction unit economics in freight audit are thin, which makes per-run LLM cost directly margin-relevant.
Must-haves: per-run and per-customer cost attribution (their pricing is transaction-adjacent); reliability on long chains.
Nice-to-haves: cross-agent observability as they hire aggressively post-raise.
ICP confidence: HIGH on role/size/stage/agent-activity. MEDIUM on pain specifics — no verbatim pain quote from him found this run. Best next step: read his own LinkedIn/engineering posts before outreach.
Sharath Veldanda — VP of Engineering, Bolster
Signal: 1/4 (LinkedIn people search for ICP agentic-AI leaders; verified via web). Source: https://www.linkedin.com/search/results/people/?keywords=VP%20Engineering%20agentic%20AI | profile https://www.linkedin.com/in/veldanda. Company size: ~93 employees, Series B ($40M total). Company: AI-powered digital risk protection / cybersecurity (phishing, brand abuse, domain & impersonation defense). Agent activity: NOTE — company profile describes AI-driven detection; agent-building is inferred from the VP's own LinkedIn headline focus ("AI Agents, Cybersecurity") plus the 2026 AI-SOC-agent market trend; not fully confirmed at company level. Pain points (INFERRED from role/headline): reliability of autonomous security agents; cost/scale of AI-driven detection running continuously. Challenges: trustworthy autonomous action in a security context. Must-haves: reliability guardrails, cost control. Nice-to-haves: observability. ICP confidence: Medium (right-size ~93-emp Series B; ICP title; agent-native activity inferred, not company-confirmed).
Signal: Bucket 1 (ICP CTO at agent company) — web research pivot (LinkedIn/Chrome unavailable). Source: https://pulse2.com/aera-technology-profile-fred-laluyaux-interview/ ; https://www.aeratechnology.com/news/aera-technology-empowers-enterprises-with-generative-ai-to-advance-decision-automation ; https://www.citybiz.co/article/848471/. Company size: 400+ employees ("Aeranauts"); Accenture-backed. Aera builds "Aera Decision Cloud" — agentic decision intelligence where AI agents autonomously make supply chain, finance and procurement decisions with human oversight, continuously learning from operational outcomes. Pain points: reliability/trust of autonomous decision agents in production; cost and governance of agents operating at enterprise machine speed; visibility into agent decisions. Challenges: scaling agentic decision automation across supply chain/finance while keeping guardrails and auditability. Must-haves: per-run/per-decision cost and reliability visibility, guardrails for autonomous action. Nice-to-haves: cross-domain agent observability. ICP confidence: High — Co-Founder & CTO at a 400-person company actively shipping decision agents.
Signal: 4 (ICP CTO at an agent-shipping company). Source: https://siliconangle.com/2026/05/20/wisdomais-new-analytics-agents-go-beyond-insights-automating-business-work-autonomous-action/ ; https://finance.yahoo.com/news/ai-data-startup-wisdomai-raised-160000787.html ; https://tracxn.com/d/companies/wisdomai | Company size: ~107 employees (Tracxn, Apr 2026); $50M Series A led by Kleiner Perkins + NVIDIA (Nov 2025), ~$73M total; founded 2023 (San Mateo). WisdomAI = Agentic Data Insights Platform; autonomous analytics agents that reason across distributed enterprise data and now take autonomous action (beyond dashboards). Co-founders: Soham Mazumdar (CEO, ex-Rubrik co-founder), Sharvanath Pathak (CTO), Kapil Chhabra (CPO), Guilherme Menezes. | Pain points: (inferred) reliability/trust of autonomous analytics agents acting on enterprise data; token/cost blowout as agents reason across large distributed data; per-run cost/observability. | Challenges: scaling many data agents reliably; controlling context/token waste over big data; production reliability + governance. | Must-haves: per-run cost visibility, context/token control, production reliability. | Nice-to-haves: agent eval + iteration; guardrails on autonomous actions. | ICP confidence: High — CTO at a ~107-emp Series A company shipping autonomous agents in production; core ICP persona.
Shawn Wen — CTO & Co-founder, PolyAI
Signal: Bucket 4-adjacent (ICP technical leader at an agent-native company). Sourced via ICP-title/company LinkedIn search after Signal 1-4 engagement searches returned mostly consultants, sub-Director engineers, or people already in Alpha Brain. Source: https://www.linkedin.com/in/shawn-wen-51b7b958/ (verified via LinkedIn people search "PolyAI co-founder CTO"). Company size: ~250-370 employees (web-verified; PolyAI raised Series D / $86M Dec 2025, >$200M total, NVIDIA NVentures-backed — note: stage is past the A-C band, otherwise strong fit). Agent-native: builds enterprise voice AI agents ("agentic-native platform for conversational assistants") deployed in production for large call centers. Pain points (inferred from role/company; no public post captured this run): per-call/per-run LLM+voice cost at high call volume, reliability of autonomous voice agents in production, latency/cost tradeoffs. Challenges: scaling many concurrent production voice agents cost-effectively; observability of cost per conversation. Must-haves: cost visibility per run, reliability guardrails. Nice-to-haves: provider portability. ICP confidence: Medium (role=CTO ✓, size ✓, agent-native ✓; funding stage Series D is beyond the A-C band; no direct engagement signal observed).
Shay Levi — Co-founder & CEO (ex-CTO of Noname Security; technical), Unframe (Unframe AI)
Signal: 4 (technical co-founder/CEO at a company deploying enterprise AI agents/solutions into production; web research this run). Source: https://www.calcalistech.com/ctechnews/article/sjzf1rfyzl ; https://pulse2.com/unframe-profile-shay-levi-interview/ ; https://www.trendingtopics.eu/unframe-enterprise-ai-startup-raises-50m-and-hits-100m-tcv-in-12-months/ . Company size: ~130 employees (~70 in Israel, rest California + Berlin; founded 2024; Series B $50M led by Highland Europe; $100M+ multi-year enterprise contracts). Unframe is an all-in-one enterprise AI platform that turns business needs into operational AI solutions/agents in days, focused on moving enterprises from pilots to full-scale production. Pain points (inferred + from public interview framing; no verbatim ICP quote captured this run): enterprises struggle to extract real value from AI and to move agents from pilot to production reliably; governance, reliability and deployment speed at scale. Challenges: production reliability/governance across many enterprise agent deployments; unit economics at scale. Must-haves: reliability, governance/control, fast production deployment. Nice-to-haves: cost/observability per deployment. ICP confidence: High (technical co-founder/CEO — was co-founder & CTO of Noname Security, acq. by Akamai ~$500M — at a ~130-emp Series B company shipping enterprise agents in production). Profile URL not captured — do not fabricate.
Shay Perera — Co-Founder & CTO, Navina
Signal: 4 (technical co-founder building/shipping agents) + 1 (podcast guest "AI as a Clinical Co-Pilot"). Source: https://theorg.com/org/navina/org-chart/shay-perera ; https://www.crunchbase.com/person/shay-perera ; https://www.calcalistech.com/ctechnews/article/ry00l6xla1e (Series C). Company size: 201-500 employees; Series C $55M (Mar 2025, led by Goldman Sachs), $100M total; ex-IDF Intelligence Corps AI lab. Navina ships an AI clinical copilot used by 10,000+ healthcare professionals across 1,300 clinics; moving toward agentic workflows over patient data. Pain points: reliability/accuracy of clinical AI at scale; scaling agent deployment across many clinics/EHRs; cost of running AI across large clinician base. Challenges: correctness in value-based-care clinical intelligence; heterogeneous data + integration; trust/reliability with clinicians. Must-haves: reliable, accurate agent output; scalable deployment. Nice-to-haves: per-agent cost/observability. ICP confidence: High — CTO/co-founder, 201-500 emp Series C, clear agent product in production at scale.
Signal: 4 (ICP senior technical leader at a company shipping AI agents). Found via LinkedIn people search; Gong ships AI agents. Distinct from co-founder Eilon Reshef (already in brain).
Source: https://www.linkedin.com/search/results/people/?keywords=Gong%20VP%20Engineering%20AI%20agents
Company size: ~1,200-1,400 employees. Within ICP 50-2,000.
Pain points (inferred from role/company context, not a direct quote): R&D leadership over agentic features shipping to enterprise customers; balancing model quality against inference cost; production reliability of agents.
Challenges: managing LLM/agent cost as feature surface grows; visibility into cost per run.
Must-haves: cost-per-run observability; reliability guardrails.
Nice-to-haves: automated model/prompt optimization; cost benchmarking.
ICP confidence: High (VP R&D at a mid-size company actively shipping AI agents).
Sherwin Yu — Head of AI and Product Engineering, Gamma
Signal: 4 (ICP writing/speaking about building agents; found via Vercel customer story, LinkedIn profile verified live 2026-09-02 — headline "Building @ Gamma || Engineering Leader and Zen practitioner || prev Benchling, Asana"; search result subtitle reads "AI + Product Engineering at Gamma").
Source: https://vercel.com/customers/gamma-builds-design-first-agents-with-vercel
Company size: ~50-100 employees; Vercel/TechCrunch report a team of roughly 20 engineers. Gamma = AI-native presentation/deck generation platform shipping "Gamma Agent", a multi-step conversational agent. Series A ($12M), later reported $2.1B valuation.
Pain points: Moving from single request-response prompting to multi-step, multi-agent orchestration exposed context and conversation-state as the binding constraint. Per case study: they "needed finer control and more persistence over conversation state... the ability to pass context from one agent to another, manage message history across sessions, and orchestrate more complex multi-step interactions than a simple request-response loop." Also quoted: "Right now, context is what separates a useful agent from a generic chat bot."
Challenges: Handing context between agents without losing fidelity; persisting message history across sessions; per-model configuration and cost tracking as agent count grows; small eng team (~20) carrying agent infra alongside product.
Must-haves: Durable cross-session agent state, agent-to-agent context passing, multi-step orchestration control, visibility into which model/config each step used.
Nice-to-haves: Cost-per-model and cost-per-agent-step reporting; regression checks on agent behaviour after prompt/model changes.
ICP confidence: High — exact Director+/Head-of title, AI-native company in band, and a textbook "1 agent -> many agents, hit the context wall" signal from a dated primary source. Headcount sits at the ICP floor, worth confirming before a heavy-touch sequence.
Shivam Khandelwal — Director of Engineering, Level AI
Signal: 4 (ICP engineering leader at an agent-shipping company; found via company-scoped LinkedIn people search "Level AI engineering").
Source: https://www.linkedin.com/search/results/people/?keywords=%22Level%20AI%22%20engineering (profile https://www.linkedin.com/in/shivamkhn/)
Company size: Level AI ~200-400 employees (Series C; contact-center AI + agent-assist/AutoQA platform shipping conversational & voice agents in production).
Pain points [INFERRED from role + company context, NOT a verbatim post]: scaling multi-agent CX/voice deployments across high call volumes; per-run/per-agent cost visibility; reliability & QA of agent outputs in production.
Challenges: holding reliability + cost-per-run steady as agent count and call volume scale across enterprise contact centers.
Must-haves: per-agent/per-run cost observability; production reliability guardrails.
Nice-to-haves: routing/model optimization to reduce token spend.
ICP confidence: High — Director of Engineering at an independent, 50-2,000-employee agent-native company; net-new (brain previously had only Level AI's CEO).
Shobhit Agrawal — SVP — Agentic AI Deployment, Netomi
Signal: 4 (senior technical leader at agent company; LinkedIn people search "Netomi VP engineering AI"). Source: https://www.linkedin.com/in/shobhitagrawal1/ . LinkedIn headline: "SVP - Agentic AI Deployment at Netomi | Enterprise Agentic AI for Customer Experience." Company size: ~100-250 employees (Netomi; enterprise CX AI agents; ICP company already in brain). Pain points (INFERRED, NOT verbatim): deploying enterprise agentic AI reliably for CX; cost-per-resolution economics; scaling agent deployments across many enterprise customers. Must-haves: per-deployment cost + reliability visibility. ICP confidence: HIGH — SVP Agentic AI Deployment, senior technical leader, agent-native, size fits. Caveat: pain inferred.
Shomron Jacob — Head of Applied Machine Learning & Platform, Iterate.ai
Signal: 1 (senior applied-AI leader at an agent-platform company; surfaced via LinkedIn people search "Head of Applied AI agents production").
Source: https://www.linkedin.com/search/results/people/?keywords=Head%20of%20Applied%20AI%20agents%20production
Company size: ~64 employees (some sources ~100); VC-backed ($6.4M); builds Interplay, a low-code agentic AI platform for multi-agent workflows. In-band (50-2,000, AI-native, shipping agents).
Pain points (inferred from role/company context; no direct post captured this run): running/scaling many enterprise multi-agent workflows, cost & latency of multi-agent runs, private/edge deployment.
Challenges: enterprise-grade reliability & observability across a growing fleet of agents.
Must-haves: cost visibility per agent/run, production reliability.
Nice-to-haves: framework-agnostic instrumentation, governance.
ICP confidence: High — Head of Applied ML & Platform (ICP title), AI-native company shipping an agent platform, in-band size.
Signal: 4 (ICP technical co-founder building/shipping agents in production — "agent swarm" for sellers). Source: https://sequoiacap.com/article/partnering-with-rox-every-seller-needs-an-agent-swarm/ ; https://techcrunch.com/2026/03/12/sales-automation-startup-rox-ai-hits-1-2b-valuation-sources-say/ ; https://tracxn.com/d/companies/rox . Company size: 130 employees (Apr 2026); ~$50M+ raised (Seed led by Sequoia, Series A led by General Catalyst, GV); $1.2B valuation (Mar 2026), unicorn in <2 yrs. AI-native GTM/sales platform running a multi-agent "agent swarm" per seller (prospecting, research, engagement) — CRM-replacement ambition. Sridharan is technical co-founder & CTO — previously led data-infrastructure at Confluent (making Kafka "faster and cheaper across clouds") and was a pre-launch engineer on Amazon Aurora, so directly steeped in cost/perf-at-scale engineering. Pain points: running many concurrent agents per user reliably; cost/quality economics of an agent swarm at scale; visibility into what each agent does and costs. Challenges: reliability + cost control across large multi-agent volume; per-run/per-seller cost attribution. Must-haves: reliability at scale, agent observability, cost efficiency. Nice-to-haves: model routing, per-run cost visibility. ICP confidence: High (technical co-founder/CTO at a 130-person AI-native company whose product is production sales agents; cost-engineering pedigree makes Alpha's cost+control wedge especially resonant). Profile URL left blank (personal LinkedIn not verified — not fabricating).
Shrivu Shankar — VP, AI, Abnormal AI
Signal: Signal 4 (LinkedIn people search at named agent company; VP-level AI leader. Sourced by role/company match)
Source: https://www.linkedin.com/search/results/people/?keywords=%22Abnormal%20Security%22%20director%20engineering%20AI%20agents
Profile: https://www.linkedin.com/in/shrivushankar/
Company size: ~1,200-1,800 (Abnormal AI / Abnormal Security; autonomous AI security agents / AI SOC analyst; Series D)
Pain points: (inferred, not directly observed) Reliability and guardrails for security agents running autonomously in production; cost-per-run at high volume
Challenges: (inferred) Preventing agent error/hallucination in safety-critical security workflows; observability
Must-haves: (inferred) Per-run cost + reliability visibility; production-grade guardrails
Nice-to-haves: (inferred) Agent-level governance and audit
ICP confidence: High (VP of AI = exact ICP; company is an in-band agent-native scale-up)
Shubham Agarwal — VP Engineering, Leena AI
Signal: 4 — ICP leader at a company actively shipping agents; amplifies his company's agent launches.
Source: https://www.linkedin.com/in/shubham-agarwal-b7021a57/recent-activity/all/ (found via https://www.linkedin.com/company/leena-ai/people/?keywords=vp%20engineering)
Company size: 201-500 employees (LinkedIn company page, verified this run). Leena AI = AI "Colleagues"/agents for enterprise HR+IT, Series B, ~$20M+ ARR.
Observed this run: reposted Adit Jain (Co-Founder & CEO) launching Leena AI's Group Chat Agent for Slack — agent reads threads, answers in-thread with sources, auto-routes personal answers to DM. LinkedIn headline: "VP Engineering at Leena AI (Hiring for multiple roles)".
Pain points: NOT directly expressed in his own words this run. Context-inferred only — running multiple production "AI Colleague" agents inside enterprise Slack/HRIS, which implies multi-agent orchestration, source grounding and per-conversation cost across a large enterprise install base.
Challenges (inferred): scaling an eng org while multiple agents run in customer production environments; hiring across multiple roles.
Must-haves (inferred): reliability + source-grounded answers in enterprise deployments.
Nice-to-haves (inferred): per-agent cost visibility as agent count grows.
ICP confidence: High — VP Engineering (core ICP title) at a 201-500 employee Series B AI-native company with 5+ agents in production. Confidence is on ICP fit; pain signal strength is only Medium (no first-person cost/reliability post found).
Shyam Rajagopalan — Co-Founder & CTO, Infinitus Systems
Signal: 4 (ICP technical co-founder at agent-native company shipping agents in production). Source: https://datainnovation.org/2025/02/5-qs-for-ankit-jain-founder-of-infinitus-systems/ ; https://www.prnewswire.com/news-releases/infinitus-launches-first-trusted-voice-ai-agents-for-healthcare-302436349.html ; https://tracxn.com/d/companies/infinitus/ . Company size: ~221 employees (Tracxn, Feb 2026; PitchBook ~186). Series C-stage, ~$51.4M raised (Kleiner Perkins, Coatue, GV). Actively building/shipping AI agents: Infinitus runs "trusted" voice AI agents for healthcare that have powered 100M+ minutes of healthcare phone conversations (benefit verification, prior auth, back-office calls), serving 44% of the Fortune 50 and 125,000+ providers — a fleet of production voice agents at very high call volume. Pain points (inferred from product domain + public statements, NOT a verbatim quote): reliability/trust of autonomous voice agents in a regulated, high-stakes healthcare context; latency and cost-per-minute economics at 100M+ minutes of scale; observability/auditability into what each agent said and did on a call. Challenges: keeping agents accurate and compliant while scaling call volume; guardrails/human oversight on risky steps. Must-haves: reliability, trust/auditability, cost efficiency per agent run at scale. Nice-to-haves: per-agent cost visibility and performance analytics. ICP confidence: High (co-founder & CTO; ~221 employees in band; large production agent fleet).
Sid Pardeshi — Co-Founder & CTO, Blitzy
Signal: 4 (ICP technical leader at a company actively building/shipping AI agents; surfaced via 2026 agent-startup funding research + LinkedIn verification — LinkedIn post/comment search this run returned mostly consultants/recruiters, so pivoted to named-company research).
Source: https://www.linkedin.com/in/sid-pardeshi/ ; company context https://en.wikipedia.org/wiki/Blitzy , https://www.bostonglobe.com/2026/05/05/business/blitzy-ai-software-cambridge-unicorn-startup/
Company size: ~80–135 employees (Cambridge, MA); Series A $200M, ~$1.4B valuation; AI-native autonomous software-dev platform that orchestrates thousands of parallel coding agents.
Pain points (inferred from role/company, not quoted): running thousands of agents in parallel makes per-run cost and token consumption a first-order constraint; reliability/validation of agent-generated code at scale; controlling compounding spend as agent count grows.
Challenges: keeping multi-agent orchestration reliable and cost-predictable across large legacy codebases; observability into what each agent run costs.
Must-haves: cost-per-run visibility and control across massive parallel agent fleets; guardrails/validation for production reliability.
Nice-to-haves: token-waste/context optimization; unified spend + reliability dashboards.
ICP confidence: High — Co-Founder & CTO (ex-NVIDIA, 27 patents), agent-native company squarely in 50–2,000 range, actively shipping agents in production.
Sigge Labor — Co-Founder & President (previously CTO), Legora
Signal: 4 (technical co-founder at an AI-native company shipping an agentic operating system). Source: https://se.linkedin.com/in/siggelabor ; https://theorg.com/org/legora/org-chart/sigge-labor ; https://www.artificiallawyer.com/2026/05/07/legora-launches-aos-agentic-operating-system/ ; https://www.law.com/legaltechnews/2026/05/07/legora-launches-agentic-ai-legal-operating-system-legora-aos/ ; https://www.cnbc.com/2026/05/19/legora-cnbc-disruptor-50-ranking.html. Company size: ~400-500 employees (grew 40 -> 400 in under 12 months; ~500 reported 2026). Founded 2023 (Stockholm), ~$866M raised, 1,000+ organisations across 50+ markets. Product: Legora aOS agentic operating system - the 'Legora Agent' autonomously runs research, drafting and busy work end-to-end from matter intake to client delivery, continuously, for review at scale. Pain points (inferred from public product positioning, not verbatim): continuously-running autonomous agents across 1,000+ customer orgs = an inference bill and a reliability surface that both scale with usage, not seats; 'review it at scale' implies a human-oversight/queueing burden. Challenges: hyper-growth (40->400+ staff in a year) while keeping an always-on agent platform reliable and economically sane; 50+ markets and 300% NRR means load compounds fast. Must-haves: per-run cost visibility and control for always-on agents; reliability and auditability on legal work product. Nice-to-haves: agents that compound/improve from reviewed output. ICP confidence: Medium-High - technical co-founder, ~400-500 employees (in band), unambiguously shipping production agents at scale. Caveat: his LinkedIn headline now reads President (title in transition from CTO), so confirm current technical ownership before outreach. Also note very large recent raise = may build in-house. NOTE: not yet contacted; research only.
Sigge Labor — Co-founder & CTO, Legora
Signal: 4 (ICP technical leader at company shifting core product to agent-run workflows). Source: https://www.businesswire.com/news/home/20250521743643/en/ , https://www.siliconrepublic.com/start-ups/legora-legal-ai-funding-series-d-sweden , https://tracxn.com/d/companies/legora . Company size: ~517 employees (May 2026), grew 40→400+ in a year across Stockholm/London/NY/Denver/Sydney/Bengaluru. Stage: Series D (~$550M round, $5.55B valuation) — later stage than core ICP but headcount well within 50–2,000. Pain points: 'work is quickly shifting to end-to-end workflows run by agents' — moving from copilot to multi-agent autonomous legal workflows at scale across top global law firms; reliability/accuracy of agent output in high-stakes legal work. Challenges: scaling many agents reliably across a fast-growing global customer base; controlling agent behavior and quality; cost of high-volume LLM calls on document-heavy legal tasks. Must-haves: reliability + control of agents at scale, observability. Nice-to-haves: per-run cost attribution, model routing. ICP confidence: Medium — strong operational fit (517-person AI-native company building production agents) but stage (Series D) is beyond the Series A–C target. (Note: profileUrl is company page.)
Sigurjón Ísaksson — Chief Technology Officer (formerly Head of AI), Definely
Signal: Bucket 4 (ICP technical leader at an agent-active company; also engages the legal-AI-agent competitive space).
Source: https://www.definely.com/blogs/definely-appoints-sigurjon-isaksson-as-chief-technology-officer ; https://www.definely.com/newsroom/definely-launches-agentic-ai-system-to-supercharge-both-accuracy-and-speed-for-lawyers
Company size: Definely ~101-200 emp (84 in Jun 2025, 100+ H2 2025), Series B $30M (Revaia, Jun 2025), London legaltech; launched an agentic AI system ('Enhance') for legal drafting/review — entered the legal-AI-agent race.
Pain points (INFERRED, not verbatim): scaling legal AI agents with high accuracy/reliability for lawyers; hiring/scaling AI engineering; accuracy-vs-cost tradeoff in agentic contract review.
Challenges: reliability & accuracy in regulated legal work; cost per agent run; scaling the technical team.
Must-haves: accuracy/reliability guardrails, cost visibility.
Nice-to-haves: model comparison, agent observability.
ICP confidence: Medium-High — CTO, Series B (clean A-C fit), right size, shipping legal AI agents. LinkedIn URL not publicly confirmed (left blank; not fabricated).
Simon Edwardsson — Co-Founder & CTO, V7 (V7 Labs)
Signal: 4 (ICP writing about agent architecture and context/token efficiency). Source: https://www.v7labs.com/about-v7 and his LinkedIn activity feed, read live 2026-08-30. Profile confirmed active at V7; note his LinkedIn public view does not display the title string, so CTO/co-founder comes from the V7 company page — worth confirming in first contact. Company size: 100+ employees; London, UK; Series A ($33M); document/agent automation for private markets, insurance and legal. Pain points (his own posts): on their Context Graph — "It connects the documents, data, entities, relationships and source evidence your firm already has, so AI can work with the full picture instead of guessing from whatever file you dropped into a prompt." On entity resolution: "How do we know if Apple (the record company) and Apple (the computer company) are two different entities quickly and cheaply with very low false positives. I think we have mostly solved it." Note "quickly and cheaply" — he frames context work explicitly in cost terms. Challenges: high-volume multi-step LLM document pipelines (tax audits, financial analysis, AP automation, investment memos) where per-document token cost lands directly on gross margin; agents that must trace an answer back to source evidence. He also writes about using coding agents to user-test his own tooling and cut a task "from ~20 steps to ~11, with the same prompt and output" — an engineer who instinctively measures agent step-count. Must-haves: context efficiency (less re-sending, fewer wasted steps), traceability back to source evidence. Nice-to-haves: per-workflow cost benchmarking across customers. ICP confidence: Medium-High — clean on stage, geo and agent-native product; the cost pain is evidenced through his own framing rather than a direct complaint, and the title needs one confirmation.
Siva Adhikarla — AVP Engineering, JSW One Platforms
Signal: 1 (ICP writing publicly and repeatedly about agent cost, agent loops and control).
Source: LinkedIn content search "agents in production" (past month) → author profile https://www.linkedin.com/in/siva-adhikarla/ (activity feed read directly this run). Post series over the last ~2 weeks:
• "AI Agents in Production" (14h old at time of scan) — "An agent isn't a chatbot with extra steps. It's a system making decisions with real consequences... Why non-determinism, cost, and failure modes compound in agent loops... scoping autonomy, capping the loop, human approval gates. Teams getting this right are not maximizing autonomy. They are scoping it tightly."
• "Moving towards LLMOps" (1w) — "If we ships LLM features without evals, tracing, and cost visibility, we are not ready to be in production."
• "AI Agent and Harness" (1w) — agent harness / control layer.
• "GenAI Architecture Patterns" (3d) — "The wrong pattern costs money and months of rework... don't default to agents."
Company size: ~399 employees (Tracxn, Mar 2026); other sources 485 (TheCompanyCheck) and 927 (PitchBook) — all inside the 50–2,000 band. JSW One Platforms = B2B commerce platform (JSW Group), Mumbai. He previously ran engineering at bigbasket.com (Associate Director) and IBM/Red Hat.
Pain points (expressed, verbatim/near-verbatim): non-determinism, cost and failure modes "compound in agent loops"; shipping LLM features "without evals, tracing, and cost visibility"; wrong architecture pattern "costs money and months of rework".
Challenges: controlling autonomy and loop depth; getting cost visibility and tracing in place before production; deciding when an agent is even the right pattern.
Must-haves: loop caps / execution budgets, cost visibility per agent run, tracing and evals, human approval gates.
Nice-to-haves: agent harness abstraction, standardised architecture decision framework for RAG vs fine-tune vs agents.
ICP confidence: Medium — AVP Engineering is VP-level and the expressed pain is the sharpest match found this run; headcount confirmed in band. Downgraded from High because JSW One Platforms is a corporate-backed B2B commerce platform, not a Series A–C SaaS/AI-native company, and I could NOT confirm 5+ agents actually running in their production stack (his content is leadership/architecture commentary, which may run ahead of deployment).
Signal: 3 (ICP at company publicly building on competitor tooling — Harvey documented "How Harvey Built Reliable AI Agents with LangSmith & Custom Tools") + 4 (building/shipping agents). Source: https://theorg.com/org/harvey-ai/org-chart/siva-gurumurthy , https://www.youtube.com/watch?v=kuXtW03cZEA , https://www.harvey.ai/blog/5-questions-with-siva-gurumurthy. Company size: ~350 employees (grew from ~100 in a year; $190M ARR; $11B valuation). 25,000+ custom agents run on Harvey executing M&A, due diligence, contract drafting, document review. CTO since May 2025 (ex-CTO Motive, ex-Director Eng Twitter Search/Recs). Pain points: building RELIABLE agents at scale, controlling/evaluating agent behavior for high-stakes legal work, tooling/observability (currently leaning on LangSmith + custom tools). Challenges: reliability and correctness of tens of thousands of custom agents; eval and guardrails for regulated legal output. Must-haves: agent reliability, observability/eval, control over tool calls. Nice-to-haves: cost attribution per agent run, cheaper routing. ICP confidence: High (CTO, decision-maker; company in size band; explicitly focused on reliable agents and already using a competitor in our category).
Signal: 1 (ICP writing about enterprise agents, reliability). Source: https://www.linkedin.com/in/sivasurend/ + podcast "Why Enterprises Need a Different Approach to AI Agents" (Chain of Thought/Galileo). Company size: ~73-174 employees (2026), Series A+ ($40M total, $250M valuation, Accenture-backed). Two-time founder (prev PowerUpCloud, acq. by L&T). Lyzr = enterprise agent infrastructure platform for building reliable, self-learning AI agents; on-prem enterprise deployments for finance/retail/government. Pain points: enterprise agent reliability, taking agents to production faster, self-learning/compounding agents. Challenges: reliability + governance for regulated enterprise agents at scale. Must-haves: production reliability, observability/control, on-prem/security. Nice-to-haves: agents that improve over time. ICP confidence: Medium (technical founder/CEO, 50-2000 emp, AI-native agent company; note Lyzr is an agent-infra platform — partially competitor-adjacent to Alpha's operating layer; validate positioning).
Signal: 4 (senior technical leader building/shipping an AI-native agent platform). Source: https://theorg.com/org/crescendo-3/teams/technology-leadership | funding/company: https://www.crescendo.ai/news/crescendo-closes-50-million-in-total-financing-with-latest-round-led-by-general-catalyst. Company size: ~1,900 employees as of 2026 (includes human concierge workforce — within 50–2,000); Series C, founded 2024, $500M valuation, General Catalyst-led. Crescendo is an AI-native customer support/contact-center platform blending conversational AI agents with human-in-the-loop; Sayko is credited on engineering of its LLM-based CX platform (alongside CTO Slava Zhakov and Tod Famous). Pain points (inferred from SVP Eng of an LLM CX agent platform): reliability/accuracy of agents in production, cost of LLM inference at contact-center volume, observability into agent cost and quality per interaction. Challenges: scaling a hybrid AI+human agent system reliably; controlling token/inference spend as automation share grows. Must-haves: production reliability, cost-per-run visibility, eval + monitoring. Nice-to-haves: model routing, context optimization. ICP confidence: Medium-High (senior technical leader, Series C, AI-native agent company in ICP size band; pain inferred from role + company). METHOD NOTE: Chrome/LinkedIn NOT connected this run; verified via web research (org chart + company news). NOTE: distinct person from existing brain entry "Slava Zhakov" (Crescendo CTO). LinkedIn URL not directly verified.
Signal: 4 (ICP technical co-founder at AI-native company building/shipping CX agents in production). Source: https://www.crunchbase.com/organization/crescendo-874d ; https://www.crescendo.ai/news/crescendo-closes-50-million-in-total-financing-with-latest-round-led-by-general-catalyst ; https://tracxn.com/d/companies/crescendo. Company size: AMBIGUOUS — Crescendo is an "augmented AI" CX platform (AI + human-in-the-loop, partnered with Alorica); GetLatka reports ~1.9K but that likely includes human support/BPO staff, while the core tech org is smaller (est. low hundreds). Series C $50M at $500M post-money (Oct 2024, led by General Catalyst). Independent. Treated as fitting 50-2,000 band for the tech org; flag headcount for confirmation. Actively building agents: AI-native customer-support platform blending conversational AI agents with human assistance in production. Zhakov is co-founder & CTO (technical decision-maker); Matt Price co-founder/CEO. Pain points (inferred): reliability/quality of CX agents in production; cost per resolution as agent volume scales; blending autonomous agents with human handoff. Challenges: production reliability across enterprise CX; cost economics per conversation. Must-haves: reliability, agent observability, cost-per-resolution visibility. Nice-to-haves: behavior control/guardrails, model routing. ICP confidence: Medium (technical co-founder/CTO at an independent AI-native agent company; softened by headcount ambiguity and LinkedIn not verified). LinkedIn URL left blank (not fabricated).
Smruti Patel — SVP of Engineering, Apollo GraphQL
Signal: 4 (ICP speaking publicly about the platform-engineering implications of agents).
Source: https://qconsf.com/speakers/smrutipatel — QCon SF 2026. Talk: "Platform Engineering's Second Act: From Vending Machine to Passport Control." Title corroborated by https://theorg.com/org/apollo-graphql/org-chart/smruti-patel and Apollo's own press release.
Company size: ~180–220 (LeadIQ ~180, May 2026; a16z jobs page "220+ employees in 10 countries"). Series D API/GraphQL platform; shipping MCP + agent-facing API tooling (Apollo MCP Server) so agents can call enterprise APIs safely.
Pain points: From her abstract — "Those pillars haven't moved. AI has just rewritten what each one requires, and the platform team's job along with it." Her framing (vending machine → passport control) is precisely the governance/control shift we sell into: platform teams moving from self-serve provisioning to gating what agents are allowed to do.
Challenges: Retooling a platform org for non-deterministic agent workloads; access control and guardrails for agents calling production APIs.
Must-haves: Policy/permission layer for agent→API calls; visibility into what agents are actually invoking.
Nice-to-haves: Cost attribution per agent-driven API call.
ICP confidence: Medium-High — SVP Engineering, company size clean, agent-adjacent product surface. Background: previously led Data Platform engineering at Stripe (scaled team 30→~180) and infra at VMware.
CAVEAT: One QCon bio phrasing reads "was most recently the SVP of Engineering at Apollo Graph," which may indicate she has departed. VERIFY current employer before outreach.
Signal: 4 (senior AI leader at an AI-agent-native company). Discovery: LinkedIn People search scoped to in-band agent companies — content-search buckets 1–3 this run surfaced mostly vendors/educators/enterprise, so pivoted to company-scoped people search. Source: https://www.linkedin.com/in/sofiezarrabi/ | Company: Cresta — unified platform for human + AI agents for CX. Company size: ~650 employees (Revelio ~618–693, 2026); Series D — in band. Pain points (INFERRED from role, not quoted this run): reliability of agents in production, cost/latency per run at enterprise contact-center scale, limited per-run cost visibility across many deployed agents. Challenges: deploying & maintaining reliable agents across many enterprise customers; moving pilots to full production. Must-haves: production observability, cost-per-run visibility, reliability guardrails. Nice-to-haves: cross-agent benchmarking, automated cost optimization. ICP confidence: High (Head-of-AI-level title; AI-native agent company; in-band size).
Sophie Roberts — Senior Vice President, Engineering, Lattice
Signal: 4 (engineering exec at a company that has shipped AI agents as a standalone product).
Source: LinkedIn people search — headline verbatim: "Senior Vice President, Engineering @ Lattice", based in Dublin, Ireland. Company agent activity verified via lattice.com/blog/lattice-ai-agent-meet-your-new-partner-in-hr, lattice.com/blog/june-2026-product-updates, and prnewswire.com/news-releases/lattice-launches-workforce-intelligence-for-ai-era-...-302796433.html.
Company size: ~532–574 employees in 2026 (Revelio ~532 as of 30 June 2026; other sources ~537 April 2026 and ~574; down from 651 in 2024). Comfortably inside the 50–2,000 band. Note: later-stage (Series F) than the stated Series A–C band.
Pain points (inferred from product surface, NOT from a personal post — she was not observed posting): Lattice launched AI Agent as a standalone product at Lattiverse (June 2026) that acts on HR data, answers employee questions and coaches managers in the flow of work, plus an agent embedded in 1:1s and a voice modality. Multiple agent surfaces across 5,000+ customers means per-customer, per-conversation cost exposure and reliability risk on sensitive HR data.
Challenges: agents operating on employee/HR data raise correctness and auditability stakes; voice modality adds latency and cost per interaction; scaling agent surfaces across a large customer base while headcount has been shrinking (651 → ~532), so efficiency pressure is real.
Must-haves: cost per agent conversation at customer granularity; reliability guarantees on agent outputs grounded in HR data.
Nice-to-haves: coaching/eval loops to improve agent quality without proportional spend increase.
ICP confidence: Medium — SVP Engineering title and headcount both qualify, and the company demonstrably ships agents in production, but no first-person pain signal was observed and it is unconfirmed that agents sit directly in her remit. Verify ownership before outreach.
Soups Ranjan — Co-founder & CEO (technical; ex-Coinbase Dir. of Data Science/Risk), Sardine
Signal: 4 (ICP technical founder building/shipping agentic AI in production). Source: https://www.sardine.ai/blog/series-c-announcement ; https://www.fintechfutures.com/venture-capital-funding/fraud-prevention-platform-sardine-secures-70m-series-c-invests-in-agentic-ai ; https://www.sardine.ai/about ; https://tracxn.com/d/companies/sardine-technologies ; https://www.linkedin.com/in/soupsranjan/ . Company size: 330 employees (May 2026); Series C ($70M led by Activant, with a16z, GV, Nyca, Moody's, Experian; $145M total). "Agentic risk platform for financial crime." Rolling out multiple production AI agents: KYC onboarding agents (edge-case resolution), sanction-screening agents (automated audits), merchant-risk agents (risk scoring/credit decisioning), disputes agents (chargeback formatting). Ranjan is technical co-founder & CEO — 20+ yrs in software engineering, data science & risk (ex-Director of Data Science & Risk at Coinbase; PhD). Pain points: reliability/accuracy of agents making high-stakes fraud/compliance decisions in a regulated domain; auditability of agent reasoning; controlling agent behavior across four distinct agent types. Challenges: production-grade trust for autonomous risk decisions; visibility into what each agent does and costs. Must-haves: reliability, auditability/traceability, behavior control. Nice-to-haves: per-run cost attribution, model routing. ICP confidence: Medium-High (technical co-founder; CEO title vs CTO is the only caveat; 330-person Series C AI-native company shipping multiple risk agents in production).
Sourav Dasgupta — Director of Engineering, Harness
Signal: 4 (ICP author shipping agents — LinkedIn post: "At Harness, we are building data platforms with agents as our primary audience... building new platform components from the ground up while reshaping existing ones"). Source: post surfaced via content search "AI agent platform engineering scale" (https://www.linkedin.com/in/sourav-dasgupta-59191036/). Company size: 501-1,000 (LinkedIn band; ~1,818 associated members). Pain points: reshaping data platforms for AI/agent consumption is non-trivial; in the AI world an agent may have the last word, requiring a significantly higher degree of reliability and much stronger governance than human-in-the-loop BI; many data platforms will struggle in this transition and hold their orgs back; foundations must be right or AI efforts fail. Challenges: first-principles redesign of platform components for an agent-primary audience; guaranteeing reliability + governance for autonomous agent decisioning. Must-haves: a reliability and governance layer for agent-consumed data/platforms; solid platform foundations before scaling autonomous agents. Nice-to-haves: abstraction-first platform tooling. ICP confidence: Medium-High (Director of Engineering, 501-1,000 emp, actively building agent-first data platforms; later-stage funding than the Series A-C ICP band — flagged).
Souvik Sen — Co-founder & CTO / Head of Engineering, Ema
Signal: 4 (ICP technical leader building & shipping agents in production). Source: https://www.linkedin.com/in/sen-souvik , https://www.ema.ai/about-us , https://tracxn.com/d/companies/ema/ , https://www.ema.ai/blog/agentic-ai/ema-revolutionizes-enterprise-ai-new-funding-ai-employee-builder-and-on-prem-deployment. Company size: 228 employees (as of May 2026). Series A, ~$61M raised (Accel, S32, Prosus, Hitachi Ventures); Mountain View; founded 2023 by Souvik Sen & Surojit Chatterjee. "Universal AI Employee" — pre-built multi-agent platform executing complex enterprise workflows across many apps; launched Builder Academy + on-prem deployment for production-ready agentic AI. Sen is technical co-founder & CTO (ex-Okta VP Engineering leading Data/ML/Devices; ex-Google, led TrustGraph ad-fraud ML). Pain points: scaling many enterprise AI-employee agents reliably to production across customers; on-prem/regulated deployment; per-workflow cost economics of multi-agent runs. Challenges: production reliability across enterprise deployments; controlling agent behavior; cost/observability at scale. Must-haves: production reliability, observability, cost visibility per agent run. Nice-to-haves: behavior guardrails, cross-app orchestration control. ICP confidence: High — technical co-founder/CTO decision-maker at a 228-person AI-native company shipping multi-agent workflows in production.
Spiros Xanthos — Co-Founder & CEO (technical; OpenTelemetry co-creator, ex-Splunk SVP/GM Observability), Resolve AI (resolve.ai)
Signal: 4 (ICP technical founder building/shipping agents in production). Source: Greylock "How Resolve AI is Building Agents to Keep the World's Software Running" (https://greylock.com/greymatter/how-resolve-ai-is-building-agents-to-keep-the-worlds-software-running/); Infinite Curiosity podcast "Putting AI On-Call for Humans". Company size: ~100–150 est. (see co-founder Mayank Agarwal entry). Technical founder (co-created OpenTelemetry; prior exits Omnition→Splunk, Pattern Insights→VMware). Publicly frames Resolve as "AI for production" / putting AI on-call. Pain points (inferred): economics of operating continuously-running agents at enterprise scale; agent reliability for autonomous incident resolution; visibility/observability into agent actions and cost. Challenges: proving trustworthy autonomous action in high-stakes production. Must-haves: reliable + affordable agent operation at scale. Nice-to-haves: continual improvement of agents. ICP confidence: High (technical co-founder/CEO of AI-native agent company shipping in production; right size).
Spurti Kanduri — Global Head of Solutions Engineering, Relevance AI
Signal: 4 (technical leader at a qualifying agent-native company; found via LinkedIn people search + web verification of company). Source: LinkedIn people search (Relevance AI); company verified via https://tracxn.com/d/companies/relevance-ai. Company size: ~124 employees (51-200 range), Series B (~$37M raised); no-code "AI workforce" platform for building & deploying AI agents. Role: Global Head of Solutions Engineering (customer-facing technical leadership — deploying/scaling customer agent workloads in production). Pain points (role-inferred): helping customers move agents from pilot to production reliably; per-agent cost/token waste as customers scale from 1 -> many agents; visibility into what agents cost per run. Challenges: reliability and cost predictability of customer-deployed agent fleets. Must-haves: production reliability; cost-per-run visibility for customer agents. Nice-to-haves: optimization/routing to reduce customer agent spend. ICP confidence: Medium (Head-level technical role, but solutions/customer-facing rather than core AI/eng; company is squarely agent-native and in size range). NOTE: pain points inferred from role/company context, not a verbatim quote.
Sreenivasulu Anantha — Head of AI, SunTec Business Solutions
Signal: 4 (ICP publishing about enterprise agentic AI and measurable outcomes). Source: https://www.linkedin.com/in/ananthasreenivasulu/ — multi-part series "Outcome Intelligence: The Next Evolution of Enterprise AI"; headline "Voice AI, Agentic AI & Intelligent Automation | Driving Measurable Business Outcomes | Trusted AI for Regulated Industries". Found via LinkedIn people search "Head of AI agents evals production SaaS" (2026-08-29). Company size: ~560–875 employees (Tracxn 559 Jul 2025; Latka 875; LeadIQ ~846 Feb 2026). BFSI pricing/billing SaaS, private. Pain points: proving measurable business outcomes from agentic AI in regulated BFSI environments — i.e. tying agent runs to outcomes rather than activity. Challenges: trust, governance and auditability for agents in regulated industries; moving from intelligent automation to autonomous agents. Must-haves: outcome-level measurement and governance over agent behaviour. Nice-to-haves: cost-per-outcome economics to support enterprise business cases. ICP confidence: Medium — Head of AI at a ~600–900 person B2B SaaS building voice + agentic AI; agent count in production not yet verified.
Sreevathsa Duglapura — SVP Engineering & GM, Ushur India, Ushur
Signal: 4 (company-scoped ICP people search on Ushur; found via LinkedIn people search, NOT an observed post — pains INFERRED from role+company). Source: https://www.linkedin.com/search/results/people/?keywords=Ushur%20director%20VP%20engineering%20AI%20agents | Company size: ~300 (Ushur, enterprise agentic-AI automation, Series C). Pain points (inferred): owning eng org shipping agents in production; scaling reliability and cost efficiency across a growing India eng team. Challenges (inferred): cross-region scaling of agent reliability; keeping per-run cost predictable. Must-haves (inferred): production reliability + cost-per-run visibility. Nice-to-haves (inferred): token/context-waste reduction. ICP confidence: High (SVP Engineering / GM = senior technical leader at agent-native company in range; verified title+company, pains inferred).
Sri Subramaniam — VP of Engineering (Life Sciences), Hippocratic AI
Signal: 4 (ICP building/shipping agents in production — found via 2026 research on healthcare voice-agent companies). Source: https://www.prnewswire.com/news-releases/hippocratic-ai-expands-life-sciences-leadership-team-as-pharma-and-medtech-demand-for-voice-ai-agents-accelerates-302778102.html (appointment announced May 2026). Company size: ~150-250 employees; heavily funded (Series B stage, healthcare voice AI, Polaris model family). Background: ex-Director of SW Dev at Amazon (Alexa AI Voice Agent teams), ex-VP AI at Credit Karma, ex-Walmart Labs founding leader. Actively building AI agents: leads engineering for pharma/biotech/medtech-specific voice agents in production. Pain points (inferred, not verbatim): reliability & safety of clinical voice agents in production; non-deterministic outputs unacceptable in regulated healthcare; scaling domain-specific agents on Polaris. Challenges: guaranteeing correctness where regression tests don't catch failures; regulated-industry guardrails. Must-haves: production reliability, guardrails, observability into agent behavior. Nice-to-haves: cost attribution per agent run. ICP confidence: High (VP Engineering, ~150-250-person Series-B AI-native company shipping healthcare agents in production).
Signal: 4 (company-scoped ICP people search on Gnani.ai; found via LinkedIn people search, NOT an observed post — pains INFERRED from role+company, not quoted). Source: https://www.linkedin.com/search/results/people/?keywords=Gnani.ai%20engineering%20AI%20agents%20director%20VP | Company size: ~250 (Gnani.ai, voice-AI-agent company, Bengaluru/SF, Samsung Ventures-backed, '30M+ voice interactions/day'). Pain points (inferred): running high-volume voice agents in production at telephony scale — per-interaction LLM/ASR/TTS cost, latency, and reliability at 30M+ daily calls. Challenges (inferred): keeping cost-per-call viable while scaling agents; real-time reliability. Must-haves (inferred): per-run/per-call cost visibility + production reliability for real-time voice agents. Nice-to-haves (inferred): token/context-waste reduction, routing to cheaper models. ICP confidence: High (VP of AI = senior AI leader; ex-Observe.ai; agent-native company in 50-2,000 range; title+company verified on LinkedIn, pains inferred).
Srinath Sridhar — Co-founder & CEO (technical, ex-Google/Meta), Regie.ai
Signal: 4 (ICP technical co-founder shipping sales agents in production). Source: https://www.regie.ai/about ; https://venturecapital.com/news/technology/regieai-launches-regieone-ai-platform-with-30m-series-b-funding-v1 ; PitchBook/LeadIQ (headcount ~99-100 as of Dec 2025). Company size: ~100 (across 5 continents). Stage: Series B ($30M; ~$50M total). Technical co-founder (ex-Google, Facebook, prior co-founder of Onera); business co-founder is Matt Millen (ex-Outreach). Actively shipping agents: RegieOne platform incl. autonomous AI Sales Agents that prospect for sales teams in production. Pain points (inferred from product + domain, NOT verbatim quotes): reliability and control of autonomous outbound agents; cost of LLM calls across high-volume prospecting; visibility into agent actions/output quality. Challenges: keeping autonomous sales agents on-brand and controllable at scale; managing token/model spend as agent volume grows. Must-haves: agent behavior control, reliability, cost visibility. Nice-to-haves: compounding improvement across campaigns. ICP confidence: Medium-High (technical co-founder & decision-maker; ~100 employees; Series B; ships autonomous agents in production).
Srinivasa D Chakravadhanula — Sr Director of Engineering & India Site Lead, SoundHound AI
Signal: 4 (senior technical leader at agent-native company shipping agents). Source: https://www.linkedin.com/in/srinivasachakravadhanula/ (LinkedIn people search "SoundHound AI director engineering agents"). Company size: ~750-950 (SoundHound AI). Ships agents: YES (Amelia agentic AI, enterprise voice/conversational agents). Pain points (inferred from role): scaling engineering org + production agents across a global site, cost/reliability of agents at enterprise scale. Challenges: 1->many agents in production, cost control. Must-haves: reliability + cost governance for production agents. Nice-to-haves: cross-team agent observability. ICP confidence: Medium (title + company confirmed; found via company-scoped search, no individual signal observed).
Signal: 4 (ICP senior technical leader at an agent-native company; discovered via LinkedIn people-search of companies actively shipping AI agents) | Source: https://www.linkedin.com/in/spatchigolla | Headline: Senior Director @ ThoughtSpot (building Spotter agentic analytics) | Ex-Amazon | SaaS, Security, Infrastructure Leadership | Company: ThoughtSpot — actively building/shipping AI agents in production | Company size: ~1,000–1,400 (estimate) | Pain points (INFERRED from role/company context — not a verbatim quote observed this run): bridging 'cool AI demo' to 'trusted production system' for agentic analytics; production reliability and trust of agent outputs on enterprise data; cost/efficiency per agent run | Challenges: making agentic analytics dependable and economical enough for strategic enterprise customers | Must-haves: production reliability/trust; cost-per-run visibility | Nice-to-haves: observability into agent behavior; context optimization | ICP confidence: Medium-High (senior technical/eng leadership at a 50–2,000-employee company shipping AI agents)
Signal: 4 (senior eng leader at an agent platform company). Source: LinkedIn people search (https://www.linkedin.com/search/results/people/?keywords=Kore.ai%20VP%20of%20AI%20OR%20head%20of%20agentic). Company size: ~1,277 employees (Kore.ai Agent Platform — agentic enterprise AI; mature/growth stage). Pain points (INFERRED): reliability and cost of running enterprise agents at scale; observability across deployments. Challenges: reliability, cost visibility, governance at scale. Must-haves: visibility/control over agent run cost + reliability. Nice-to-haves: model routing/optimization, evals. ICP confidence: Medium (agent platform, right-size; late/growth stage beyond Series A-C — flagged; VP Eng is a valid senior technical buyer).
Signal: 4 (ICP technical leader at a company shipping AI agents). Method: web/company-leadership research (LinkedIn content/people search low-yield this run). Source: https://avaamo.ai/ , DQ India interview 'Beyond Chatbots: how Avaamo is Transforming Enterprise AI with Autonomous Agents', his posts #aiagents #digitalworkforce #avaamo. Alt profile: /in/sriram-chakravarthy-11b730268/. Company size: 142 employees (May 2026). Series A (Intel Capital). Agent activity: autonomous enterprise AI agents / 'digital workforce' across healthcare, banking, telecom (confirmed). Pain points (INFERRED, not verbatim): reliability/determinism of autonomous agents in regulated enterprise workflows; safety/guardrails (Avaamo GPT positioned as 'safest path'); scaling agents across use cases; token/cost visibility. Challenges: enterprise-grade trust + auditability. Must-haves: guardrails + reliability in production. Nice-to-haves: per-run cost control. ICP confidence: Medium-High (right size 142, agent-native, CTO; pain points inferred).
Signal: 4 (ICP technical co-founder/CTO at agent-OS company shipping agents in production). Source: Gravity Q3-2026 AI-agent funding tracker (https://gravity.fast/blog/ai-agent-funding-tracker-q3-2026/) + GeekWire (https://www.geekwire.com/2026/after-hiring-aws-exec-and-raising-107m-seed-round-virginia-startup-plants-flag-in-seattle-area/) + LinkedIn. Company size: 56 employees (confirmed). Trase = agentic operating system for regulated industries (healthcare + defense); orchestration, governance, security across cloud/on-prem/edge; $107M seed (ARCH Venture Partners, Red Cell); early customers Duke Health, US Navy; agents on Google Cloud Marketplace. Koneru is ex-AWS GM of Bedrock Agentic AI Infra & GenAI Services, ex-Google/Salesforce senior eng director. Pain points: running many agents reliably where a wrong answer carries serious consequences; orchestration/governance/security across heterogeneous environments. Challenges: production reliability + compliance for agents in regulated verticals. Must-haves: agent observability, guardrails, control at scale. Nice-to-haves: per-run cost visibility, model routing. ICP confidence: High (CTO, agent-native, 56-person Series/seed-stage company shipping 5+ agents in production).
Signal: 4 — technical co-founder publicly discussing shipping agents in production (MLOps Community podcast) and building Dust's "Deep Dive" agents; posts regularly about production agent architecture.
Source: MLOps Community podcast (Co-Founder & CTO of Dust episode) + https://tech.eu/2026/05/18/dust-raises-40m-series-b-to-build-the-multiplayer-operating-system-for-enterprise-ai/
Company size: 144 employees (Series B, $40M announced May 2026, ~$60M total; platform used by 51,000 workers across 3,000+ companies).
Pain points: production quality gates, safety vs. cost tradeoffs, managing long-tail agent failures, deployment patterns as agents go multi-agent/"multiplayer".
Challenges: keeping multi-agent systems reliable and cost-efficient across an entire org.
Must-haves: control + observability over agent behavior at scale.
Nice-to-haves: per-agent cost attribution.
ICP confidence: High — technical co-founder/CTO at a 144-person Series B shipping 5+ agents in production. Note: Dust is an agent-platform vendor, so partner/competitor-adjacent.
Signal: 1 (ICP CTO at a company publicly focused on agent reliability at scale). Source: https://www.parloa.com/parloa-in-the-press/parloa-navigator-and-parloa-lens-launch-to-connect-building-reliable-ai-agents-and-trusting-them-at-scale/ and https://openai.com/index/parloa/ . Company size: est. ~350-450 (Berlin, Series C, $120M+ raised; enterprise voice-agent platform handling millions of customer interactions). Pain points: reliability of voice agents in production (instruction-following consistency, API-calling consistency, latency, edge cases); building trust in agents at scale; moving beyond abstract benchmarks to production-grade evaluation. Challenges: measuring & guaranteeing agent reliability across millions of live interactions; simulation/eval pipelines that mirror production; managing latency+performance under real load. Must-haves: production reliability + trust/observability at scale (their Navigator/Lens products are literally "build reliable agents / trust them at scale"). Nice-to-haves: per-interaction cost efficiency. ICP confidence: High — technical co-founder/CTO decision-maker at a ~400-person AI-native company shipping agents in production; reliability-at-scale is their stated core problem, directly on Alpha's wedge.
Stelios Modes — Co-Founder & CTO, Rillet
Signal: 4 (ICP building/shipping AI agents in production). Source: https://www.rillet.com/blog/rillet-raises-70m-series-b-from-andreessen-horowitz-and-iconiq ; https://finance.yahoo.com/news/rillet-raises-70m-replace-20th-123000646.html . Company size: ~191 employees (Series B $70M co-led by a16z & ICONIQ, Aug 2025; Sequoia, Oak HC/FT). AI-native ERP for finance teams where AI agents + human expertise close the books, reconcile and report. Pain points (INFERRED): reliability/accuracy of finance agents on the general ledger where errors are unacceptable; per-run cost of agentic accounting workflows at customer scale; auditability of what each agent did. Challenges (INFERRED): scaling agent automation across many customers' books without cost/reliability blowup. Must-haves (INFERRED): defensible, auditable agent actions + per-run cost visibility. Nice-to-haves (INFERRED): fleet-wide observability across agent types. ICP confidence: High (technical co-founder/CTO, Series B, ~191 emp in-band, agents explicitly in production). LinkedIn URL not surfaced in unauthenticated search — not fabricated.
Stephan Kletzl — Co-Founder & CTO, UserGems
Signal: Bucket 4 (ICP technical co-founder personally demoing multi-agent product capabilities).
Source: https://www.usergems.com/news/usergems-hosts-usergems-insiders ; https://www.usergems.com/ ; https://www.cbinsights.com/company/usergems ; https://tracxn.com/d/companies/usergems
Company size: ~100–103 employees (Apr 2026). ~$22.4–24.4M raised; $20M Series A led by Craft Ventures with Battery Ventures (Oct 2021), YC seed. SF-based, founded 2019 with Christian Kletzl (CEO).
Agent activity (verified): UserGems ships "Gem-E," a multi-agent AI system for outbound/ABM — named agents include a Research Agent (pulls custom insights on accounts) and an Orchestration Agent (builds campaigns from a description, flags missing campaigns, suggests tweaks), plus intent-scoring and personalized-outbound agents. Stephan personally demoed new Gem-E agent functionality at the UserGems Insiders event (2026).
Pain points (inferred from architecture, not a quote): research agents that enrich every account in a customer's TAM are an unbounded token workload — cost scales with prospect volume, not seats. Multi-agent orchestration across research → prioritization → outbound is exactly the 1→5+ agents scaling wall in the ICP definition.
Challenges: keeping per-account research cost below the value of the pipeline generated; agent output quality (personalization accuracy) directly determines customer-visible quality; cost attribution per customer account.
Must-haves: per-agent and per-account cost visibility; guardrails on runaway research loops; quality regression detection across agents.
Nice-to-haves: caching/dedup of account research across customers; cheaper model routing for bulk enrichment steps.
ICP confidence: Medium-High — CTO + technical co-founder, ~100 emp (in range), genuine multi-agent production product. Downgraded from High because last disclosed round is a 2021 Series A (bottom of the A–C band, and stale — may be bootstrapped/profitable since). LinkedIn personal URL not confirmed; company page recorded instead. VERIFY personal profile before outreach.
Stephen Whitworth — Co-Founder & CEO (technical; ex-Monzo ML/data engineer), incident.io
Signal: 4 (ICP technical co-founder building/shipping incident-response AI agents; found via web research — LinkedIn search blocked this run). Source: https://www.insightpartners.com/ideas/incident-io-raises-62m-to-build-ai-agents-that-resolve-incidents-with-you/ ; https://www.linkedin.com/in/stephenwhitworth/ ; https://incident.io/blog/podcast-a-trip-down-memory-lane-with-the-founders . Company size: ~150-200 (mid-size; well-funded scale-up). Stage/funding: Series B ($62M round; Insight Partners). Actively building/shipping agents: incident.io is building a new generation of AI agents ("AI teammates") that investigate incidents, surface root causes, and recommend/apply fixes for on-call teams. Technical co-founder (ex-Monzo engineer; built the platform from a 2020 side project). NOTE: DIFFERENT person from existing incident.io record Pete Hamilton (CTO, id 229) — net-new. Pain points (role-inferred, public framing = "AI agents that resolve incidents with you"): reliability/trust of autonomous agents acting in production incidents; cost of always-on/continuously-running agents; observability into agent actions. Challenges: safe autonomous remediation in high-stakes incidents; scaling agent reliability across customers. Must-haves: production reliability, guardrails, cost/observability per agent run. Nice-to-haves: inference cost efficiency. ICP confidence: High (technical co-founder/CEO of AI-native agent company shipping in production; right size).
Signal: 1 (ICP co-founder writing about agent reliability in production). Source: https://www.linkedin.com/posts/shind_when-jamie-hall-and-i-started-lorikeet-we-activity-7315493651864788994-8b5Y ; https://prodsens.live/2026/05/28/building-lorikeet-how-ai-humility-and-a-dual-agent-architecture-are-redefining-customer-support/ (profileUrl derived from his post vanity handle 'shind'). Company size: Lorikeet ~50-80 employees; Series A ($58M raised). Pain points: high-stakes agent reliability in regulated industries (fintech/healthtech/energy); avoiding hallucination on customer-facing actions. Challenges: dual-agent (Concierge+Coach) reliability; correct escalation; production trust. Must-haves: reliable, auditable agent behavior on high-risk workflows. Nice-to-haves: continuous agent improvement tooling. ICP confidence: Medium (co-founder/CEO decision-maker; strong reliability signal; company at lower bound of the size range; technical CTO Jamie Hall already tracked separately).
Steve Yazicioglu — Head of Forward Deployed Engineering, Candid Health
Signal: 4 (ICP leader at a company actively shipping AI agents). Found via LinkedIn company People-tab sweep of Candid Health.
Source: https://www.linkedin.com/company/candid-health/people/?keywords=engineering
Company size: 51-200 employees (per Candid Health LinkedIn company page). Candid Health = AI-native medical billing / revenue cycle; agents in production against real payer workflows.
Pain points: NONE DIRECTLY OBSERVED THIS RUN. Inferred only: FDE leaders own agent behaviour per customer deployment, so they feel per-customer agent cost and per-customer failure modes first and most directly.
Challenges (inferred): agent behaviour drifting per customer configuration; no clean per-customer view of what each deployment's agents cost to run.
Must-haves (inferred): per-deployment cost and reliability visibility; fast diagnosis of which step in an agent run failed at a specific customer.
Nice-to-haves (inferred): reusable eval sets built from production traces per customer.
ICP confidence: Medium. Head-of-Engineering-function level and a strong agent company, but "Forward Deployed Engineering" is a customer-facing eng org rather than core platform — likely an influencer/champion rather than the budget owner. Pain is inferred, not expressed.
Signal: 4 (ICP building/shipping agents). Source: https://research.contrary.com/company/cognition and https://en.wikipedia.org/wiki/Cognition_AI. Company size: ~150-400 (AI-native, Devin autonomous coding agent; enterprise usage up 10x in 2026, ~$492M ARR run-rate). Pain points: running autonomous coding agents reliably in production (Devin must check/test its own output to be production-safe); scaling agent volume as enterprise adoption 10x's. Challenges: agent reliability at scale, cost of large autonomous coding runs, keeping non-deterministic agents production-grade. Must-haves: reliability/verification of agent output, cost control on high-volume agent runs. Nice-to-haves: per-run cost visibility, efficiency gains. ICP confidence: High (CTO/co-founder at AI-native company shipping agents in production, target size band). LinkedIn slug not confirmed — profile_url blank to avoid fabrication.
Signal: 4 (ICP technical leader shipping conversational agents at massive scale across housing & healthcare via email/text/phone). Source: https://www.linkedin.com/in/stoyan-tony-stoyanov-07690a53/ ; https://www.bvp.com/news/building-the-backbone-of-operational-excellence-with-eliseais-vertical-automation . Company size: ~599 employees; Series D, >$1B valuation (NOTE: funding stage exceeds the A-C target, but headcount is within 50-2000 and agent-shipping is core). Pain points: reliability of conversational agents at very high volume; multi-channel orchestration; accuracy in regulated housing/healthcare workflows. Challenges: scaling agent volume without quality degradation; cost per conversation at scale; cross-channel observability. Must-haves: reliable multi-channel agents, monitoring, guardrails. Nice-to-haves: per-run cost optimization. ICP confidence: Medium (role/size/agent-shipping strong; funding stage D beyond A-C).
Signal: 4 (adapted — LinkedIn auth unavailable this run; identified via web research + verification). Source: https://en.wikipedia.org/wiki/Cursor_(company) ; https://www.crunchbase.com/organization/anysphere ; https://goldhouse.org/people/aman-sanger-sualeh-asif/. Company size: several hundred to ~1,000+ employees (within 50–2,000); Anysphere/Cursor is a leading AI coding-agent company running agentic coding at massive scale (very high token throughput). Co-founder and senior technical leader (titled CTO / CPO across sources). Pain points (inferred from the product's nature, not a verbatim quote): agent cost/token economics at extreme scale; reliability of autonomous multi-step coding agents; visibility into cost-per-run. Challenges: keeping agentic coding reliable and economical as usage explodes; controlling compounding token spend across parallel agent runs. Must-haves: cost-per-run visibility and token efficiency; reliability at scale. Nice-to-haves: fine-grained agent observability. ICP confidence: High — technical co-founder/CTO at a flagship agent-native company in the size range. LinkedIn URL not captured this run (title varies CTO vs CPO across sources — verify).
Subho Mukherjee — Chief Science Officer & Co-Founder, Hippocratic AI
Signal: 4 (ICP technical/AI leader at agent-native company shipping agents in production). Source: https://research.contrary.com/company/hippocratic-ai ; https://hippocraticai.com/team/ ; https://tracxn.com/d/companies/hippocratic-ai/ (312 employees, May 2026). Company size: ~312 employees. Series B, $141M raised at $1.64B valuation (Kleiner Perkins, a16z, General Catalyst). Building/shipping AI agents: safety-focused healthcare LLM powering patient-facing voice/chat agents (intake, scheduling, chronic-care follow-up) rolled out via an "AI agent app store" — i.e., scaling from a few agents to many. Subho Mukherjee is Chief Science Officer & co-founder (Head-of-AI-equivalent technical decision-maker); CTO is Saad Godil (already in brain). Pain points (inferred from product domain + public statements, NOT a verbatim quote): safety/reliability of patient-facing agents (they run 500k+ test calls with 7,000+ clinicians for output testing); scaling and governing a growing catalog of agents; cost/observability across many agents. Challenges: rigorous eval/testing at scale; controlling agent behavior across an app-store of agents. Must-haves: reliability/safety, evaluation infrastructure, control at scale. Nice-to-haves: per-agent cost visibility. ICP confidence: Medium-High (co-founder & Chief Science Officer / AI decision-maker; 312 employees in band; large and growing production agent fleet).
Suchet Bargoti — Director of Inspection and Mapping, Skydio
Signal: 4 (ADAPTED — public conference speaker listing, not LinkedIn; LinkedIn was unavailable this run).
Source: https://www.ai.engineer/worldsfair/schedule — AI Engineer World's Fair 2026, "Robotics & World Models" track, Day 3 (Wed 7/1).
Talk (verbatim): "From Manual Drones to Autonomous Multi-Agent Missions".
Company size: ~959–1,065 employees (Revelio 959 as of March 2026; Tracxn 1,039 as of 30 June 2026; RocketReach 1,065). Inside the 50–2,000 band. Late-stage private (Series F-era), not Series A–C.
Pain points: INFERRED FROM TALK TITLE, NOT A VERBATIM QUOTE — the explicit arc is single-unit autonomy to coordinated multi-agent missions, i.e. the 1 -> 5+ agents wall.
Challenges: coordination and reliability once multiple autonomous agents operate on one mission.
Must-haves: reliability and control at multi-agent scale.
Nice-to-haves: run-level cost/telemetry attribution.
ICP confidence: Medium — Director-level, qualifying headcount, and a clean multi-agent-scaling signal; downgraded because this is robotics/embodied autonomy rather than LLM token economics, so the cost-blowout pain may not translate.
NO LinkedIn URL captured — do not fabricate one.
Sudarshan Kamath — Co-founder & CEO (technical), Smallest.ai
Signal: 1/4 (technical co-founder/CEO at a company shipping voice agents; web research this run, LinkedIn not connected). Source: https://techcrunch.com/2026/07/31/smallest-ai-raises-13m-to-build-ultra-fast-voice-ai-that-sounds-genuinely-human/ ; https://www.sierraventures.com/content/sierra-ventures-our-early-stage-investment-in-smallest-ai . Company size: ~60 employees (Series A, 2026; total ~$21M raised). Smallest.ai builds a full-stack voice AI platform for enterprises to build/deploy/scale real-time voice agents. Pain points (inferred; no verbatim quote this run): latency + cost economics of real-time voice agents at scale, production reliability, controlling per-call/token spend. Challenges: making concurrent voice agents fast, cheap, and reliable enough for enterprise production. Must-haves: low cost-per-run, latency, reliability. Nice-to-haves: cost-per-run visibility. ICP confidence: Medium-High (technical co-founder/CEO — IIT Guwahati, ex-Bosch engineer — at a 50+ emp Series A voice-agent company; C-suite technical founder, valid per ICP). Profile URL not captured — do not fabricate.
Sujit Karpe — CTO and Co-Founder, iMocha
Signal: 4 (technical co-founder/CTO shipping AI agents). Source: https://www.linkedin.com/in/sujitkarpe/ — authored post describing the iMocha AI Readiness Agent that identified 480 employees below required AI fluency and then autonomously created personalized learning paths, scheduled assessments, notified managers, and tracked progress; framed as AI agents that take action, from AI insights to AI execution, hashtag AgenticAI. Company size: ~300-500 (iMocha, skills-intelligence SaaS, Pune). Pain points: moving enterprise AI from insight/dashboards to reliable agentic execution at scale; trust/governance of autonomous action. Challenges: scaling agentic AI in enterprise, reliability, governance. Must-haves: agents that reliably take action, governance. Nice-to-haves: cost/observability per agent run. ICP confidence: Medium-High (technical co-founder and CTO directly shipping production agents at a 50-2,000-emp SaaS).
Suneeta Mall — Head of AI Engineering, Harrison.ai
Signal: 4 (ICP writing publicly about building and running agents). LinkedIn URL is printed on her own site footer and is verified — the strongest-sourced profile URL in this batch.
Source: https://suneeta-mall.github.io/about/ and https://suneeta-mall.github.io/blog/category/agentic-ai/
Company size: ~230-240 mid-2026 (PitchBook/LeadIQ). NOTE: Harrison.ai's own LinkedIn company page states the 51-200 band, so headcount is an estimate — either figure is inside 50-2,000. Sydney-based medical AI, Series B/C.
Pain points: Agent economics. Runs an ongoing Agentic AI blog series and built FableFlow, an open-source multi-agent production pipeline, with explicit economics reasoning — she calls out that "the actual economics" was "the biggest gap in the original analysis."
Challenges: Multi-agent pipeline orchestration and reasoning about real per-run cost rather than idealised architecture.
Must-haves: Honest cost accounting for multi-agent pipelines.
Nice-to-haves: Reusable patterns for agent pipeline design.
ICP confidence: Medium — Head-of-AI-Engineering title, in band, genuinely agent-native and technically deep. Downgraded from High because the cost evidence comes from a personal open-source side project, not from confirmed Harrison.ai production pain; Harrison.ai's own agent estate is unverified.
Sunil Chandra — Chief of AI & VP of Engineering, Mindtickle
Signal: 4 (senior AI/eng leader at a company shipping an autonomous agent operating system in production). Source: https://www.mindtickle.com/ai/ ; production evidence: https://aws.amazon.com/solutions/case-studies/generative-ai-mindtickle/
Company size: ~424-704 employees (2025-2026 aggregators) — in range. Series E ($100M, 2021) — later stage than the A-C band, flagged. Revenue enablement / GTM agents.
Pain points: Model portability and switching friction across providers, plus security posture for production LLM workloads. On-record from the AWS case study: "AWS's architecture and authentication mechanism makes switching between models seamless, and its strong security policies and tools allow us to build solutions more securely." Note: this is positively framed vendor-case-study language, not a candid complaint — weakest pain evidence in this run.
Challenges: ElevateOS deploys autonomous agents that coach reps, guide deals, enable buyers and unify the GTM stack — a genuine multi-agent surface across a large customer base; managing model choice and cost across those agents sits in his remit.
Must-haves: Ability to swap models per agent without breaking output quality; secure, governed production LLM operations.
Nice-to-haves: Cost-per-agent and cost-per-customer visibility; unified evals across agent types.
ICP confidence: Medium — title (Chief of AI + VP Eng) and headcount are clean fits and agents are demonstrably in production, but funding stage is Series E and the pain evidence is marketing-framed. Needs a stronger signal before prioritising.
Signal: 4 (ICP technical leader at a company actively shipping a fleet of agents). Source: https://www.prnewswire.com/news-releases/ironclad-unveils-new-ai-assistant-and-agents-to-bring-contract-intelligence-to-every-agreement-302717729.html ; https://ironcladapp.com/resources/articles/ai-agentic-launch ; https://www.law.com/legaltechnews/2026/03/19/ironclad-launches-ironclad-assistant-expands-agentic-ai-capabilities/ ; appointment: https://www.prnewswire.com/news-releases/ironclad-taps-former-google-snowflake-docusign-skadden-executives-for-key-leadership-roles-302520979.html ; LinkedIn: https://www.linkedin.com/in/sunita-verma-1a4491/. Company size: ~796-866 employees (Crustdata Jun-2026 866; Revelio Mar-2026 796; +33% YoY). Series E, AI contracting platform, 2B+ contracts processed. Agents in production (9+ named): Jurist drafting/review/research agents, Intake Agent, Redlining Agent, Conversational Search, plus Renewal / Cost Savings / Archive Agents in early access (launched Mar-2026). Pain points (inferred from public product positioning + launch messaging, not verbatim): running a growing fleet of distinct agents across one platform with 'agents push work forward, with you in control' as the explicit stance - i.e. per-agent control, guardrails and visibility across many agents at once; measuring per-agent impact (they publish agent-level metrics like 50% reduction in intake submission time), which implies a need for per-run cost/outcome attribution. Challenges: shipping agent after agent into a regulated, high-consequence workflow (legal agreements) where a wrong autonomous action is costly; 65%+ of customers already adopted AI so agent volume scales with the customer base. Must-haves: control/guardrails and auditability per agent; reliability on customer contract data; ability to prove per-agent ROI to enterprise buyers. Nice-to-haves: cheaper per-run inference; cross-agent orchestration and shared context. ICP confidence: High - CTO title, ~866 employees (in band), demonstrably 5+ agents in production, and a 25-year AI/infra background (ex-CTO Character.AI, ex-VP Google Core Labs) so she is a technical buyer who will understand a harness. NOTE: not yet contacted; research only.
Signal: 4 (ICP speaking about how they build and ship agents in production).
Source: AI Engineer World's Fair 2026 speaker record — https://ai.engineer/worldsfair/2026/speakers.json ; session "How Forward Deployed Engineering is done at Decagon" (Forward Deployed Engineering track).
Company size: ~405 employees (Revelio, Mar 2026); 434 (May 2026); grew 770% from 47 in 2023. Series D — $250M, Jan 2026, $4.5B valuation (Bloomberg/Forbes). AI customer-support agents are the entire product; agents run in production across a large enterprise customer base.
Pain points: NOT directly quoted — session abstract is empty in the source dataset. Inferred from role only: as FDE CTO he owns per-customer agent deployments, which is exactly where per-tenant agent cost, reliability and regression問題 surface. TREAT AS UNVERIFIED until a primary quote is found.
Challenges (inferred, unverified): standing up and maintaining many customer-specific agent configurations; keeping deployed agents reliable and economical per account.
Must-haves (inferred, unverified): per-customer/per-run cost and reliability visibility.
Nice-to-haves (inferred, unverified): reusable deployment tooling across FDE engagements.
ICP confidence: High on fit (CTO-level technical leader, 405 emp, pure agent company, agents at real production scale). Caveats: (a) Series D, one stage past the stated Series A–C band; (b) no LinkedIn URL in source dataset and no first-party pain quote yet — verify both before outreach. Do not fabricate a profile URL.
Surendranath C — Senior Director of Engineering, Gupshup
Signal: 4 (company-scoped LinkedIn people search — ICP technical leader at agent-shipping company). Source: https://www.linkedin.com/in/surendranath-c-b666363 (via https://www.linkedin.com/company/gupshup/people/?keywords=director%20engineering). Company size: LinkedIn band 1K-5K employees (Gupshup; conversational AI + AI agents / 'Conversational AI Agents for Every Customer Conversation'; est. ~1,500-2,000). Pain points (INFERRED from role+company, not a personally observed post this run): LLM/agent cost at very high message volume across BFSI/enterprise messaging; per-agent cost visibility; reliability/latency of conversational agents in production. Challenges: controlling spend and quality across a large multi-agent conversational platform; scaling many enterprise deployments. Must-haves: per-run cost observability, reliability guardrails. Nice-to-haves: model routing / token optimization. ICP confidence: Medium (clear Senior Director of Engineering = ICP title; company size band 1K-5K straddles the 2,000 ceiling, so size is near/over the cap — flagged. Brain already had Kunal Patke, SVP Engineering; this is net-new).
Suresh Parameshwar — Head of Engineering, Ema (Ema Unlimited)
Signal: 4 (senior eng leader at AI-native company shipping enterprise agents). Source: https://www.prweb.com/releases/ema-appoints-suresh-parameshwar-as-head-of-engineering-to-drive-scalable-enterprise-grade-ai-solutions-302457372.html ; https://www.ema.ai/blog/engineering-in-ai/welcoming-suresh-parameshwar-to-ema ; headcount via search (Ema ~246 employees as of July 2026). Company size: ~246 employees (founded 2023; ~$61M raised, Seed + Series A; builds "universal AI employees" / agentic AI platform automating enterprise knowledge work across many functions). Role: Head of Engineering (appointed May 2025), 25+ yrs building/scaling systems at Microsoft (Skype/Teams messaging & presence infra for hundreds of millions of users) and Productiv. Pain points (inferred from role + product domain, NOT a verbatim quote): making enterprise-grade agentic systems reliable and scalable in production; multi-agent orchestration across HR/IT/sales/finance workflows; cost/latency at enterprise scale. Challenges: scaling reliable agents across large enterprise deployments. Must-haves: reliability, enterprise-grade scalability/governance. Nice-to-haves: per-run cost visibility/observability. ICP confidence: High (named Head of Engineering, ICP title; company in 50-2,000 band actively shipping agents; Series A). Found via web research — LinkedIn/Chrome not connected this run.
Suresh Ponnusamy — Head of Platform Engineering, Atomicwork
Signal: 3 (ICP engaging with competitor content — named reference customer in a Maxim AI case study).
Source: https://www.getmaxim.ai/blog/scaling-enterprise-support-atomicworks-journey-to-seamless-ai-quality-with-maxim/
Company size: ~120–150 (ZoomInfo ~150, May 2026; PitchBook ~120).
Pain points (direct quotes): "Maxim has been a game-changer for our AI quality journey. From the start, multiple teams have relied on Maxim for comprehensive end-to-end testing and monitoring of all our AI features, enabling us to scale efficiently." And a hard buying requirement: "In the enterprise space, privacy and security are critical. Maxim's on-premises deployment within our VPC gives us all the benefits of advanced AI evaluation while keeping our data entirely within our secure environment — a critical requirement for our largest customers."
Challenges: Multiple teams shipping AI features in parallel with no shared testing/monitoring layer; enterprise customers will not accept agent telemetry leaving their VPC.
Must-haves: End-to-end agent testing and monitoring usable by multiple teams at once; VPC / on-prem deployment option (this is a hard gate for their enterprise deals).
Nice-to-haves: Cross-team quality reporting.
ICP confidence: Medium — company size and agent activity are a clean fit and the pain is well articulated, but "Head of Platform Engineering" is adjacent to rather than exactly on the target title list, and he is the second contact at an account we already touch (Jeegar Shah).
IMPORTANT for positioning: the VPC/on-prem requirement is a deal-gating constraint for this account — lead with deployment model, not features.
Surojit Chatterjee — Co-founder & CEO (ex-Coinbase CPO, ex-Google), Ema
Signal: 1 (decision-maker posting publicly about shipping teams of AI agents in production). Source: https://www.linkedin.com/posts/surojitchatterjee_agenticbusinesstransformation-aiemployees-activity-7381364940944703488-Dx50 and https://www.pmf.show/blog/ema-ai-agents-fortune-500-enterprise-sales-surojit-chatterjee. Company size: 228 employees (May 2026); Series A, $61M raised (Accel, S32, Prosus). Builds "AI employees" — teams of AI agents connecting to Workday/Salesforce/ServiceNow, automating handoffs across enterprise systems (55k-employee Hitachi deployment). Pain points: reliable multi-agent execution across many enterprise systems, automating manual handoffs, governance/policy adherence. Challenges: scaling agent teams across many Fortune 500 customers reliably. Must-haves: cross-system agent observability, reliability + policy guardrails, cost per outcome visibility. Nice-to-haves: easier agent-builder tooling for customers. ICP confidence: High (ships 5+ agents in production per customer; 228 employees; classic buyer). Note: co-founder Souvik Sen already in brain; Surojit is a distinct new contact.
Sushant Randive — Head, Gen AI Products and Services, Haptik
Signal: 4 (company-scoped ICP people search on Haptik; found via LinkedIn people search, NOT an observed post — pains INFERRED from role+company). Source: https://www.linkedin.com/search/results/people/?keywords=Haptik%20director%20VP%20head%20engineering%20AI%20agents | Company size: ~400 (Jio Haptik, conversational/GenAI CX agents; operates as ~300-500-person unit — CAVEAT: owned by Reliance Jio, so not an independent Series A-C company). Pain points (inferred): productionizing GenAI CX agents at consumer scale; cost-per-conversation and reliability. Challenges (inferred): reliability + cost at high message volume. Must-haves (inferred): production reliability + per-run/per-conversation cost visibility. Nice-to-haves (inferred): context-waste reduction, routing. ICP confidence: Medium-High (Head of GenAI Products = AI-focused product leader at agent company in size range; caveat = Jio-owned, not independent Series A-C; verified title+company, pains inferred).
Signal: 4 (ICP technical leader at a company shipping AI agents). NOTE ON METHOD: LinkedIn content/people search was low-yield this run (junior ICs/students dominate); found via web/funding research + company leadership verification. Source: https://www.haptik.ai/ (Enterprise-Grade CX AI Agents), Jio Haptik WhatsApp AI agents for SMBs launch 2025, https://www.linkedin.com/in/swapan-rajdev-64a0591a/. Company size: ~223 (Haptik business unit, Apr 2025; ~306-person team per getlatka 2025). Owned by Jio Platforms/Reliance but the Haptik unit sits in the 50-2,000 band. Agent activity: ships enterprise CX AI agents across chat + voice/WhatsApp channels (confirmed). Pain points (INFERRED from role+company context, not verbatim): reliability/accuracy of CX agents in production across many tenants; cost/efficiency of high-volume chat+voice agents; scaling agents across SMB + enterprise. Challenges: multi-tenant agent quality control at scale; multilingual reliability. Must-haves: per-agent reliability + cost visibility in production. Nice-to-haves: honest model comparison / routing to cut token spend. ICP confidence: Medium-High (right size, agent-native, CTO; caveat: Jio-owned, pain points inferred).
Signal: 4 (ICP building/shipping agents in production). Source: https://www.linkedin.com/in/voiceaiagent and https://www.observe.ai/about-us. Company size: 201-500 employees, Series C, AI-native contact-center CX platform shipping purpose-built AI agents in production (voice/chat agents, auto-QA, real-time assist) evaluating 100% of interactions. Technical co-founder. Pain points (inferred from role + agent fleet scale, not a verbatim quote): agent reliability/accuracy in regulated CX, LLM inference cost across high call/chat volume, visibility into per-interaction agent cost and behavior. Challenges: moving customers from pilot to production and keeping accuracy high enough for unsupervised operation while controlling spend. Must-haves: production reliability guardrails, cost-per-run visibility. Nice-to-haves: unified observability across a growing fleet of agents. ICP confidence: High — clean technical decision-maker at confirmed 201-500 headcount, AI-native shipping 5+ agents.
Swayam Prakash Behera — Group Vice President of Software Engineering, Netcore Cloud
Signal: 3 (ICP engaging with competitor/adjacent agent infrastructure — named AWS Bedrock Agents customer case study). NOTE: LinkedIn/Chrome not connected this run. LinkedIn URL not verified.
Source: https://aws.amazon.com/solutions/case-studies/netcore-bedrock-case-study/
Company size: 1,752 employees (Tracxn, Mar 2026) — top of our band but inside it. Martech/CX SaaS, India. Funding profile does not map cleanly to Series A-C (founded 1998, largely bootstrapped/PE-style growth) — flagged below.
What he said (verbatim, short): "Netcore's multi-agentic generative AI capabilities on Amazon Bedrock."
Pain points: latency and data-exposure risk routing agent traffic through public model APIs; coordinating multiple specialized agents against proprietary in-house models.
Challenges: integrating a multi-agent marketing assistant with fine-tuned proprietary deep learning models without public API exposure; per-agent model selection at scale.
Must-haves: secure/serverless foundation-model access, RAG integration, per-agent model routing.
Nice-to-haves: faster MVP delivery, vendor implementation support.
ICP confidence: Medium — VP-level technical leader, headcount in band, genuine multi-agent production deployment and a competitor-adjacent stack; downgraded from High because the company's stage (mature, non-VC-staged) sits outside the Series A-C definition.
Sybille Fuks — Global Director, Agent Architecture, Parloa
Signal: 4 (Director-level agent-focused leader at an in-band company shipping voice/CX AI agents; identified via Parloa LinkedIn people search). Source: https://www.linkedin.com/in/sybille-fuks/ . Company size: ~380 employees (2026, growing to ~600; TechCrunch/company); Parloa ships agentic AI for customer experience (voice/chat/messaging agents) in production; raised $120M Series C (2025) and $350M Series D (Jan 2026) — note: now later-stage than the A-C guideline but headcount solidly in-band. Pain points (inferred from role; no direct post/quote read this run): designing/operating large fleets of customer-facing voice agents => reliability in production, containing per-conversation LLM cost, context management. Challenges: agent reliability + cost-per-conversation at enterprise scale. Must-haves: per-agent/per-run cost visibility, reliability guardrails, observability. Nice-to-haves: cost optimization across many concurrent agents. ICP confidence: Medium (Director-level, agent-architecture focus with linguistics/conversational-design background; company in-band and agent-native; pains inferred not quoted).
Signal: 4 (ICP building/shipping agents). Source: https://www.glean.com/about and https://www.glean.com/blog/engineering-agents-feb-drop-2026. Company size: ~900-1,200 (AI-native enterprise work assistant; building context-aware agents; under 2,000 headcount). Pain points: enabling a distributed agent-development model across app teams; central AI team must provide eval infrastructure, prompt-tuning and recommendation primitives so teams can ship agents reliably. Challenges: standardizing eval/observability so many teams can build agents without each reinventing reliability; controlling agent behavior and cost across a large org. Must-haves: shared eval + observability infrastructure, prompt/cost tooling. Nice-to-haves: per-team cost attribution across multi-tenant agent workloads. ICP confidence: Medium (clear technical decision-maker at AI-native company shipping agents and within size band, but later funding stage than the Series A-C sweet spot). LinkedIn slug not confirmed — profile_url blank to avoid fabrication.
Taivo Pungas — Chief Technology Officer, Pactum AI
Signal: 4 (ICP writing/speaking publicly about shipping agents to production). Found via web research — LinkedIn/Chrome unavailable this run.
Source: https://nex.day/speaker/taivo-pungas/ ; https://pactum.com/about (leadership page) ; his own writing at https://www.taivo.ai/beyond-prototypes-shipping-trustworthy-ai/
Company size: ~166–167 employees (Tracxn 166 as of 31 May 2026; PitchBook and LeadIQ both 167). Series C. Autonomous supplier-negotiation agents for procurement/supply chain — customers include Walmart, Maersk, Honeywell; 50+ Global 2000 enterprises.
Agent evidence: platform ships four agent lines (Alignment, Procurement, Price List, Campaigns). Speaker bio: "CTO of Pactum AI, where AI agents autonomously negotiate tens of billions of dollars in supplier deals for Fortune 500 companies. He sees AI from both sides: building it into the product and deploying it across the entire company as an internal AI transformation initiative." Previously rebuilt Veriff's core product around AI; founded NFTPort. Speaks specifically on "Beyond prototypes: shipping trustworthy AI." 76–82% of surveyed suppliers preferred the agent interface.
Pain points: the prototype-to-production gap for AI features; trustworthiness/reliability of autonomous agents that transact real money; needing every negotiation action to be transparent, trackable and auditable in real time.
Challenges: keeping autonomous agents inside hard enterprise guardrails (pricing, terms, supplier, outcome thresholds validated before a negotiation starts); integrating agents natively into SAP Ariba/Coupa procure-to-pay rather than as a silo; running an internal AI transformation across a 166-person org while also shipping the product.
Must-haves: pre-negotiation input validation and policy enforcement; full audit trail of which data influenced each agent decision; ERP-native integration; agents that escalate complex situations to humans.
Nice-to-haves: agents that learn across thousands of negotiations to refine offer structuring; supplier-facing UX quality; internal agent tooling for non-engineering teams.
ICP confidence: High — CTO, Series C, ~166 people, AI-native, agents demonstrably in production at Fortune 500 scale, and he publicly frames his own topic as the prototype→production trust gap.
Signal: 1 (ICP personally publishing engineering content on agent evals/design). NOTE: LinkedIn/Chrome unavailable this run — found via company tech blog + press. Japanese name: 平村 健勝. Public on X as @hiratake55.
Source: https://tech.revcomm.co.jp/look-back-at-2025 (byline: "RevComm の CTO の平村") ; https://thebridge.jp/2025/04/fighting-customer-harassment-with-ai-revcomm-cto-hiramura-talks-the-future-of-business-conversations-hot100-accenture-ventures-vol-1
Company size: 303 as of March 2026 (269 Japan / 34 Indonesia) — RevComm official company page. Tokyo, Japan. Founding member, joined June 2018. CAVEAT: funding figures found (¥1.65B cumulative, ¥8.67B valuation) are from a 2021 article and are STALE — do not quote a current stage.
Agents in production: MEDIUM-STRONG. July 2025 shipped "MiiTel Synapse" = MiiTel Synapse Copilot + MiiTel Synapse Agent — voice-communication-specialised AI agents automating customer support. Synapse Agent was beta at announcement, so full production is partially aspirational; internal agent tooling is clearly live. Oct 2025 he stood up a new engineering unit, SDU (Strategic Development Unit), whose stated mission is productionising generative AI / voice-dialogue AI / LLM products fast.
Pain points: [evidenced, org-level, via his own tech blog] This is one of very few non-US eng orgs publishing real agent-ops methodology. Posts include "LLM-as-a-Judgeで「えこひいき」が起きる？— セルフバイアスを統計的に測る手法の紹介" (statistically measuring self-bias in LLM-as-a-Judge) and "AI エージェント開発で設計の一貫性をどう守るか" (maintaining design consistency in agent development). His team sent engineers to Datadog DASH specifically for AI monitoring. So: eval trustworthiness, agent design consistency, and agent observability are all demonstrably live concerns. [inferred] Cross-language/voice-agent latency and cost at MiiTel's call volume.
Challenges: Trusting your own evals (he is publishing on judge bias — that is a team that has been burned by bad evals); keeping architecture consistent as agent count grows; monitoring agents in a latency-critical voice product.
Must-haves: Trustworthy eval/scoring that isn't self-biased; agent observability and monitoring tooling; consistent design/config patterns across a growing agent fleet.
Nice-to-haves: Per-call cost attribution for voice agents; regression detection across model upgrades.
ICP confidence: High on title/size/engineering voice — he personally writes the eng year-in-review and the org publishes genuine agent-ops content. Medium on "agents fully in production" (Synapse Agent was beta) and on funding stage (data is stale).
Takuya Yoshioka — Senior Director of Research, AssemblyAI
Signal: 4 (senior technical AI leader at a voice-AI / voice-agent company, found via LinkedIn currentCompany-facet People search). Profile: https://www.linkedin.com/in/ty274/ . Source: https://www.linkedin.com/search/results/people/?currentCompany=%5B%2218644094%22%5D (AssemblyAI). Company size: 51-200 (LinkedIn). Pain points (inferred from role + company): inference/model cost of realtime voice AI at production scale; latency + reliability of streaming voice pipelines that power customer voice agents; per-run cost of always-on voice sessions. Challenges: keeping realtime model quality high while controlling GPU/inference spend. Must-haves: visibility into cost per session/run and reliability guardrails. Nice-to-haves: routing between model sizes by workload. ICP confidence: Medium-High (Senior Director of Research; AssemblyAI is voice-AI infra shipping a Voice Agent API - agent-adjacent; 2nd-degree connection; verified this run). Pain points INFERRED from role+company, not fabricated.
Tal Shapira — Co-Founder & CTO, Reco (Reco AI)
Signal: 4 (ICP technical co-founder/CTO at an agent-native company). Source: https://www.linkedin.com/in/tal-shapira/ ; discovered via agent-company funding research + LinkedIn people-search verification, 2026-07-27 run. Company size: ~170 employees (Apr 2026; ~100 in early 2026), split Israel/US; Series B, $85M total (30M Series B led by Zeev Ventures). Reco = agentic AI security for SaaS ('SaaS Security Automation with AI Agents' / 'Agentic AI Security for SaaS'); ships AI agents that discover and secure agent+SaaS activity. CTO is Ph.D. (EE, Tel Aviv), ex-head of a cyber R&D group. Pain points (INFERRED from role+company, not verbatim): observing and controlling autonomous agents operating across SaaS (shadow AI, agent-to-agent); cost + reliability of security agents at scale; visibility into what agents do and cost. Challenges: authorization at the tool-call level for agents; scaling many agents safely. Must-haves: agent observability + in-path control/authorization. Nice-to-haves: cost-per-run metrics; vendor-neutral control plane. ICP confidence: High (technical co-founder/CTO; Series B; ~170 emp; agent-native security).
Tamal Biswas — VP of Cloud Platform and Infrastructure, Calix
Signal: 1 (ICP authoring about agent cost blowout and consumption control). Source: LinkedIn post 2w before 2026-09-03, "Your AI Bill Is Not a Pricing Problem. It Is a Consumption Problem." (18 reactions) — surfaced via content search for "cost per agent"/"per-agent cost". Profile: https://www.linkedin.com/in/tamalbiswas/ (headline: "VP of Cloud Platform and Infrastructure @ Calix | Building AI-Native Data & Cloud Platforms at Scale | Ex-Salesforce, Adobe, GE, Intel"). Company size: 1,926 employees as of March 2026 (Revelio Labs / stockanalysis.com, +13.9% YoY from 1,676) — inside the 50–2,000 band but close to the ceiling and growing, so re-verify next run. Stage: public (NYSE: CALX), not Series A–C — noted as a deviation from the stated ICP stage filter. Agent activity: Calix already has an "Agentic Platform & ML Frameworks" engineering function (SN Raju Kattari, Director of Engineering — Agentic Platform & ML Frameworks, already in this People Library), which corroborates active agent build-out; Tamal owns the cloud platform/infra layer those agents run on. Pain points (his words): "Your AI vendor is getting cheaper. Your AI bill is getting bigger"; cites Gartner projecting "inference costs per agentic workflow will rise more than fivefold through 2028"; EY's $0.04 linear workflow vs ~$1.20 agentic workflow; "most AI cost programs are aimed at the wrong target... negotiating a 15% vendor discount while consumption triples". Challenges: consumption growth outrunning unit-price declines; organising cost levers into an operating model (how much are we buying / what per unit / when / who's allowed to buy); getting teams to pull the compounding levers early. Must-haves: consumption-side controls rather than procurement-side discounts; caching, context reduction and batch scheduling instrumented and measured; visibility into which agentic workflows drive spend. Nice-to-haves: entitlement/approval model for who can spend on agents; forecast of inference cost per workflow over time. ICP confidence: Medium-High — VP-level platform owner at a company actively building an agentic platform, writing unprompted and specifically about agent cost per workflow; downgraded from High only because Calix is a public networking/broadband SaaS rather than an AI-native Series A–C company and headcount sits near the ceiling.
Signal: 4 (senior technical/product leader at a qualifying agent company; found via web verification of Glean leadership). Source: https://www.businesswire.com/news/home/20240326392278/en/ ; https://www.glean.com/about. Company size: ~900 employees; agentic enterprise "Work AI" platform (Glean Agents / assistant) deployed in production across enterprises; ~$7B+ valuation. Role oversees ALL R&D — global engineering, product, design, data science (ex-Google VP, ex-Slack CPO). Pain points (role-inferred): cost of agentic runs at enterprise scale across many customers; reliability + governance of agents in production; scaling the agent platform. Challenges: keeping per-run agent cost and reliability predictable as enterprise agent usage scales; governance/observability at platform scale. Must-haves: cost-per-run visibility; production reliability + governance. Nice-to-haves: inference optimization / routing across the platform. ICP confidence: High (President of Product & Technology — above VP, owns engineering + AI — at an agent-native 50-2,000-emp company). NOTE: pain points inferred from role/company context, not a verbatim quote.
Tanmai Gopal — Co-founder & CEO (technical co-founder; also co-founder of Hasura), PromptQL (Hasura)
Signal: 1 (ICP writing publicly about token waste, context and agent control — the single closest message-market fit found this run). Title verified live on LinkedIn people search 2026-09-02: "Co-founder & CEO at PromptQL | Building Team AI to manage shared context | For AI native teams - Fortune 500s to Silicon Valley", United States.
Source: https://hasura.io/blog/@tanmaig — post "Stop tokenmaxxing. Start contextmaxxing." (23 Jun 2026), verbatim: "The alpha has shifted from using AI to maintaining useful context." Also "43% accuracy with Opus-4.6 & friends - will Text-to-SQL ever be good enough?" (25 Mar 2026), citing Berkeley DAB scores (Opus-4.6 43%, Gemini-3-Pro 38%, GPT-5.2 25%).
Company size: ~70 ("The team has 70 people", eesel AI PromptQL teardown, Feb 2026, https://www.eesel.ai/blog/promptql). Bengaluru + San Francisco. Series B $25M / Series C $100M.
Pain points: He is publicly arguing that token spend is the wrong optimisation target and context quality is the real one — thealpha.ai's exact wedge, in his own words. PromptQL's core architectural bet is that the LLM only writes a query plan which is then executed as real code outside the model, i.e. determinism over raw token throughput. He also publishes hard accuracy numbers showing frontier models under 50% on realistic agent tasks.
Challenges: proving agent reliability to enterprise buyers with hard numbers; context bloat and token waste; multi-model spend on a 70-person team.
Must-haves: token-waste and context-window visibility; repeatable eval evidence he can put in front of Fortune 500 buyers.
Nice-to-haves: per-customer cost attribution for their own multi-tenant agent product; model routing.
ICP confidence: Medium — technical co-founder (valid C-suite target), AI-native, Series B/C, and the strongest public pain signal in this run. Discount: ~70 employees sits right at the bottom of the 50-2,000 band, and as an agent-platform vendor himself he may build rather than buy. Net-new name AND net-new company.
Tao "Tony" Tong — Chief Technology Officer & Chief AI Officer, Auditoria.AI
Signal: 4 (ICP at a company shipping finance agents in production; company publicly articulating agent autonomy/governance maturity gaps). Source: https://blog.auditoria.ai/governed-autonomy-cfos-ai-agents ; funding: https://www.auditoria.ai/pr-auditoria-ai-raises-38m-in-series-b-funding-to-usher-agentic-ai-era-for-enterprise-finance-teams/
Company size: ~108 employees across 5 continents (Jan 2026) — in range. Series B, $38M. Agentic AI for enterprise finance (AP/AR/collections).
Pain points: Gap between claimed and actual agent autonomy, and the absence of end-to-end audit trails. Company framing: "the marketing decks all claim Stage 4, and the actual product capability sits somewhere between Stage 2 and Stage 3"; buyers should demand "the audit trail for a single transaction, end to end" and "a named customer running at Stage 4 in production for at least 12 months." Note: this quote is attributed to co-founder/CEO Rohit Gupta, so treat as company-level evidence rather than a personal Tong quote.
Challenges: Running AP agents against multi-entity finance postures (classify, code, PO-match, route exceptions) where every step must reconcile; proving sustained production autonomy over long horizons.
Must-haves: End-to-end transaction-level audit trail across agent steps; demonstrable production reliability over months not demos.
Nice-to-haves: Cost-per-invoice / cost-per-agent-run attribution; autonomy-level scoring per workflow.
ICP confidence: Medium-High — CTO/Chief-AI title, headcount, Series B stage and named production customers (Boddie Noell Enterprises, Otto Car Ltd) all verified; pain quote is company/CEO-level rather than personal.
Ted Nielsen — Chief Technology and Product Officer, Bidgely
Signal: 4 (ICP at a company where agentic AI is the explicitly stated technology strategy). NOTE: LinkedIn/Chrome unavailable this run — found via BusinessWire. Promoted to CTPO July 2026.
Source: https://finance.yahoo.com/energy/articles/bidgely-builds-executive-leadership-team-152900501.html (BusinessWire, 14 July 2026 — "the promotion of Ted Nielsen to Chief Technology and Product Officer")
Company size: ~290 (Tracxn, as of June 2026). Scale corroborated: serves 50M+ homes, 35+ energy providers. Los Altos, CA. Vertical AI for electric/gas utilities (UtilityAI platform). Funding: ~$120M+ raised, acquired Grid4C. CAVEAT: round letter NOT confirmed from a fetched page — treat "Series C / late-stage private" as approximate. This is the weakest field in this record.
Agents in production: MEDIUM-STRONG at company level. His promotion is explicitly framed as "Bidgely's deliberate strategy to unify its technology and product roadmap under a single leader as the company deepens its investment in generative and agentic AI capabilities." Separate BusinessWire coverage (Jan 2026, "Bidgely Redefines Energy AI in 2025: From Machine Learning to Agentic AI") describes autonomous agents executing workflows across billing, service and grid operations, and UtilityAI Pro as a platform that HOSTS a utility's own agents and copilots. NOT confirmed: agents running autonomously in production at named customers.
Pain points: [not stated by him — no personal quote found] [evidenced, org-level] Consolidating tech + product under one leader specifically to push agentic AI; new COO Creighton Oyler (ex-Oracle Energy & Utilities GM) said "Utilities are done experimenting with AI pilots — they're moving to enterprise-wide deployment" — i.e. pilot-to-production is the stated company-level pressure. [inferred] Running agents inside customers' own data environments (AWS/Snowflake/Databricks) plus third-party apps creates a real orchestration, observability and cost-attribution problem — especially since UtilityAI Pro hosts the UTILITY'S agents, meaning he inherits multi-tenant agent operations.
Challenges: Moving 35+ utility customers from pilots to enterprise-wide agent deployment; operating agents inside customer-controlled data environments; multi-tenant agent hosting (UtilityAI Pro) with per-customer cost and reliability accountability; regulated-utility scrutiny of autonomous grid/billing actions.
Must-haves: Multi-tenant agent observability and per-customer cost attribution; pilot-to-production reliability evidence for regulated utility buyers; orchestration across customer cloud environments.
Nice-to-haves: Self-serve agent cost dashboards for utility customers; benchmarking agent performance across the 35+ provider base.
ICP confidence: Medium — clean C-level technical title, right size (~290), right vertical (energy/utilities), and agentic AI is an explicitly stated company strategy. Downgraded because there is no personal statement from him, no confirmation of agents running autonomously in production at customers, and the funding stage is unverified.
Tejas Manohar — Co-Founder & Co-CEO, Hightouch
Signal: 4 (ICP technical co-founder shipping agents in production). Source: https://techcrunch.com/2026/04/15/hightouch-reaches-100m-arr-fueled-by-marketing-tools-powered-by-ai/ ; https://hightouch.com/blog/hightouch-agents-ai-for-marketers ; https://www.businesswire.com/news/home/20251112501540/en/Hightouch-Introduces-New-Agents-Platform-to-10x-Marketer-Productivity. Company size: ~380 employees (Apr 2026); $100M ARR; Series D $2.75B valuation. Ex-Segment engineering manager (technical). Hightouch Agents = an AI platform running always-on marketing agents (audience research, on-brand creative generation, campaign execution across ads/email/SMS/web) — 5+ agents in production. Pain points (inferred): orchestrating many always-on autonomous marketing agents reliably; controlling agent behavior and brand safety at scale; agent run cost as usage scales. Challenges: reliability + governance of autonomous agents acting in live marketing systems. Must-haves: control/observability over agent actions, production reliability. Nice-to-haves: per-run cost visibility. ICP confidence: High (technical co-founder/co-CEO at a 380-person agentic-AI company shipping agents in production).
Tejaswi Tenneti — Head of AI/ML, Ambience Healthcare
Signal: 4 (ICP AI leader at an AI-native company shipping clinical AI agents; found via LinkedIn people search "Ambience Healthcare OR Abridge director of AI engineering agents" this run).
Source: https://www.linkedin.com/search/results/people/?keywords=Ambience%20Healthcare%20OR%20Abridge%20director%20of%20AI%20engineering%20agents — listed as "Head of AI/ML, Ambience Healthcare", San Carlos CA, ~7K followers, company actively hiring ("See open roles").
Company size: 201–500 employees (LinkedIn company page, verified this run). Ambience Healthcare = AI-native ambient clinical documentation / clinical AI company, Series C ($243M led by Oak HC/FT + a16z + OpenAI Startup Fund), running AI agents across documentation, coding and clinical decision support in production at health-system scale.
Pain points: NO public pain post captured this run — do not treat as expressed. Qualification is role + company based.
Challenges (inferred from role scope, flagged as inference): owning AI/ML for multi-agent clinical workflows in a regulated setting implies eval/regression load, per-encounter inference cost, and accuracy-vs-cost tradeoffs at health-system volume.
Must-haves: unverified this run.
Nice-to-haves: unverified this run.
ICP confidence: High — Head of AI (above the ICP Director floor), AI-native Series C company at 201–500 employees with a documented multi-agent production footprint. Note: Brendan Fortuner (Head of Engineering) and Nikhil Buduma (Co-founder & CEO) are already in the People Library for this account — Tejaswi is a net-new second entry point.
Thejas Bhat — Director, Data Engineering and AI, Uniphore
Signal: 4 (ICP senior technical leader at an agent-native company; discovered via LinkedIn people-search of companies actively shipping AI agents) | Source: https://www.linkedin.com/in/thejashr | Headline: Director, Data Engineering and AI at Uniphore | Company: Uniphore — actively building/shipping AI agents in production | Company size: ~800–1,000 (estimate) | Pain points (INFERRED from role/company context — not a verbatim quote observed this run): cost/efficiency of AI + data pipelines feeding agents; reliability and observability of agent outputs in production | Challenges: controlling inference/token cost while keeping agent data pipelines reliable at scale | Must-haves: cost-per-run visibility; observability | Nice-to-haves: context/token-waste optimization | ICP confidence: Medium-High (senior technical/eng leadership at a 50–2,000-employee company shipping AI agents)
Thiago Scalone — Partner & Director (Engineering) — previously Head of Engineering / Director of Engineering, CloudWalk
Signal: 1 (ICP company publicly writing about agent/token economics). Title verified live on LinkedIn people search 2026-09-02: "Partner And Director at CloudWalk, Inc.", Brazil. NOTE: theorg.com lists him as CTO — LinkedIn is the authoritative read and says Partner & Director, so treat "CTO" as unconfirmed.
Source: https://www.businesswire.com/news/home/20260311778452/en/ and https://www.businesswire.com/news/home/20260513790695/en/ ; https://theorg.com/org/cloudwalk/org-chart/thiago-scalone
Company size: 720 (CloudWalk 2025 results press release, verbatim: "$1.8 million in revenue per employee with a lean team of 720 people"). São Paulo, Brazil. AI-native fintech (InfinitePay / Pierre.finance / JIM.com), 7M MAU.
Pain points: Company press states CloudWalk processes "in excess of 60 billion tokens per day in production" with "dozens of agents" across ops, marketing, sales and engineering. CEO Luis Silva: "AI is not a feature we added to Cloudwalk - it is how the business runs." They built their own GPU cluster explicitly for "a structural cost advantage in inference economics" and publicise >90% GPU utilisation — inference cost is already a board-level metric, which means cost-per-agent-run attribution is a live, funded problem.
Challenges: attributing cost per agent and per run across three products and dozens of agents; reliability of autonomous credit/fraud/support agents inside a Brazilian Central-Bank-regulated product; balancing owned GPU capacity against frontier-API burst.
Must-haves: per-run and per-agent cost attribution; support for self-hosted models alongside multiple providers; audit trail suitable for regulated finance.
Nice-to-haves: routing/fallback between own cluster and frontier APIs; per-agent SLOs.
ICP confidence: High — Director+ (verified), 720 employees (in band), AI-native, dozens of agents in production, and token economics already public and explicit. Net-new name AND net-new company vs the 1,071-record library.
Thiyagaraj T — Director, Engineering, Eightfold AI
Signal: 1 (ICP quoted publicly on agent cost governance). Found via web research; title VERIFIED live on LinkedIn people search 2026-08-31 — result reads "Thiyagaraj T (thiyaga) — Director, Engineering at Eightfold AI, Bengaluru". This RESOLVES a title conflict: an earlier secondary source listed him as "Engineering Manager"; LinkedIn confirms Director. Profile URL came from a search-result listing and was not href-verified directly — re-check before outreach.
Source: https://inc42.com/features/enterprise-ai-spend-gets-a-reality-check/
Company size: ~1,000-1,500 (Glassdoor / Inc42 company profile). Top of the 50-2,000 band.
Pain points: Uncontrolled AI spend at feature level. Runs a pre-launch gate — no feature may use an AI model until a team states what it's building, which model, and the projected monthly cost, with platform owners challenging the estimate first. One projected cost went from $2,000 to $400 (80% cut) before launch.
Challenges: Runtime enforcement — per-team monthly budget caps, every call logged and compared against the original estimate. That reconciliation is being done manually/in-house.
Must-haves: Per-team budget caps with enforcement; logged calls reconciled against forecast.
Nice-to-haves: Automated estimate-vs-actual drift alerting; model-choice recommendations at design time.
ICP confidence: Medium-High — Director-level confirmed, in band, and near-perfect cost-governance evidence. Downgraded from High because Eightfold is late-stage (not Series A-C) and sits at the top of the headcount range.
Thomas Kinsella — Co-founder & Chief Customer Officer, Tines
Signal: 1 (ICP-adjacent founder writing first-hand about agent consumption cost). Source: https://www.tines.com/blog/building-an-ai-soc-with-tines/ (authored by him). Company size: 548 as of 31 Mar 2026 (Tracxn). Stage: Series C, $125M (per Tines' own announcement post). Dublin, Ireland. Pain points: Kinsella describes building an internal AI SOC prototype in Tines' own production tenant and lists five lessons, of which lesson 5 is "Consumption control matters" — regardless of which model is used, AI comes at a cost, and they deliberately evaluated which NON-AI deterministic workflows performed just as well as AI-enabled ones specifically to find where they could cut cost. He also notes generative AI tools scored lowest in satisfaction in the 2025 SANS SOC Survey. Challenges: choosing autonomous vs. deterministic execution paths per workflow step; proving the cost-efficiency of agentic steps against a deterministic baseline; governance and audit for agent actions in a security context. Must-haves: cost attribution per workflow and per agent step, an AI-vs-non-AI cost-to-value comparison, audit trails on agent actions. Nice-to-haves: consumption/budget dashboards, per-tenant cost breakdown. ICP confidence: Medium (company, Series C stage, 548 headcount and the cost-consciousness signal are all verified, and he is a co-founder with a first-person published account of agent consumption cost — BUT Chief Customer Officer is NOT a match to the ICP title bucket, so he is a champion/entry point rather than the technical buyer). NOTE: Tines already has Eoin Hinchy (Co-Founder & CEO) in the library; Kinsella is net-new. Best use: route to whoever owns Tines' agent platform engineering.
Signal: 4 (ICP senior AI leader at a company shipping AI agents). Source: https://www.linkedin.com/in/thijs-waanders (found via NiCE Cognigy People directory). Company size: 201-500 employees (verified); Cognigy ships contact-center AI agents. Pain points (inferred from VP AI Transformation remit): scaling customer agent deployments reliably from pilot to production, cost/ROI visibility on agent programs, moving from 1 to many agents. Challenges: proving reliability and controlling spend as enterprise agent rollouts scale. Must-haves: per-run cost + reliability visibility across deployed agents. Nice-to-haves: benchmarking / cost attribution across customers. ICP confidence: Medium-High (VP-level AI leadership at a verified agent company; role is partly customer-facing).
Tim Armandpour — Chief Technology Officer, PagerDuty
Signal: 4 (ICP at a company shipping agents in production; quoted in press on agent operations) + 3 (company uses Arize and LangSmith for agent observability).
Source: name, title and LinkedIn hyperlink verified on https://www.pagerduty.com/leadership/ ; quoted alongside Chief AI Officer João Freitas in https://www.techjournal.uk/p/pagerdutys-ai-chief-says-outage-response (10 Jul 2026). Agent evidence: https://www.pagerduty.com/blog/ai/meet-your-virtual-responder-pagerdutys-sre-agent-for-ai-driven-reliability/ and Mar 2026 press release announcing 30+ AI partners including Anthropic, Cursor and LangChain.
Company size: 1,155 employees as of 31 Jan 2026 (S&P Global / FY2026 10-K via stockanalysis.com/stocks/pd/employees), down 7% YoY.
Pain points: No verbatim quote from Armandpour captured in this run — he appears in the same July 2026 TechJournal piece where the agent-reliability framing is laid out, but the quotes verified were Freitas's. Company-level pain: agent non-determinism and error propagation across agent-to-agent chains; the need for deterministic workflow fallbacks around non-deterministic agents.
Challenges: As CTO he owns the technology bet on three production agents (Insights, Shift, SRE) shipped into customer incident response, plus a 30+ partner AI ecosystem, against a shrinking engineering org.
Must-haves: A defensible reliability story for agents operating in customers' highest-stakes workflows; vendor/partner strategy for the agent stack rather than building everything in-house.
Nice-to-haves: Consolidated cost and performance view across the agent portfolio; reduced build-vs-buy burden on the platform team.
ICP confidence: Medium-High — CTO title verified on the company's own leadership page with a real LinkedIn link, headcount verified, agents demonstrably in production. Downgraded only because no personal pain quote was captured. He is the exec-sponsor path; João Freitas (Chief AI Officer, ID 1003) is the champion and the better first touch.
Signal: 4 (ICP technical decision-maker at a company actively shipping agents in production). Source: https://theorg.com/org/cresta/org-chart/tim-shi and https://cresta.com/llm-info. Company size: ~693 employees (Palo Alto; contact-center AI; deploys production LLMs to Fortune 500 since 2018; $100M+ revenue). Actively ships AI agents — Cresta Virtual Agent, Agent Copilot, and Cresta Conductor (build/optimize conversational agents from discovery through post-launch). Pain points: reliability of agents in live enterprise contact centers, shipping agents "with confidence," post-launch optimization at scale. Challenges: taking agents from prototype to reliable production across many enterprise tenants; measuring/optimizing agent quality and cost per interaction. Must-haves: production reliability, observability into what agents do per run, per-tenant control. Nice-to-haves: cheaper model routing for low-risk turns, faster build/optimize loop. ICP confidence: High (ex-OpenAI MTS; clear technical decision-maker; company squarely in size band and shipping multiple agents in production).
Tim Smith — Co-Founder & Chief Technology Officer, Medable
Signal: 4 (ICP technical co-founder shipping a staged agent fleet). Source: https://www.businesswire.com/news/home/20260615413055/en/Medable-Introduces-Digital-Data-Flow-Agent-Laying-the-Foundation-for-End-to-End-Agentic-Clinical-Development ; https://www.medable.com/newsroom/medable-launches-the-industrys-first-agentic-ai-platform-and-cra-agent-removes-bottlenecks-in-clinical-development ; https://www.medable.com/platform/agent-studio ; https://www.mobihealthnews.com/news/medable-launches-agentic-ai-platform-clinical-development ; LinkedIn: https://www.linkedin.com/in/tsmith11/ . Company size: ~352 (Jun-2026, Unify/LeadIQ) to ~501 (Dec-2025, other sources) - in band on both figures. Decentralised clinical trial platform for pharma/life sciences.
Pain points: INFERRED (no 2026 verbatim pain quote found; only an older Studio-launch quote about "eliminate one of the biggest hurdles to broad adoption"). Agent Studio is a NO-CODE agent builder handed to clinical (non-engineering) staff - meaning agent count and token spend grow outside engineering's control, which is the classic cost-blowout and governance shape.
Challenges: staged multi-agent rollout - Agent Studio platform, then CRA Agent (trial monitoring, runs without continuous human prompting), then Digital Data Flow Agent (15 Jun 2026, protocol -> CDISC USDM 4.0 JSON), plus an Agentic Accelerator Program pushing customers to deploy their own agents across the clinical lifecycle. Every agent runs on GxP/FDA-regulated trial data where a wrong action is a regulatory event.
Must-haves: audit trail and human accountability per agent action (their own materials stress qualified humans remain accountable); reliability on long-running unattended monitoring agents.
Nice-to-haves: per-customer/per-trial cost visibility as customers self-build agents in Studio.
ICP confidence: HIGH. Technical co-founder, headcount confirmed in band from two independent sources, three named agents plus a customer-facing agent-building platform.
Angle: the strongest untapped pain here is customer-built agents in Studio - Medable carries the runtime cost and the blame for other people's agents.
Timothée Lacroix — Co-Founder & CTO, Mistral AI
Signal: 1 (CTO / research commentary on cost-efficient agentic production; SLMs as first-class citizens; ARM agentic reasoning)
Source: https://mistral.ai/about/ ; https://www.neonriver.com/timothee-lacroix-2025/
Company size: AI-native, ~hundreds of employees (within 50-2000); caveat: late-stage funding
Pain points: cost-efficiency of agentic production systems; right-sizing models for agent tasks
Challenges: making agentic reasoning reliable AND affordable in production
Must-haves: efficient model routing/sizing for agents; reliable agentic reasoning
Nice-to-haves: observability across agentic workflows
ICP confidence: Medium (CTO building agents at right size; caveat: frontier lab may build in-house)
Signal: 1 (ICP publicly writing about agentic architecture / building agents on structured business data)
Source: https://www.nue.io/company/press/agentic-revenue-architecture/ ; https://www.unite.ai/author/tinakung/
Company size: 51-100 employees (Series A, Inovia-led)
Pain points: making AI agents reason/recommend/execute reliably on structured revenue data; secure, auditable agent execution
Challenges: trustworthy autonomous execution on financial data; adoption/usability of agentic systems in revenue ops
Must-haves: auditable & secure agent execution environment; context-awareness across the revenue stack
Nice-to-haves: agents that adapt to each customer's revenue framework
ICP confidence: High (CTO at 51-100 emp Series A SaaS actively shipping production agents)
Signal: 4 (ICP technical/product leader at a qualifying agent company; found via LinkedIn people search + web verification). Source: https://theorg.com/org/crescendo-3/org-chart/tod-famous ; https://www.crescendo.ai/company. Company size: ~1,945 employees (AI-native, fully-managed CX contact center; passed $100M revenue; agents in production for enterprise CX). Prior: SVP/Head of Product Genesys Multicloud, VP PM Cisco contact center. Pain points (role-inferred from outcome-based-pricing CX agent model): per-resolution / per-run LLM cost directly hits gross margin because Crescendo guarantees outcomes; unpredictable agent cost erodes the outcome-priced model; production reliability of voice+chat agents across many enterprise clients. Challenges: keeping cost-per-resolution predictable while scaling agent volume; reliability/quality at enterprise scale. Must-haves: visibility into cost per run/resolution; production reliability guardrails. Nice-to-haves: model routing / inference-cost optimization to protect margin. ICP confidence: High (product C-suite at an agent-native 50-2,000-emp company; cost-per-outcome economics are core to their business). NOTE: pain points are inferred from role/company context, not a verbatim quote.
Signal: 1 (ICP speaking publicly on a named panel about moving agents from experiment to production). NOTE: LinkedIn/Chrome unavailable this run — LinkedIn URL surfaced in search results, not opened/verified.
Source: https://asugsvsummit.com/speakers/todd-tobin
Company size: ~190-302 (Tracxn 302 as of 31 May 2026; PitchBook 190). Scale corroborated by his ASU+GSV bio: "over 6 million educators across more than 13,000 schools and districts." Series B — $45M led by Valor Equity Partners (~$65M total). K-12 edtech AI.
Agents in production: CONFIRMED via panel + bio. He is a panelist on an ASU+GSV session literally titled "Agents in the Real World: From Experiments to Business Outcomes." His bio frames his job as "building the systems that give AI the right context, knowledge, and guardrails to deliver genuine outcomes in classrooms at scale." Works with Anthropic and AWS "to tackle the real-world challenges of deploying AI in education, where the stakes are high and the margin for error is low." Prior: co-founded Craftsy, led engineering at HomeAdvisor — 25 years of scaling.
Pain points: [evidenced] The panel title IS the pain — getting agents from experiment to business outcome. His own bio names context plumbing, knowledge, and guardrails as the systems he has to build, and explicitly flags low margin for error at classroom scale. [inferred] Per-run cost at 6M-educator volume; eval and regression discipline across model upgrades.
Challenges: Proving business outcomes (not demos) from agents; supplying correct context/knowledge to agents at 13,000-district scale; guardrails where the end user is a child and error tolerance is near zero; cost at very high free/low-ARPU user volume.
Must-haves: Context management and guardrail infrastructure; evidence of outcomes, not activity; cost per run at massive educator volume.
Nice-to-haves: Model-upgrade regression testing; per-district/per-feature cost attribution.
ICP confidence: High — CTO, Series B, ~200-300 employees, and on public record specifically about production agents vs. experiments. One of only a handful this run with a first-person/named-panel agent-ops voice.
Signal: 4 (ICP writing/speaking about shipping agents; conference speaker page + live LinkedIn verification 2026-09-02, headline "Head of Engineering at Linear", Brooklyn NY).
Source: https://2026.agentconference.com/speaker/Tom-Moor and https://2026.agentconference.com/schedule — panel "The Hidden Infrastructure Required to Scale AI Coding Agents" (alongside OpenAI's Codex deployment lead). LinkedIn profile URL href-verified from search results.
Company size: 51-200 (Linear LinkedIn company page, read live 2026-09-02; page also shows ~75% employee growth trend). Series B/C, US/remote-first.
Pain points: The panel title is literally the 1->5+ agents wall — what infrastructure you need once coding agents stop being a demo. Linear's own LinkedIn page now positions the product as "The product development system for teams and agents" and it shipped third-party app approvals so admins can gate what agents can reach — i.e. agent governance is already a live product surface for them, and internally they are running agents at volume.
Challenges: infrastructure to scale coding agents past one or two; visibility into what each agent run costs and whether it succeeded; access control for agents touching company data.
Must-haves: run-level observability and reliability signal for coding agents; per-agent identity and policy.
Nice-to-haves: cost caps per agent; cost-per-merged-PR style ROI reporting.
ICP confidence: High — exact title (Head of Engineering), 51-200 employees (in band), Series B/C, US, and public speaking evidence on precisely the scaling-agents pain. Net-new name (Linear already in library via Tuomas Artman).
Signal: 4 (ICP product leader; SaaStr AI Annual 2026 Deploy Summit kickoff on "what's actually working with AI agents in production, not what's possible").
Source: https://www.saastr.com/the-first-44-speakers-for-saastr-ai-annual-2026-the-founders-and-operators-actually-shipping-ai-at-scale/
Company size: ~840-915 employees (infra co; ships v0 agentic product + AI SDK powering thousands of AI apps).
Pain points: gap between what agents can demo vs what reliably works in production.
Challenges: taking agentic demos to dependable production behavior at scale.
Must-haves: production reliability + observability for agents.
Nice-to-haves: cost predictability across agent runs.
ICP confidence: Medium (AI-focused product exec with deep technical background (ex-React lead), agent-shipping company in size range).
Tomer Teller — VP Product, Zenity
Signal: 1 (ICP writing publicly about agent control/observability at scale).
Source: https://zenity.io/blog/agentic-ai-hype (15 Apr 2026); title verified on https://zenity.io/company and in https://www.businesswire.com/news/home/20260727514033/en/
Company size: 230+ worldwide (~150 in Israel) — verified in Zenity's own Series C release (https://zenity.io/company-overview/newsroom/company-news/zenity-raises-125-million-to-secure-the-era-of-1-billion-ai-agents) and Calcalist/CTech 3 Aug 2026. Series C, $125M raised Aug 2026.
Pain points: (verbatim) "human oversight of agents doesn't scale... That works when you have ten agents. It breaks down when there are hundreds, thousands." Also (verbatim) "They don't have full traceability of what the agent reasoned, what it decided, what it acted on. No session replay. No risk-scored evidence chain that holds up to scrutiny."
Challenges: Manual/human-in-the-loop oversight collapses past ~10 agents; no per-agent decision trace; evidence chains don't survive audit scrutiny.
Must-haves: Automated, machine-scale agent oversight; full reasoning/decision/action traceability; session replay; risk-scored evidence chain.
Nice-to-haves: Risk scoring tuned per agent; audit-ready export.
ICP confidence: High — exact ICP title (VP Product, AI-focused), verified headcount inside band, Series C, company ships agent-governance product and runs agents itself, and he has publicly articulated the 1→many scaling wall in Alpha's exact language.
Watch-out: Zenity sells AI agent security/governance — partial competitive overlap on the observability story. Wedge is cost/run economics, not security.
Tommi Holmgren — VP of Product (Agents and Automation), Sema4.ai
Signal: 4 (AI-focused VP of Product at an agent-platform company, amplifying production-agent content).
Source: https://www.linkedin.com/in/tholmgren/recent-activity/all/ — reposts of Sema4.ai content including a Snowflake Silicon Valley AI Hub developer workshop, "Building Production-Ready AI Agents on Snowflake: A Developer Deep Dive," covering agents built with Python, SQL and natural-language runbooks, zero-copy enterprise data access, and Snowpark Container Services / Cortex AI for "secure, scalable automation." Also amplifies Sema4.ai material on traditional data pipelines failing to keep up with the speed, scale and precision finance/business teams need. Company page: https://www.linkedin.com/company/sema4-ai/about/
Company size: 51-200 employees (LinkedIn company page, verified this run). Sema4.ai sells an enterprise platform to "build, run, and manage AI agents at scale" — i.e. the buyer-side of the same problem Alpha addresses.
Pain points: The content he amplifies centres on getting agents from workshop to production safely and on data pipelines that cannot keep up with agent-speed demands. No cost-per-run language observed.
Challenges: Making agents production-ready and governable for enterprise (finance) customers with runbook-defined behaviour and controlled data access.
Must-haves: Production-readiness and secure/scalable execution as a product property, not a deployment afterthought.
Nice-to-haves: Developer-friendly authoring (natural-language runbooks, SQL/Python parity).
ICP confidence: Medium — VP of Product with an explicitly AI/agent remit (in ICP), 51-200 employee agent-platform company. Downgraded from High because Sema4.ai is itself an agent-management platform vendor, so it is as likely a competitor/partner as a customer; qualify that before treating as a prospect.
Tony Lee — Chief Technology Officer (leads ML, engineering, product, design), Hyperscience
Signal: Bucket 1 (ICP CTO at agent company) — web research pivot (LinkedIn/Chrome unavailable). Source: https://www.hyperscience.ai/newsroom/from-idp-to-intelligent-inference-spring-2026-release/ ; https://www.hyperscience.ai/about-us/. Company size: ~250-400 employees. Hyperscience's Hypercell Spring 2026 release moves the platform from IDP to "agentic enterprise" — a unified orchestration layer for AI agents that "intelligently balances accuracy with cost." Tony Lee (CTO) leads product, ML, engineering and design. Pain points: explicitly balancing agent accuracy vs. cost; trusted data foundation and high-performance infra for agents; transition from simple automation to agentic. Challenges: orchestrating reliable document/inference agents at enterprise scale while controlling cost. Must-haves: accuracy-vs-cost per-run visibility, reliability/orchestration observability, guardrails. Nice-to-haves: unified cost/latency dashboards across agent workflows. ICP confidence: High — CTO owning ML+engineering at a 50-2,000 emp company shipping agentic platform. (Note: Hyperscience's Peter Levi, Director of Engineering, already in library; Tony Lee is net-new.)
Signal: 4 (adapted method — found via LinkedIn company People-page directory of an in-band agent-native company; LinkedIn content/post search was low-yield this run, consistent with prior runs). Source: https://www.linkedin.com/company/perplexity-ai/people/ and https://www.linkedin.com/in/tonygwu/ (prev. OpenAI). Company size: 201-500 employees (LinkedIn header). In-band, agent-native (ships answer engine, Comet browser agent, Deep Research agent in production). Pain points (INFERRED from role+company, no quotes): per-query/per-agent-run LLM cost at consumer scale (agentic answer engine + Comet agentic browser + Deep Research agent, millions of runs); token/context efficiency; reliability & citation accuracy of agent outputs; latency; near-zero per-run/per-agent cost visibility. Challenges: keeping unit economics sane as agent usage compounds; orchestrating multi-step search->browse->research agents reliably at scale. Must-haves: per-run/per-agent cost observability; reliability guardrails at massive scale. Nice-to-haves: model routing/cost optimization; agent governance. ICP confidence: High (VP Eng at in-band agent-native company).
Signal: 4 (ICP at a company publishing on agent reliability at scale). Source: LinkedIn people search verified live 2026-08-30 — "Tony Zhu, CTO and Co-founder of WIZ.AI, ASEAN'S BEST AI PARTNER & VOICE AI SOLUTION, Singapore, Current: CTO at WIZ.AI". Exact profile URL not captured (3rd-degree connection, URL not exposed in search results) — left blank rather than guessed; resolve before outreach. Company size: ~165 employees (Tracxn); Singapore; conversational/voice AI agents; Series B led by SMBC. Pain points: company reports "100M+ AI calls monthly" — extreme per-run volume where a fraction of a cent per call is a material line item. They publish a "Voice AI Reliability Benchmark" and a piece titled "From Pilot to Scale: Enterprise AI Agent Deployment", i.e. they are publicly positioned on exactly the pilot-to-production reliability wall in the ICP definition. Challenges: voice agents are the most latency- and cost-sensitive agent shape; scaling from pilot deployments to enterprise-wide across ASEAN telco and banking customers. Must-haves: per-call cost visibility at 100M+ monthly volume; reliability measurement they can put in front of enterprise buyers. Nice-to-haves: model routing/arbitrage across voice and LLM providers. ICP confidence: High — technical co-founder CTO, right size and stage, agents in production at very high volume, and the company's own content marketing is about agent reliability. NOTE: was previously discarded by an automated surname-only dedupe collision with "Kay Zhu | Genspark"; that was a false positive, this is a different person at a company not otherwise in the library.
Toshish Jawale — Head of AI Engineering (ex-CTO, Symbl.ai), Invoca
Signal: 4 — ICP-title technical AI leader at an agent-shipping company (found via people search: 'Head of Engineering agentic AI'). Source: https://www.linkedin.com/search/results/people/?keywords=Head%20of%20Engineering%20agentic%20AI%20startup . Company size: ~500-700 (est). Invoca = revenue-execution / conversation-intelligence SaaS; headline cites 'Agentic Systems, Conversational AI, Multi-Modality, Speech'. Pain points (INFERRED): scaling agentic + speech/conversational AI in production; multi-modal agent reliability; cost. Challenges: production agentic systems at scale. Must-haves: reliability + cost control for agents. ICP confidence: Medium-High — Head of AI Eng and former startup CTO at in-range agent company; pains inferred from role.
Signal: 4 (ICP technical co-founder/CTO at agent-native company). NOTE: Found via web research (funding/press + company sources); LinkedIn/Chrome was NOT connected this run, so prescribed LinkedIn searches were unavailable. Source: https://techcrunch.com/2026/03/10/mandiants-founder-just-raised-190m-for-his-autonomous-ai-agent-security-startup/ and https://www.prnewswire.com/news-releases/armadin-secures-record-breaking-189-9m-in-seed-and-series-a-funding-to-combat-the-era-of-ai-driven-hyperattacks-302709318.html and https://www.securityweek.com/kevin-mandias-armadin-launches-with-189-9-million-in-funding/. Company size: ~60+ (reported hiring 60+ employees in first 6 months; Kevin Mandia's/ex-Mandiant startup). Stage: Series A ($189.9M combined Seed+Series A led by Accel, w/ GV, Kleiner Perkins, Menlo, In-Q-Tel — described as largest combined Seed+A in cybersecurity history). Actively building AI agents: creates and manages autonomous AI security agents that continuously scan for threats; already working with Fortune 100 companies. Pain points (inferred from domain/market positioning, NOT a verbatim quote): reliability and control of autonomous security agents in production; scaling agent fleets to counter AI-driven 'hyperattacks'; visibility into agent behavior at enterprise scale. Challenges: making autonomous agents trustworthy in high-stakes SOC work; orchestrating many agents reliably. Must-haves: reliability, control, observability of agents in production. Nice-to-haves: cost-per-run visibility. ICP confidence: High (CTO, agent-native security company, ~60 emp in-band, Series A, autonomous agents in production).
Travis Rehl — CTO, Innovative Solutions
Signal: 3 (ICP engaging with competitor/adjacent inference-vendor content — Fireworks AI customer story).
Source: https://fireworks.ai/blog/innovative-solutions
Company size: ~780 employees (RocketReach/ZoomInfo, 2026). AWS Premier Partner / cloud + AI services firm, Rochester NY, US. Running 4–10B tokens/month across multi-agent workloads.
Pain points: (1) "Our number one COGS is AI cost" — inference spend is the single largest cost of goods sold; (2) costs scaling linearly with usage and acquisitions, compressing margin; (3) "We're spending a ton of time around the unit economics of multi-agent systems" — engineering time consumed by cost modelling rather than product.
Challenges: keeping per-customer margin intact as multi-agent token volume grows into the billions per month; attributing spend to specific agents/customers; deciding model routing between frontier and cheaper/open models.
Must-haves: per-agent and per-customer cost attribution; unit-economics visibility on multi-agent runs; levers to cut inference cost without degrading output quality.
Nice-to-haves: automated model routing; forecasting of agent spend ahead of the invoice.
ICP confidence: High — CTO title, headcount squarely in range, and the most explicit agent-cost-blowout quote found this run. Caveat: company is a services/MSP firm rather than a Series A–C SaaS product company, so the buying motion may differ. LinkedIn URL not found in search results — not guessed.
Verification note: headcount from third-party aggregators only; not confirmed against a first-party source.
Trey Holterman — Co-founder & CEO (technical; Stanford CS), Tennr
Signal: 4 (technical co-founder/CEO at a company shipping AI agents in production; web research this run). Source: https://techfundingnews.com/top-10-us-ai-agents-2026-fastest-scaling-category-52b-by-2030/ ; https://fortune.com/2025/06/18/tennr-health-tech-ai-patient-referral-ivp-a16z-lightspeed-iconiq-series-c/ ; https://theorg.com/org/tennr . Company size: 205 employees (NYC; founded 2021; Series C $101M at $605M valuation, led by IVP with a16z, GV, Lightspeed, ICONIQ). Product: RaeLM, a vision-language agent that reads unstructured patient records and generates payer-compliant documentation/prior-auth to maximize first-pass approval — agents in production across healthcare providers. Pain points (inferred from product; no verbatim quote this run): reliability/accuracy of document agents on messy clinical inputs, first-pass approval rates, scaling agent throughput across payers. Challenges: production-grade accuracy on high-stakes healthcare docs; scaling agents across many payer rule-sets. Must-haves: reliability/accuracy, auditability. Nice-to-haves: per-run cost visibility, throughput scaling. ICP confidence: Medium-High (technical co-founder/CEO at a 205-emp Series C company shipping agents in production; C-suite technical founder valid per ICP). Co-founder Tyler Johnson (CTO) is already in the People Library. Profile URL not captured — do not fabricate.
Tsung-Hsien Wen — Co-founder & CTO, PolyAI
Signal: 4 (ICP CTO at a qualifying company shipping agents in production at scale). Source: https://app.dealroom.co/news/note/polyai-ceo-nikola-mrk-i-on-scaling-enterprise-voice-agents-from-science-fiction-sales-to-50m-arr-and-the-future-of-ai-contact-centers and https://www.cmswire.com/customer-experience/polyai-raises-86m-series-d-for-enterprise-voice-ai/ . Company size: est. ~250-350 (London; Series D Dec 2025 $86M at $750M valuation, NVIDIA NVentures participating; ~$50M ARR). 100+ enterprise customers (Marriott, Caesars, UniCredit, PG&E), 2,000+ live deployments across 45 languages / 25+ countries. Pain points: cost scaling on per-minute basis as call volume grows (more agent calls = higher cost); production reliability of voice agents across many languages/deployments. Challenges: managing inference/cost economics at 2,000+ live deployments; consistent reliability at global scale; margin as usage scales. Must-haves: cost control/visibility per deployment; reliability at scale. Nice-to-haves: routing/model-selection efficiency across languages. ICP confidence: Medium — technical co-founder/CTO decision-maker at a ~300-person AI-native voice-agent company; strongly in ICP on stage/size, though public cost/reliability commentary is via CEO not CTO directly.
Tumas Rackaitis — Co-founder & CTO, Rogo
Signal: 4 (ICP technical co-founder building & shipping agents in production at scale). Source: https://www.prnewswire.com/news-releases/rogo-raises-160m-series-d-to-scale-the-agentic-platform-for-finance-302756546.html , https://siliconangle.com/2026/04/29/rogo-raises-160m-speed-financial-analysis-ai-agents/ , https://openai.com/index/rogo/ , https://aws.amazon.com/startups/learn/rogo-delivers-secure-ai-with-amazon-bedrock..., https://tracxn.com/d/companies/rogo/. Company size: 121 employees (as of Mar 2026). Funding: Series D $160M (Apr 2026, Kleiner Perkins/Sequoia/Thrive; $300M+ total) — NOTE: later stage than the A–C sweet spot, but headcount (121) and agent-building squarely fit ICP. Secure enterprise agentic AI platform for finance automating research, financial modeling and data-intensive workflows; used by 35,000+ professionals at 250+ institutions (Rothschild, Jefferies, Lazard, Moelis, Nomura). Runs on OpenAI o1 + Amazon Bedrock. Rackaitis is technical co-founder & CTO. Pain points: reliability/accuracy of finance research agents in a high-stakes regulated domain; inference/token cost as usage scales across 250+ institutions and 35K+ users; security/traceability of agent outputs. Challenges: production reliability + auditability for regulated finance; cost economics at scale; controlling multi-step reasoning agents. Must-haves: reliability/accuracy, auditability, cost control per agent run. Nice-to-haves: cost/latency visibility per workflow, model-routing efficiency. ICP confidence: Medium-High — technical co-founder/CTO decision-maker at a 121-person AI-native company shipping finance agents in production (Series D flagged as slightly beyond stated stage band).
Tuomas Artman — Co-Founder & CTO, Linear
Signal: 4 (technical co-founder at qualifying company actively developing 'Linear for AI Agents')
Source: https://linear.app/now/author/tuomas-artman ; https://www.menlotimes.com/post/enterprise-project-management-platform-linear-raises-82-million-at-1-25-billion-valuation
Company size: lean (~100-150 est.), Series C (~$134M total raised)
Pain points: integrating agents natively into product workflow so agents work alongside humans reliably
Challenges: agent reliability, permissions & governance inside a production system used by 33k+ companies
Must-haves: reliable agent execution & governance within product workflows
Nice-to-haves: visibility into agent actions/state
ICP confidence: Medium (CTO, Series C, right size, actively developing agent capabilities). Note: LinkedIn URL unconfirmed; Crunchbase profile linked instead.
Signal: 1 (ICP senior eng leader speaking about scaling/controlling agents and subagents at scale). Source: AI Engineer World's Fair 2026 (speaker/session data: https://www.ai.engineer/worldsfair/2026/speakers.json) — mainstage talk "Unlock Agent Autonomy: The Runtime for AI-Native Systems" (agents read/write entire codebases; subagents spawn to chase flaky tests, refactor modules, triage incidents — often unsupervised). Company size: ~1,028 employees (within 50–2,000); private, mature. Building agent runtime/tooling and running agents internally across SDLC. Pain points: running many autonomous agents/subagents at scale, often unsupervised; need for a controlled runtime and boundaries. Challenges: giving agents autonomy while keeping behavior controlled and observable across laptops/CI/cloud. Must-haves: control and boundaries over agent execution at scale. Nice-to-haves: standardized agent runtime. ICP confidence: Medium (EVP Engineering at a ~1k-person company building agent runtime and running 5+ agents across its SDLC; more infra-oriented and mature-stage, so not a perfect A–C fit).
Tyler Akidau — Chief Technology Officer, Redpanda Data
Signal: 4 (ICP speaking publicly about shipping agents reliably in production).
Source: https://newyork.qcon.ai/speakers/newyork2025 and https://ai.qconferences.com/schedule/newyork2025 — talk: "Taming Agentic AI: Lessons from Streaming Architecture for Scalable, Safe Systems."
Company size: ~188 (Tracxn/Unify 2026; range across sources 152–204). Series C data streaming infrastructure company, building agentic AI features on top of its platform.
Pain points: From the talk abstract — "Agentic AI is everywhere—but deploying it safely and reliably inside production systems is another story. The move from clever demos to autonomous agents that operate within enterprise infrastructure introduces new risks."
Challenges: The demo-to-production reliability gap; running non-deterministic agents inside infrastructure that customers depend on for correctness and ordering guarantees.
Must-haves: Runtime controls and safety boundaries for autonomous agents operating against production systems; reliability guarantees comparable to what streaming systems already promise.
Nice-to-haves: Reusing streaming/stream-processing primitives (replay, exactly-once, backpressure) as the substrate for agent orchestration.
ICP confidence: High — CTO, ~188 employees, well within band; publicly framing agent production-reliability as the core problem, which is our exact narrative. Strong technical credibility (ex-Google, Apache Beam / streaming systems author) so messaging must be substantive, not marketing-led.
Tyler Folkman — Chief AI Officer, JobNimbus
Signal: 1 (ICP writing about agent costs / compounding / control — first-person owned content).
Source: https://tylerfolkman.substack.com/p/i-cut-my-ai-agent-costs-7x-without (fetched and verified 2026-08-29, published 2026-05-03). Related posts on same Substack "The AI Architect": https://tylerfolkman.substack.com/p/i-tested-6-ai-models-across-3-providers ("I Routed 2,415 AI Agent Turns Across 6 Models. It Cost $76.77"), https://tylerfolkman.substack.com/p/my-cheaper-ai-coding-handoff-cost, https://tylerfolkman.substack.com/p/i-dont-use-one-ai-model-for-everything
Company size: ~283-292 employees (LeadIQ / PitchBook, Jul 2026). SaaS for contractors; growth-equity backed (Mainsail Partners, Sumeru Equity Partners) rather than classic Series A-C, but squarely inside the 50-2,000 band. Shipping AI features under NimbusLabs early access.
Pain points (verbatim from his own post): "cost = price_per_token x (tokens_you_need + tokens_you_waste)" — "I was optimizing the first term while the second term was 80% of the total." Also: "Start a session, load CLAUDE.md, AGENTS.md, skills directory, all MCP tools, and I was 30,000 tokens in before writing a line of code. Saying 'hello' cost 30K tokens." And: "the model got worse the longer I went. Context filled with MCP dumps, file listings, git diffs from five edits ago, test output I already read."
Challenges: he explicitly names three things that did NOT work — model downshifting ("Cheaper per gallon doesn't help when 80% is leaking"), "be concise" prompts ("Models ignore this. Input tokens don't change because the garbage filling context is tool output, not model output"), and usage caps ("Like cutting the credit limit instead of fixing spending habits"). This is the exact objection map for anyone selling a cheaper-model or a spend-cap wedge.
Must-haves: per-run / per-task cost visibility across model routes; tool-output compression before it hits context; a cost dashboard measuring tokens per unit of work (he built his own "tokens per PR" tracker).
Nice-to-haves: automated routing recommendations; eval gates and fallback logic around routed models; insulation against cheap-token pricing going away.
ICP confidence: High — Director+ title (Chief AI Officer), headcount in band, and he has independently written the exact thesis Alpha sells (waste not price, control at the harness layer). He has already hand-built a partial Alpha out of rtk + context-mode + manual routing + a homemade cost dashboard, which is the highest-converting profile in the library.
CAVEAT recorded honestly: title and employer verified via search-result metadata and his public bio, not via a directly loaded LinkedIn profile page. The Substack content and authorship are directly verified.
Signal: 4 (ICP building/shipping AI agents; found via web research — LinkedIn/Chrome NOT connected this run, prescribed LinkedIn searches unavailable; pivoted to web research + primary-source verification). Source: https://theorg.com/org/voiceflow + https://tracxn.com/d/companies/voiceflow (co-founder & CTO of Voiceflow; deduped against full ~781-person library — not present). Company size: ~54 employees (low end of the 50–2,000 band but qualifies); Toronto; ~$35M raised (Series A); Voiceflow is an agent-building/design platform enterprises use to build and ship AI agents. Pain points (inferred from company focus, NOT a verbatim quote): giving builders reliability + cost/token visibility across many customer-built agents; orchestrating multi-step agents; keeping deployed agents accurate. Challenges: scaling an agent-building platform; per-agent cost + behavior visibility for customers. Must-haves: reliability, cost/token visibility per agent. Nice-to-haves: multi-agent orchestration, evals/observability. ICP confidence: Medium-High (technical co-founder/CTO at an agent-platform company; headcount ~54 sits at the lower edge of the size band).
Tyler Johnson — Co-Founder & CTO, Tennr
Signal: 1 (technical co-founder publicly addressing agent reliability limits of frontier models). Source: https://www.healthcareaiguy.com/p/company-deep-dive-tennr ; https://foundationcapital.com/ideas/making-healthcare-move-faster-with-ai-(trey-holterman-co-founder-ceo-tennr). Company size: ~150-250 (Tennr, YC W23, Series C; healthcare back-office automation — patient intake, referrals, benefits). Pain points: frontier models can't reliably reason across unstructured medical records, payer rules, and the long tail of edge cases — Tennr built its own vision-language model (RaeLM) trained on 100M+ docs; production reliability on messy real-world data. Challenges: edge-case coverage; reliability on unstructured inputs; scaling high-volume document/agent processing. Must-haves: reliable reasoning over unstructured data, control over model behavior, eval on edge cases. Nice-to-haves: cost control on high-volume agent/document processing. ICP confidence: High — CTO/technical co-founder, Series C, 50-2000 employees, shipping agents in production.
Signal: 4 (ICP founder shipping voice/chat agents in production). Source: https://fortune.com/2026/04/27/avoca-ai-agents-missed-calls-hvac-plumbing-roofing-kleiner-perkins-chen-shrivastava-braswell/ ; https://www.ycombinator.com/companies/avoca ; https://www.prnewswire.com/news-releases/avoca-raises-125m-at-1b-valuation-to-power-americas-services-economy-with-ai-302753962.html. Company size: ~85 employees (NYC); $125M+ raised, $1B valuation, Series B (Meritech, General Catalyst; Series A led by Kleiner Perkins); 800+ customers, ~$1B in job bookings managed. Ex-Nuro PM. Avoca ships AI voice/chat/SMS/email agents for service industries (HVAC, plumbing, roofing) handling calls/scheduling/marketing/follow-ups — multiple agents in production at scale. Pain points (inferred): reliability + low latency of real-time voice agents at high call volume; cost per interaction across 800+ customers; scaling agent quality. Challenges: production reliability across live phone traffic. Must-haves: reliability, low-latency responses, cost control per call. Nice-to-haves: per-agent cost visibility, observability. ICP confidence: High (founder/co-CEO at an 85-person Series B AI-agent-native company shipping voice agents in production).
Signal: 4 — ICP-title technical AI leader at an agent-shipping company (found via people search: 'Head of Engineering agentic AI'). Source: https://www.linkedin.com/search/results/people/?keywords=Head%20of%20Engineering%20agentic%20AI%20startup . Company size: ~700-1,000 (est). Uniphore = enterprise conversational & agentic AI platform (brain already lists other Uniphore contacts, confirming active ICP account). Pain points (INFERRED from role+company, not a quoted post): scaling agentic/conversational AI in production; agent reliability; cost & latency at enterprise scale. Challenges: operating multi-agent conversational systems reliably. Must-haves: production reliability + cost/latency visibility for agents. ICP confidence: Medium-High — Head of AI Eng at in-range agent-native company; pains inferred from role.
Umut Gültepe — Head of Product, Platform, Abnormal AI
Signal: Signal 4 (LinkedIn people search at named agent company; Head of Product-Platform, AI-focused. Sourced by role/company match)
Source: https://www.linkedin.com/search/results/people/?keywords=%22Abnormal%20AI%22%20head%20of%20engineering
Profile: https://www.linkedin.com/in/gultepe/
Company size: ~1,200-1,800 (Abnormal AI; AI security agents; Series D)
Pain points: (inferred, not directly observed) Building the platform that runs security agents reliably and economically at scale
Challenges: (inferred) Balancing cost, reliability and governance of the agent platform
Must-haves: (inferred) Cost + reliability visibility built into the platform
Nice-to-haves: (inferred) Guardrails and cost governance as platform features
ICP confidence: Medium (Head of Product-Platform is AI-focused product leadership at an in-ICP agent company; slightly outside the pure eng/AI-leader core)
Uri Knorovich — Co-Founder & CEO (technical), Nimble (Nimble Way)
Signal: 4 (web-research proxy — LinkedIn/Chrome unavailable this run; found via Feb 2026 Series B announcement + company profile). Source: https://techcrunch.com/2026/02/24/nimble-way-raises-47m-to-give-ai-agents-better-cleaner-data/ ; https://www.norwest.com/blog/making-web-search-for-ai-agents-reliable-why-were-investing-in-nimble/ . Company size: ~136 employees (Tracxn/Crunchbase, 2026); Series B ($47M round, $75M total). Technical co-founder: former Head of Cyber Security AI Dept, Israeli Military Intelligence; years across AI + large-scale data systems.
Pain points (from company positioning): AI agents get unreliable/dirty web data → hallucination and bad actions in production; enterprises need validated, structured real-time web data agents can trust.
Challenges: making web search/data reliable enough for autonomous agents at enterprise scale; verifying and structuring results.
Must-haves: reliable, validated data feeding production agents; trust/accuracy.
Nice-to-haves: cost-efficient data retrieval at scale.
ICP confidence: High — technical co-founder/CEO at a confirmed 50-2,000-emp company whose entire product targets AI-agent reliability.
Signal: 4 (ICP senior technical AI leader at an agent-native company; surfaced via ICP-title + agent-company research after raw LinkedIn post/comment scans returned no ICP-matching authors this run). Source: https://theorg.com/org/aisera/org-chart/utkarsh-contractor ; https://aisera.com/company/ . Company size: ~250-500 (Aisera, agentic AI for enterprise, Series D). Pain points (inferred from role/company, not verbatim quotes): keeping domain-specific AI agents reliable in production; controlling LLM/inference cost as agent volume scales; per-run cost + quality visibility across many enterprise agents. Challenges: productionizing multimodal/domain LLM agents at enterprise scale and reliability. Must-haves: production reliability + guardrails, cost-per-run observability across a growing agent fleet. Nice-to-haves: automated benchmarking/eval, model routing to cut spend. ICP confidence: High (Field CTO/VP AI leading Agentic AI at an in-range agent-native company; verified current via multiple sources).
Uttam Kumar Bhatta — Senior Director of Engineering, Kore.ai
Signal: 4 (net-new ICP senior technical leader at an agent platform company). Source: https://www.linkedin.com/in/uttambhatta/ ; LinkedIn people search (Kore.ai engineering), 2026-07-27 run. Headline: 'Senior Director of Engineering | Agentic AI - Voice - LLM | 20+ years | Building AI-First Organizations | Ex-Kony'. Company size: ~1,000 employees (mature Series D-stage agent platform) - NOTE larger/later-stage than the Series A-C guideline but within 50-2,000 and agent-native. Kore.ai ships enterprise multi-agent/voice/LLM in production. Pain points (INFERRED, not verbatim): reliability + cost of voice/LLM agents at enterprise scale; visibility into per-run cost; retry/latency tax on voice agents. Challenges: scaling agent fleets reliably; cost control across deployments. Must-haves: cost-per-run + reliability observability. Nice-to-haves: model comparison on cost-per-completed-task. ICP confidence: High/Medium (Senior Director of Engineering, agentic AI; large/mature company).
Vadim Korolik — Head of Engineering, Observability, LaunchDarkly
Signal: 4 (senior technical leader at a company actively shipping agent runtime-control product).
Source: LinkedIn people search — headline read verbatim: "Head of Eng. for Observability @ LaunchDarkly | Prev. CTO at Highlight (Acquired)". Company agent activity verified via launchdarkly.com/blog/introducing-agentcontrol/, siliconangle.com/2026/05/19/launchdarkly-launches-runtime-control-layer-agentic-ai-era/, helpnetsecurity.com/2026/05/19/launchdarkly-agentcontrol/.
Company size: ~466 employees as of February 2026.
Pain points: agents in production behaving badly between deploys; needing to change agent behaviour, route to a different model or trigger a fallback mid-conversation-turn rather than after the fact; scoring agent output on cost as well as quality.
Challenges: sub-200ms config propagation to intervene before a bad response reaches a customer; building the observability layer that makes agent cost/quality legible.
Must-haves: runtime control and real-time cost/quality scoring of agent responses; per-config cost attribution.
Nice-to-haves: automatic escalation to a more capable (more expensive) config only when quality drops below threshold — i.e. spend routed by need.
ICP confidence: High — Head-of-Engineering level, 466-person company, directly owns the observability surface for a shipped agent-control product. Ex-CTO of Highlight (acquired), so a repeat technical buyer.
CAVEAT: LaunchDarkly AgentControl overlaps with parts of thealpha.ai's agent operating layer — competitive/adjacent, worth understanding before outreach.
Signal: 4 (ICP technical decision-maker at agent-native company shipping agents in production). Source: https://www.linkedin.com/in/vnivargi ; https://www.moveworks.com/us/en/company/vaibhav-nivargi. Company size: ~1,000–1,400 (Moveworks; 350+ enterprise customers, 5M+ workers served; agentic RAG + generative AI platform). NOTE: Moveworks acquired by ServiceNow (Dec 2025) — still operates as the agent-building org. Pain points (inferred from domain/role, not a verbatim quote): reliability of agentic RAG at enterprise scale; cost/observability across a large multi-agent fleet; making agents dependable across 350+ enterprise tenants. Challenges: integrating ML/agents into products at scale; consistent quality across tenants. Must-haves: production reliability, scalability, observability. Nice-to-haves: compounding intelligence / institutional memory. ICP confidence: Medium (founder & CTO of an at-scale agent platform in the size band; softened only because of the ServiceNow acquisition into a large parent).
Signal: 3 (adjacent to competitor space — agent guardrails/observability/control). Source: https://www.linkedin.com/in/vaikkunth/ and https://www.cbinsights.com/company/dynamo-ai. Company size: ~40-58 employees (borderline lower bound of ICP), Series A ($15M). Technical co-founder — MIT PhD. Dynamo AI builds guardrails/eval/observability for generative & agentic AI (DynamoGuard, DynamoEval, and AgentWarden, launched 2025, targeting agent-specific risks). Pain points / focus (from company positioning, not a verbatim quote): controlling agent behavior at scale, real-time guardrails, agent observability and risk in high-stakes/regulated industries. Challenges: making agents safe and controllable in production. Must-haves: agent control/observability. Nice-to-haves: cost visibility. ICP confidence: Medium — technical decision-maker, but company is an agent-safety vendor near the 50-employee floor rather than a team operating a large internal agent fleet; watch as a partner/competitor-adjacent signal as much as a pure prospect.
Vara Kumar Namburu — Co-founder & Head of R&D and Solutions, Whatfix
Signal: 4 (ICP speaking publicly about building and shipping agents). Found via web research — LinkedIn/Chrome unavailable this run.
Source: https://www.dqindia.com/interview/whatfix-launches-role-aware-ai-agents-to-drive-enterprise-efficiency-10466623
Company size: 1,298 employees as of Mar 2026 (Revelio Labs, down 1.9% YoY). Bengaluru, India + San Jose. Late-stage (past Series C) — stage is the only ICP stretch. ~85 Fortune 500 customers.
Agent evidence: launched three role-aware production agents — Authoring Agent, Insights Agent, Guidance Agent — built on ScreenSense, their proprietary context/intent inference layer. Stack: predominantly Azure OpenAI LLMs + RAG, with in-house models where performance/scale demands it, layered with Whatfix enterprise data. Agents embedded in existing workflows rather than as a separate copilot. Reported results: 30% authoring efficiency gain; insights from days to minutes. Expects agents to drive ~25% of revenue. Long-term goal: "zero-click outcomes."
Verbatim: "One of the biggest challenges with LLMs is moving from demonstration to reality. A demo often looks impressive, but making the solution work consistently across customer use cases requires a lot of effort. Unlike traditional software development, building AI-led solutions requires a different approach to evaluation, testing, and security. The answers from LLMs are not always predictable, so our R&D had to adapt with new evaluation frameworks and safeguards."
Pain points: demo-to-production gap — inconsistent agent behaviour across heterogeneous customer environments; LLM non-determinism breaking conventional QA; users forced to context-switch into separate copilots; analyst bottleneck for data questions.
Challenges: building new evaluation frameworks and safeguards from scratch because traditional software testing doesn't transfer; security review of AI-led features for Fortune 500 buyers; rolling agents out gradually across a large installed base without disrupting UX; rising integration demand (Confluence, Google Docs, Salesforce Knowledge) as customers want agents grounded in their own data.
Must-haves: repeatable evaluation/testing frameworks for non-deterministic systems; enterprise data grounding via deep integrations; role/persona-aware agent design; in-workflow embedding with no context switching; enterprise security and safeguards.
Nice-to-haves: mixing commercial LLMs with self-built models for high-scale paths; expanding automation scope from 30-minute tasks to full-day tasks; source-referenced answers for auditability.
ICP confidence: High — 1,298-employee Indian SaaS scaleup, technical co-founder running R&D, agents shipped and monetized in production, and he names the demo-to-production reliability gap unprompted.
Signal: 4 (technical co-founder/CTO at company shipping agentic AI in production). Source: https://startupintros.com/orgs/akasa ; https://akasa.com/ ; company/team searches (2026). Company size: ~200 employees; healthcare revenue-cycle automation using generative/agentic AI (medical coding, prior authorization, claims), deployed across 650+ hospitals and 6,500+ outpatient facilities representing $120B+ in net patient revenue. Founders: Malinka Walaliyadde (CEO), Varun Ganapathi (CTO), Andy Atwal (VP Eng), Ben Beadle-Ryby. Varun Ganapathi: Stanford AI PhD, ex-Google, co-founder of Udacity's early team — strongly technical. Pain points (inferred from product domain, NOT verbatim quotes): reliability/accuracy of agents acting on claims & clinical/billing data in a regulated, high-stakes setting; auditability and control of autonomous agent actions; per-run cost across very high transaction volume. Challenges: reliable, compliant, cost-efficient agents at hospital-system scale. Must-haves: reliability, governance/auditability, control. Nice-to-haves: per-run cost visibility/observability. ICP confidence: High (named CTO/technical co-founder; ~200-emp company in band actively shipping agentic workflows in production). LinkedIn URL not captured this run — do not fabricate.
Varun Kacholia — Co-Founder & Chief Technology Officer, Eightfold AI
Signal: 4 (ICP at a company with named agents in enterprise production).
Source: https://eightfold.ai/company/eightfold-ai-leadership/ ; https://eightfold.ai/blog/predicitions-ai-in-hr-2026/ ; https://artificialintelligencecompanies.com/co/eightfold-ai/
LinkedIn: https://www.linkedin.com/in/varunkacholia
Company size: 501-1,000 employees (Santa Clara HQ). $410M raised, $2.1B valuation (Series D led by General Catalyst + Lightspeed). Note: later-stage than the A-C band — flagged, not disqualifying.
Agent evidence: Recruiter Agent and Sourcing Agent in enterprise production — autonomous passive-candidate sourcing and screening at scale, reported 40-60% time-to-hire reduction. Company is publicly framing 2026 as agents moving from pilots to production.
Pain points (INFERRED): agents that autonomously screen and interview candidates sit under EEOC / NYC Local Law 144 / EU AI Act high-risk classification — every agent decision must be explainable and auditable by law, not by preference; sourcing agents run continuously over very large candidate corpora = high sustained token volume.
Challenges: proving non-discrimination in a probabilistic system; multi-tenant enterprise deployments each with different policy constraints.
Must-haves: decision-level audit trail and explainability; policy/guardrail enforcement per tenant.
Nice-to-haves: per-tenant cost attribution for agent runs.
ICP confidence: MEDIUM-HIGH. Role and agent-activity are ideal; headcount in band. Downgraded from High because Series D + deep in-house ML org (ex-Google/Facebook ranking leaders) = strong build-in-house risk.
Varun Vummadi — CEO & Co-Founder, Giga (GigaML)
Signal: Bucket 1 (ICP building/shipping agents — found via 2026 agent-startup research + LinkedIn verification). Source: https://www.linkedin.com/in/varunvummadi/ ; funding/customer confirmation https://finance.yahoo.com/news/giga-raises-61m-funding-reinvent-171500172.html . Company size: ~66 employees (Tracxn, Jan 2026); Series A $61M (Redpoint, YC S23, Nexus Venture Partners). Giga ships enterprise voice + chat support/ops AI agents in production for customers incl. DoorDash and Zepto, claiming 90%+ real-world resolution. Technical co-founder (IIT Kharagpur CS, Forbes 30U30). Pain points: production reliability of agents (90%+ resolution is their headline KPI — reliability is existential), cost per resolution at high volume, keeping quality high while scaling across many enterprise workflows. Challenges: proving agents beat 20+ competing vendors on real-world resolution; controlling cost/latency of high-volume voice+chat runs. Must-haves: per-run cost + reliability visibility across a growing fleet of production agents; guardrails against runaway/looping agents. Nice-to-haves: cross-agent observability and benchmarking as they scale 1→N deployments. ICP confidence: High (technical co-founder/CEO at a ~66-person Series A company actively shipping agents in production).
Signal: 4 (ICP technical leader building/shipping agents). Found via LinkedIn people search ('CTO agentic AI startup') + web verification — content-search buckets 1-3 were low-yield this run. Source: https://www.linkedin.com/in/vedavyas-panneershelvam-22080214/ | Company: Phaidra — reinforcement-learning AI agents that autonomously control industrial facilities & data centers ('AI factories'). Company size: ~100-150 (Series B, ~$120M raised; Collaborative Fund, Index, NVIDIA, Sony). Pain points (inferred from role/product, not verbatim quotes): running long-horizon control agents 24/7 in production; compute/cost of continuous inference at scale; reliability & safety of autonomous agents controlling physical infrastructure; visibility into agent decisions and cost-per-run. Challenges: scaling from pilot to many concurrent plant agents; guaranteeing uptime; compute-budget control. Must-haves: production reliability, safety guardrails, cost-per-run visibility. Nice-to-haves: cross-agent benchmarking, drift detection. ICP confidence: High — technical co-founder/CTO at a Series B, agent-native company inside the size band.
Signal: 4 (ICP technical leader found via LinkedIn people search at a named agent company). Source: https://www.linkedin.com/search/results/people/?keywords=Kore.ai%20VP%20engineering%20agents%20platform . Profile: https://www.linkedin.com/in/venkata-mutnuru/ . Company size: ~1,000-1,200 (enterprise agentic AI platform; RAG, agentic, distributed systems). Pain points (inferred from role+company, no direct quote): reliability + cost of RAG/agentic workloads at enterprise scale; observability into per-agent behavior. Challenges: controlling token/inference cost as agent usage grows; keeping agents reliable across enterprise tenants. Must-haves: cost-per-run + token visibility; production reliability tooling. Nice-to-haves: guardrails, shadow testing, spend anomaly alerts. ICP confidence: Medium-High (AVP Eng on the AI Platform team at a large independent agentic-AI company; 2nd-degree, in-band role). Found by scheduled task icp-prospect-signal-scanner run 2026-08-03.
Venkat Peri — Head of Agentic AI, Advisor360
Signal: 1/4 (LinkedIn people search for ICP agentic-AI leaders; verified via web). Source: https://www.linkedin.com/search/results/people/?keywords=Head%20of%20Agentic%20AI | profile https://www.linkedin.com/in/venkatperi. Company size: ~500-1,000 employees (mature/PE-backed private; PAST A-C stage). Agent activity: Advisor360 is building a "fully agentic operating system" for wealth management — Digital Workforce portfolio (Meeting IQ, Email IQ, Account Opening IQ, next-gen operational agent workflows), rolling out 2026. Pain points (INFERRED from role/company, not a verbatim quote): scaling operational agents across advisory workflows reliably; governance/control of member/advisor-facing agents; cost as agent workflows expand. Challenges: reliability + governance of a broad agentic OS in regulated wealth management. Must-haves: control/governance, reliability, cost visibility. Nice-to-haves: efficiency/routing. ICP confidence: Medium (Head of Agentic AI at a right-size ~500-1,000-emp company clearly building agents, but the company is a mature/PE-stage private co, not Series A-C).
Signal: 4 (technical founder; SaaStr AI Annual 2026 speaker — flagged that 67% of eng leaders raised AI investment yet 30% of engineers are still stuck on repetitive infra work).
Source: https://www.saastr.com/the-first-44-speakers-for-saastr-ai-annual-2026-the-founders-and-operators-actually-shipping-ai-at-scale/
Company size: ~137-158 employees (Series B DevOps automation; ships DevOps/infra AI agents).
Pain points: repetitive infra toil persists despite AI investment; hard to prove ROI on rising AI spend.
Challenges: scaling AI agents for infra/DevOps reliably and measurably.
Must-haves: demonstrable productivity/output from agents (not just consumption).
Nice-to-haves: cost efficiency at scale.
ICP confidence: Medium (technical founder/CEO at agent-shipping company in size range).
Venkata Koppaka — Co-Founder & CTO, Tenex AI
Signal: 3 (ICP CTO in the agentic-security space adjacent to competitors like AWS AgentCore; found via 2026 SOC-agent launch/funding sweep). Source: https://tenex.ai/tenex-ai-appoints-venkata-koppaka-as-cto-to-accelerate-ai-powered-mdr-during-rapid-expansion/ , https://www.linkedin.com/in/venkata-subba-rao-koppaka-24826b13/ . Company size: <200 employees (company committed to 250+ new hires in 2026); Sarasota FL / SF Bay; Series B $250M (2026). Tenex = AI-native MDR/SOC that runs autonomous triage/investigation agents (human-led). Koppaka was a founding engineer of Google Chronicle/Google SecOps and led autonomous-agent + NLP work in Google Cloud. Pain points (INFERRED): scaling autonomous SOC agents reliably at MDR volume; agent decisions must be defensible/auditable to enterprise security customers. Challenges (INFERRED): accuracy of agent triage; operational efficiency as both agents and analyst headcount scale. Must-haves (INFERRED): reliability, per-action audit trail, per-run cost visibility at scale. Nice-to-haves (INFERRED): unified cost/quality view across the agent fleet. ICP confidence: HIGH (named CTO/co-founder with deep agent pedigree; in-band headcount; autonomous agents core to product).
Venkatesh Yadav — VP, AI Applications and Delivery, H2O.ai
Signal: 4 (AI leader at AI-native company shipping agents; found via people search 'VP of AI customer support agents SaaS'). Source: https://www.linkedin.com/in/yadavvenkatesh/ . Company size: ~250-400 employees (H2O.ai, AI-native, ships agentic AI incl. h2oGPTe agents — in range). Role: VP, AI Applications & Delivery. Pain points (inferred from role/company, no direct post captured this run): delivering agentic AI applications to enterprise customers in production — reliability, cost-per-run, and scaling multiple agent deployments. Challenges: making customer agent deployments cost-predictable and reliable. Must-haves: per-deployment cost visibility + production observability. Nice-to-haves: governance across customer tenants. ICP confidence: Medium (senior AI leader at in-range AI-native company shipping agents; role leans delivery — pain inferred, not an authored post).
Signal: 4 (ICP technical co-founder/CTO at a company shipping agentic AI in production). NOTE ON METHOD: LinkedIn browsing unavailable this run (Chrome extension not connected); found/verified via web research. Pain points INFERRED from Prophet's domain + CTO podcast, not verbatim quotes.
Source: https://venturebeat.com/ai/ai-vs-ai-prophet-security-raises-30m-to-replace-human-analysts-with-autonomous-defenders ; https://www.prophetsecurity.ai/blog/prophet-security-raises-30-million-series-a-led-by-accel ; https://www.northsecurity.xyz/p/prophet-security-ai-analyst
Company size: ~40-70 (Series A, $30M led by Accel, July 2025; multi-agent agentic AI SOC platform — triage/hunting/detection agents in production). NOTE: Series A headcount may sit near the 50-emp floor — confirm before outreach. Prior background: Sreekanti was VP Eng at StackRox (acq. by Red Hat); ~decade eng leadership at Oracle/Red Hat.
Pain points (inferred): running multiple autonomous SOC agents reliably in production; controlling per-investigation LLM cost as alert volume scales; trust/accuracy of autonomous agent decisions.
Challenges (inferred): scaling from a few agents to a multi-agent mesh without cost blowout; observability into agent cost/behavior per investigation.
Must-haves (inferred): reliability + cost-per-run visibility for production security agents.
Nice-to-haves (inferred): cost attribution across agent types.
ICP confidence: Medium — textbook CTO ICP persona at an agent-native company; caveat is early-stage headcount possibly near/below 50.
Victor Duprez — Senior Director of Engineering, AI, Gorgias
Signal: 4 (ICP building/shipping agents in production). Source: https://theorg.com/org/gorgias/org-chart/victor-duprez . Company size: ~350-500 (Series C ecommerce CX platform; multi-team engineering org). Pain points: scaling conversational AI agents to ~1.6M automated customer conversations/month; needs higher reliability guarantees and skills-based routing as agent volume grows — classic 1-to-many agent scaling wall. Challenges: reliability at high automated-conversation volume, routing/orchestration across many agents, maintaining resolution quality at scale. Must-haves: production reliability, control over agent behavior at scale, per-conversation observability. Nice-to-haves: cost attribution per resolution. ICP confidence: High — Senior Director of Engineering (AI) is a clear decision-maker at an in-range company shipping agents at high volume.
Vijay Bharadwaj — Chief Data Scientist, Machinify
Signal: 4 (ICP shipping agents). NOTE: LinkedIn/Chrome NOT connected this run — found via web research; LinkedIn URL surfaced in search results and is recorded as-is, not browsed. Source: https://www.machinify.com/resources/qa-with-chief-data-scientist-vijay-bharadwaj ; https://www.machinify.com/bio/vijay-bharadwaj ; https://theorg.com/org/machinify/org-chart/vijay-bharadwaj ; agentic AI positioning https://www.machinify.com/resources/a-healthcare-evolution-modernizing-your-health-plan-with-an-ai-operating-system. Company size: ~662-923 employees (LeadIQ ~740 Mar 2026 and ~923 Jul 2026; SignalHire 662). NOTE headcount sources disagree widely (PitchBook lists 110 — treated as stale); all credible current sources put it inside the 50-2,000 band. Backed by New Mountain Capital; healthcare payment integrity. Agentic AI confirmed: Machinify publicly states it is using agentic AI to change clinical data extraction — systems that decompose a task, build a structured plan, execute steps in sequence and validate intermediate results. Background: previously VP of Product Innovation - Algorithms at Netflix (7+ yrs). Pain points (INFERRED from company-published material, NOT personal quotes): multi-step planning agents over clinical documents where each step must be validated — step-level validation is a direct token multiplier; accuracy is regulated and auditable. Challenges: agents that re-plan and re-read context per step; cost per claim/chart at payer volumes. Must-haves: per-step and per-run cost attribution; validation/reliability tooling for multi-step plans. Nice-to-haves: caching or context reuse across similar documents. ICP confidence: Medium-High — title is Chief Data Scientist (senior technical AI leader, above Director bar, owns the AI function) rather than a literal CTO/VP-Eng/Head-of-AI string; company size band confirmed but sources vary; agentic AI in production confirmed. UNVERIFIED: exact headcount, personal pain statements.
Signal: 1 (ICP publicly discussing agentic AI economics/ROI). Source: https://www.linkedin.com/in/vijums/ and https://ushur.ai/ai-agents. Company size: ~213-240 employees (Series C, $108M raised, founded 2014, Santa Clara). Actively building AI agents: yes — Ushur AI Agents complete complex requests end-to-end in healthcare, insurance, financial services. Pain points: making agentic AI pay off on margins (he is publicly focused on "agentic AI and its importance for firms looking to maximize profit margins"); deploying reliable end-to-end agents in regulated verticals without handoffs. Challenges: production reliability + compliance for autonomous agents in regulated industries; cost-to-value of agent deployments. Must-haves: agents that reliably complete complex regulated workflows end-to-end; ROI/margin visibility. Nice-to-haves: agent observability and per-workflow cost attribution. ICP confidence: High (VP of AI at Series C, ~220-employee company shipping agents in production). Note: pain points inferred from public role/talks, not a single verbatim quote.
Signal: 1 (ICP quoted publicly on the pilot-to-scale wall for agents).
Source: https://m.thewire.in/article/ptiprnews/kaptures-survey-of-cx-leaders-reveals-massive-ai-ambition-but-only-7-have-scaled-agentic-ai-enterprise-wide (PRNewswire, 8 Dec 2025); title verified on https://www.kapture.cx/about-us/ (updated 23 Jun 2026).
Company size: ~613 employees as of Jul 2026 per LeadIQ (https://leadiq.com/c/kapture-cx/5a1d99c02300005b00889eaa); Tracxn showed 325 as of Aug 2025. Estimate 500–650 global — inside band. $10M pre-Series B Jul 2026 led by Bajaj Finserv Ventures.
Pain points: (verbatim) "The new question is, 'How do we scale sophisticated, outcome-driven, human-like AI fast enough?' Organisations want measurable transformation, not experiments that stay locked in innovation labs." Same release reports 50% of CX leaders stuck in pilots and only 7% scaled agentic AI enterprise-wide.
Challenges: Agents stuck in pilot; no measurable outcome data to justify scaling; the 1→many transition is the explicit bottleneck he names.
Must-haves: Measurable per-agent outcomes and economics; a path out of pilot purgatory into enterprise-wide deployment.
Nice-to-haves: Benchmarks against peer CX deployments.
ICP confidence: Medium-High — technical co-founder + CPO at a company inside the headcount band actively shipping agentic CX, with a public quote naming the exact scaling wall in Alpha's ICP definition. Not High only because the quote is Dec 2025 rather than within the last quarter.
Account note: second Kapture contact alongside Sanchit Sood (Chief AI Officer). Kapture ships its own agent observability platform — competitive overlap; wedge on run economics, not monitoring.
Vikas Salaria — Director – AI Products, DigitalNet.ai
Signal: 4 (ICP-title AI leader at a qualifying company shipping agents). Source: https://www.linkedin.com/in/vikassalaria/ (LinkedIn people search "Director of AI agents LLM production"). Company size: ~1,100-1,150 employees, 300+ AI/data specialists (DigitalNet.ai, Bethesda MD; 2,000+ trained AI agents on 50+ models). Confirmed 50-2,000 band and actively shipping agents. Role: Director – AI Products (owns AI agent/product roadmap; Bengaluru). Pain points (inferred from role/company, no direct quote): productizing and shipping many AI agents reliably; cost per agent/run at fleet scale; visibility/observability across a large agent portfolio. Challenges: turning 2,000+ agents into dependable, cost-predictable products across verticals. Must-haves: reliability, per-run cost visibility, agent observability/governance. Nice-to-haves: model right-sizing, reusable agent scaffolding. ICP confidence: Medium-High (Director-level AI product leader at an in-range, clearly agent-native/agent-heavy company; strong fit, pains inferred not quoted this run). profile_url captured in Source too.
Vikram Verma — VP Engineering & Site Leader (India), Level AI
Signal: 4 (ICP engineering leader at an agent-shipping company; surfaced via company-scoped LinkedIn people search, confirmed via own recent activity).
Source: https://www.linkedin.com/in/vikramverma/recent-activity/all/ (post ~1 week ago promoting Level AI's "Level Up" CX leaders event, Nov 4-6 San Jose; quotes practitioners on using live conversation data to reshape roadmaps in real time). Company page: https://www.linkedin.com/company/level-ai/about/
Company size: 51-200 employees (LinkedIn company page, verified this run). Level AI ships a full-stack agentic CX platform (virtual agents + agent-assist + QA) in production.
Pain points: Not directly expressed in the public posts reviewed. Contextual pain from his remit: running a multi-team agentic CX platform org and site, where virtual agents, agent-assist and QA all run against live customer conversations.
Challenges: Owns delivery for the India site across an agentic platform that must operate at enterprise CX volume — implies per-run cost and reliability exposure, but this is inference, not a stated quote.
Must-haves: Unknown / not stated publicly.
Nice-to-haves: Unknown / not stated publicly.
ICP confidence: High — VP of Engineering (above Director bar), company is 51-200 employees, AI-native, and demonstrably shipping agents in production. Confidence is on fit, not on an observed pain quote; treat pain discovery as an open question for outreach.
Ex-Yubi / Yahoo / Informatica; IIT-D CS, IIM-B.
Viktor Qvarfordt — VP of Engineering, Sana (Sana Labs)
Signal: 4 (senior eng leader at company building/shipping agents). Source: https://se.linkedin.com/in/viktor-qvarfordt ; https://craft.co/sana-labs/executives ; https://sanalabs.com/products/sana/ai-agents. Company size: ~495 employees (Stockholm/London/NY); Series C $55M (Nov 2024, NEA), $130M+ total. Sana ships "Sana Agents" — build-your-own expert AI agents for work (knowledge access, automation) deployed across enterprises. Pain points (role/company inferred; no direct public cost quote captured): scaling agentic AI across enterprise knowledge work; reliability/accuracy of agent answers; cost of running agents at 495-person + enterprise-customer scale. Challenges: enterprise-grade reliability and governance; multi-agent orchestration. Must-haves: reliable agents, scalable deployment, cost predictability. Nice-to-haves: per-agent cost/observability. ICP confidence: Medium — strong fit on role (VP Eng), size (~495, Series C), and clear agent product; downgraded from High only because no direct cost/reliability pain quote from Viktor was captured this run (fit-based, not signal-based).
Signal: 4 (ICP technical leader hired explicitly to scale an AI agent portfolio). Source: https://www.project44.com/press-releases/project44-names-vinay-mathur-as-chief-technology-officer/ ; https://www.globenewswire.com/news-release/2026/06/09/3308889/0/en/project44-names-Vinay-Mathur-as-Chief-Technology-Officer.html ; https://www.project44.com/press-releases/project44-launches-ai-freight-procurement-agent-to-cut-freight-spend-and-accelerate-sourcing/ . Company size: ~789 employees across 5 continents (Jun-2026). Supply-chain visibility / Decision Intelligence platform; later-stage (Series F) - stage deviation, headcount in range.
Pain points: INFERRED (no verbatim pain quote found - do not attribute one). His appointment release states he leads engineering, platform infrastructure and applied AI for "the company's expanding portfolio of AI agents for shippers and logistics service providers" - i.e. he owns a growing agent fleet, not one agent.
Challenges: fleet is growing by acquisition as well as build - project44 acquired LunaPath in Apr 2026, adding AI execution agents that must be integrated onto the existing platform. Heterogeneous agents from two codebases on one platform is exactly the 1->5+ wall. AI Freight Procurement Agent shipped Feb/Mar 2026 into the TMS, automating carrier selection, rate benchmarking and negotiation - high-volume, always-on, per-run cost scales with freight volume.
Must-haves: unified visibility and control across agents inherited from acquisition plus agents built in-house; reliability on agents that negotiate and commit spend.
Nice-to-haves: per-run cost attribution per shipper/LSP tenant.
ICP confidence: HIGH on role and company fit; MEDIUM on expressed pain (inferred from mandate and product releases, not stated). LinkedIn profile URL NOT confirmed this run - do not guess it before outreach.
Timing: appointed 9 Jun 2026, ex-Head of Product Engineering at Stripe. A brand-new CTO ~6 weeks into an explicit agent-scaling mandate is an unusually open buying window; he is still choosing his platform stack.
Vinay Perneti — VP of Engineering, Augment Code
Signal: 4 (ICP building/shipping agents in production, speaking publicly). Source: https://www.iheart.com/podcast/1333-ai-giants-a-live-intervie-317881978/episode/vinay-perneti-vp-engineering-augment-code-317881996/ . Company size: ~156 employees, Series B ($227M raised) — in range. Augment Code builds AI coding agents that understand entire codebases (IDE + CLI, autonomous task coordination). Pain points: persistent memory / large-context "context engine" for production coding agents; company content notes "AI costs don't fit traditional annual forecasting models — usage is spiky, hard to predict, and growing fast." Challenges: unpredictable/spiky agent cost, context management, scaling autonomous coding agents. Must-haves: cost predictability and context/token efficiency. Nice-to-haves: per-task cost visibility. ICP confidence: High (exact VP Engineering role, agent-native company, right size, public on production agent cost/context/scaling).
Vineet Singh — Co-Founder & CTO, Darwinbox
Signal: 4 (ICP building/shipping agents) — found via web research; LinkedIn/Chrome extension NOT connected this run; verified via Infomance, PeopleMatters, Tracxn, Agentic List 2026.
Source: https://www.infomance.com/news/darwinbox-promotes-cto-vineet-singh-to-co-founder/ ; https://www.agentconference.com/agenticlist/2026 (HR & Talent Agents)
Company size: ~1,000–1,500 employees (Hyderabad, India; ~$149M–$307M raised across sources; ~$1B valuation; founded 2015). HCM/HR platform shipping agentic AI across the employee lifecycle.
Role note: Elevated from CTO to Co-Founder in recognition of scaling Darwinbox's tech & product; remains top technical leader.
Pain points (INFERRED from domain/role, not verbatim quotes): running HR-workflow agents reliably across large multi-country enterprise deployments; controlling LLM cost per run as agentic features roll out to a big HCM user base; data/permission-safe autonomous actions.
Challenges: reliability & governance of agents acting on sensitive HR data; cost scaling across enterprise tenants.
Must-haves: production reliability, cost predictability, guardrails/governance.
Nice-to-haves: per-run cost visibility and agent observability.
ICP confidence: Medium-High — CTO/technical co-founder at a company (near upper bound of 50–2,000-emp band) actively shipping agentic AI; headcount figures vary by source.
Vipul Lonkar — Senior Director of Engineering — Platform, Avaamo
Signal: 4 (Sr Director of Engineering building agent/voice-AI platform; found via LinkedIn people-search). Source: https://www.linkedin.com/in/vipull/ (headline: 'Senior Director of Engineering — Platform @ Avaamo | AI Agents, Voice AI, RAG | AI-Native Org Builder'). Company size: 201-500 (LinkedIn-verified). Building agents: yes - Avaamo ships enterprise conversational/voice AI agents. Pain points (inferred from role/headline, not a quoted post): scaling multi-agent + voice-AI platform in production; reliability & latency of voice agents; RAG/context token cost. Challenges: production reliability of voice agents at platform scale. Must-haves: reliable low-latency agent platform; context/token cost efficiency. Nice-to-haves: per-run observability across agents. ICP confidence: Medium-High (Sr Director of Eng at 201-500-emp agent-native company building AI-agent/voice platform).
Signal: 4 (ICP technical co-founder at an AI-native RevAI platform shipping agents). Source: https://6sense.com/about-us/team/ ; https://www.linkedin.com/in/viralbajaria/ ; https://www.crunchbase.com/person/viral-bajaria-2. Company size: ~1,500 employees ($5B+ B2B AI platform, processes >1 trillion intent signals daily). Pain points (inferred from public product/positioning): running AI across an enormous signal graph and shipping Revenue AI agents; scaling autonomous GTM agents reliably over massive data. Challenges: reliability and cost of agents operating on trillion-signal-scale data; controlling autonomous agent behavior. Must-haves: reliable, controllable agents at data scale. Nice-to-haves: per-agent cost visibility. ICP confidence: Medium (technical co-founder, but public title is shifting from CTO to Chief Innovation & Ecosystem Officer — decision-maker fit slightly ambiguous; 50-2,000 emp; agents in production).
Viren Baraiya — Co-Founder & CTO, Orkes
Signal: 1 (writing/speaking about agent reliability in production) + 4. Source: https://www.ai.engineer/worldsfair/schedule ; funding verified via businesswire (Orkes $60M Series B, Apr 2026). Company size: ~73 employees (Series B; built Netflix Conductor). Orkes is an agentic workflow orchestration platform focused on deploying AI agents to production with reliability. Pain points (inferred from company thesis): agents failing/stalling in production, no durability across multi-step agent workflows, poor visibility into agent runs, scaling from a few to many agents. Challenges: giving enterprises production-grade agent reliability. Must-haves: durable execution, observability, control over agent behavior at scale. Nice-to-haves: cost/latency visibility per agent run. ICP confidence: Medium-High — technical co-founder/CTO at a 73-person Series B in the agent-reliability space; note Orkes is orchestration-adjacent (possible partner/competitor dynamic).
Vishal Madan — VP Engineering, Information Security and Integrations, iMocha
Signal: 4 (company actively shipping agentic AI). Source: https://www.linkedin.com/in/madanvishal/ — reposts of iMocha AI Readiness Agent (autonomous action: identifies below-fluency employees, builds learning paths, schedules assessments, notifies managers, tracks progress) and Task/Skills Intelligence; hashtag AgenticAI, from AI insights to AI execution. Company size: ~300-500 (iMocha, skills-intelligence SaaS, Pune). Pain points: scaling AI-native SaaS to enterprise-ready; moving from dashboards to agents that take action; AI governance, security and integrations. Challenges: building skills data the business trusts; enterprise-grade governance/reliability. Must-haves: reliable agentic execution, security/governance. Nice-to-haves: per-agent-run cost visibility. ICP confidence: Medium (VP Eng at mid-size SaaS demonstrably shipping agentic AI).
Vishal Parikh — Co-founder & Chief Product Officer, Hippocratic AI
Signal: 4 (ICP technical co-founder building/shipping AI agents; found via company primary sources, not LinkedIn — Chrome/LinkedIn not connected this run). Source: https://hippocraticai.com/team/ and https://unicornscreener.vc/blog/7-ai-agent-startups-funded-by-top-vcs-in-2026. Company size: ~300 (Series C, $126M raise Nov 2025, $3.5B valuation; 50+ health systems, 1,000+ clinical use cases, 115M+ patient interactions — clearly shipping many agents in production). Role note: Co-founder & CPO who "leads product and engineering across all of technology including Polaris, infrastructure, interfaces, and tools" (Stanford masters; deep AI/engineering background) — qualifies as senior technical/product leader (VP/Head of Product AI-focused + technical co-founder). Pain points: clinical-grade safety and reliability at scale; keeping a "no safety issues" record across 100M+ live agent interactions; production-grade multimodal/voice agents. Challenges: reliability across a large fleet of clinical agents and multiple underlying models; safety guardrails and evaluation at scale. Must-haves: safety/eval guardrails, production reliability, observability across an agent fleet. Nice-to-haves: cost/observability per interaction or per run. ICP confidence: High (clean technical co-founder leading product+engineering at a mid-size Series C agent-native company).
Signal: 1 (ICP quoted on agent token-cost blowout). Source: TechCrunch "The token bill comes due" (2026-06-05, https://techcrunch.com/2026/06/05/the-token-bill-comes-due-inside-the-industry-scramble-to-manage-ais-runaway-costs/); LinkedIn https://www.linkedin.com/in/vitalygordon. Company size: ~75 (Series A engineering-intelligence platform building autonomous AI software-engineer agents; ex-VP Eng Salesforce / founder of Salesforce Einstein). Pain points: engineers burning ~$40k/mo on tokens with unclear ROI ("I genuinely don't know whether I should stop him or tell everyone to be like him"); output up but bugs/rewrites up; heavy users spend 10x tokens for ~2x productivity. Challenges: proving ROI of AI dev tools; measuring business value of shipped AI code; token-spend visibility at scale. Must-haves: agent/token cost visibility tied to engineering outcomes. Nice-to-haves: model routing / efficiency benchmarking. ICP confidence: Medium-High — technical co-founder/CEO of an agent-building company, vocal on the exact pain; note Faros is partly observability-adjacent.
Signal: LinkedIn People search (agent-native company + agents), closest to Bucket 4. Source: https://www.linkedin.com/in/vitaly-shagurin/ (found via people search 'Parloa engineering AI agents'). Company size: ~470-506 employees (verified via Tracxn/PitchBook/RocketReach, 2026). Parloa = Berlin voice AI agent platform for contact centers (agent-native, Series C, in-band). Pain points: pre-production agent testing/simulation and compressing product feedback loops (his recent post cites Bain synthetic-customer simulations to pre-test agent concepts, refine weak ideas earlier), agent quality/evaluation. Challenges: reliably validating agent behavior before shipping; scaling agentic CX quality. Must-haves: agent evaluation/simulation, reliability. Nice-to-haves: cost visibility, observability. ICP confidence: Medium. Note: title is 'Product Leader - Agentic AI' (senior product owner for the core agentic product; Director/VP level not explicitly confirmed on profile). Company clearly in-band + agent-native.
Vivek Muppalla — VP AI Engineering, Hippocratic AI
Signal: 1 (ICP senior technical leader publicly speaking about agent reliability in production). Source: AI Engineer World's Fair 2026 (speaker/session data: https://www.ai.engineer/worldsfair/2026/speakers.json) — talk "200 Million Patient Interactions Later: What the Generic Voice Stack Misses" (a healthcare voice agent can be right on the benchmark and wrong in production). Company size: ~312 employees, Series C ($3.5B val, Nov 2025) — AI-native healthcare voice-agent company. Pain points: gap between benchmark scores and real-world production behavior of agents at massive scale (200M+ patient interactions); generic voice/agent stacks miss production edge cases. Challenges: keeping large fleets of patient-facing agents reliable and safe in production. Must-haves: production-grade reliability/observability into what agents actually do per interaction. Nice-to-haves: better eval-to-production correlation. ICP confidence: High (VP AI Engineering at a 312-person Series C AI-native company shipping agents at scale; reliability pain is core thesis fit).
Vivek Raju Muppalla — VP AI Engineering, Hippocratic AI
Signal: 1 (ICP speaking publicly about agent reliability in production). Source: https://www.agentconference.com/speaker/Vivek-Raju-Muppalla . Company size: ~181 employees (Sept 2025, ~107% YoY growth), Series C — in range. Safety-focused generative AI voice agents for healthcare (patient/clinical calls). Pain points: pushing "four and five nines of reliability in agentic performance"; latency budget, turn-taking, and safety stakes for healthcare voice agents. Challenges: guaranteeing reliability/safety of autonomous voice agents talking to patients. Must-haves: very high reliability + evaluation of agent behavior in production. Nice-to-haves: observability into failure modes across live calls. ICP confidence: High (VP-level AI engineering leader, ICP company size/stage, public on agent reliability).
Vladimir Poliakov — Head of Engineering (Vision AI), Sword Health
Signal: Bucket 4-adjacent (ICP engineering leader at agent-native healthtech). Sourced via LinkedIn people search "Sword Health AI engineering director". Source: https://www.linkedin.com/in/pauleech/ . Company size: ~1,000-1,520 employees (Sword Health, Series D+ unicorn; size in 50-2000 band). Agent-native: leads Vision AI engineering behind the Phoenix production care agent. Pain points (inferred from role/company; no public post captured this run): production reliability of real-time vision+LLM agents, inference cost at scale, observability of agent runs. Challenges: keeping many concurrent production agents reliable and cost-bounded. Must-haves: reliability guardrails, cost-per-run observability. Nice-to-haves: portability, compounding memory. ICP confidence: Medium (Head of Engineering ✓, size ✓, agent-native ✓; stage Series D+; no direct engagement signal observed).
Volodymyr Giginiak — Co-Founder & CTO, Wordsmith
Signal: 4 (ICP technical leader at an agent-native company absent from the brain). Source: gap-company research after brain dedup (Wordsmith not in the current list); title verified on LinkedIn ('Volodymyr Giginiak — CTO @ Wordsmith', London). Wordsmith (Edinburgh/London) is an in-house legal AI platform deploying 'AI paralegal' agents across legal tasks (contract review, document Q&A, etc.). Funding: $70M Series B (Jun 2026, Highland Europe / Index Ventures); founded 2023; scaling toward ~300 staff by end-2026 (currently ~100-200, in range). Volodymyr Giginiak is Co-Founder & CTO (10+ yrs Meta/Instagram, 6+ yrs Microsoft). Pain points: reliability/accuracy of legal agents on high-stakes output; cost per run; observability/evals; scaling agents across many in-house legal teams. ICP confidence: High (Co-founder & CTO; vertical AI agents; headcount in range). Note: pains inferred from role/company profile, not a personal quote.
Vrajesh N Sejpal — Director of Engineering, Data Science, Ushur
Signal: 4 (company-scoped ICP people search; found via LinkedIn people search, NOT an observed post — pains INFERRED from role+company). Source: https://www.linkedin.com/search/results/people/?keywords=Ushur%20director%20VP%20engineering%20AI%20agents | Company size: ~300 (Ushur). CAVEAT: company (Ushur) INFERRED from search context/mutual connections — headline did not explicitly name employer; treat as unconfirmed. Pain points (inferred): data-science/ML behind production agents; model cost and reliability. Challenges (inferred): reliable agent behavior; cost of inference at scale. Must-haves (inferred): reliability + cost-per-run visibility. Nice-to-haves (inferred): token/context-waste reduction. ICP confidence: Medium (Director of Engineering, Data Science; agent-native company likely Ushur but employer UNCONFIRMED; title verified, company + pains inferred).
Walden Yan — Co-Founder, Cognition AI (Devin)
LinkedIn: https://www.linkedin.com/in/waldenyan. Signal: 4 (ICP building & shipping autonomous agents in production at scale). Source: https://research.contrary.com/company/cognition ; https://jobsbyculture.com/blog/working-at-cognition-2026. Company size: ~305 employees (March 2026; $25B+ val, ~$492M ARR run-rate for Devin). Pain points: scaling reliability of a fully autonomous coding agent (Devin) across huge production volume; cost per agent run as usage explodes; long-running multi-step agent sessions going off-track. Challenges: keeping autonomous agents reliable and cost-bounded under real production load; visibility into what agents do across many concurrent sessions. Must-haves: production reliability, per-run cost control, agent observability at scale. Nice-to-haves: guardrails for long-horizon autonomy. ICP confidence: High (technical co-founder, IOI gold medalist, ~305-employee company shipping agents at massive scale).
Warren Van Winckel — Director of Engineering, AI Platform, Upwork
Signal: 4 (ICP eng leader at company shipping agents; found via people search 'Director of Engineering AI agents SaaS production'). Source: https://www.linkedin.com/in/warrenvw/ . Company size: ~1,500 employees (Upwork, public marketplace shipping AI agent 'Uma' — in range). Role: Director of Engineering for the AI Platform. Pain points (inferred from role/company, no direct post captured this run): running an AI platform that powers agent features in production at marketplace scale — reliability, latency, and cost of LLM/agent calls. Challenges: controlling agent/LLM spend and reliability as usage scales across the platform. Must-haves: production observability + cost controls per agent/feature. Nice-to-haves: multi-model routing/fallbacks. ICP confidence: Medium (Director-level technical AI-platform leader at in-range company shipping agents; pain inferred from role).
Signal: 4 (ICP building/shipping agents). Source: https://writer.com/blog/humans-of-ai-waseem-alshikh/ and https://writer.com/blog/writer-action-agent-press-release/. Company size: ~250-350 (Series C, $1.9B val, AI-native enterprise platform). Pain points: building, activating and SUPERVISING autonomous agents grounded in enterprise company data across hundreds of enterprises; keeping agent behavior controllable at scale. Challenges: shipping an "Autonomous Super Agent" reliably into large regulated enterprises; grounding agents in proprietary data; cost of running agents at enterprise volume. Must-haves: agent supervision/observability, reliability, control over agent behavior. Nice-to-haves: cost efficiency on their in-house LLM family vs frontier models. ICP confidence: High (CTO/co-founder at AI-native company shipping agents in production, target size band). LinkedIn slug not confirmed — profile_url left blank to avoid fabrication.
Waseem Alshikh — Co-founder & CTO, WRITER
Signal: 1 (ICP writing publicly about what breaks in enterprise AI evaluation and control).
Source: https://www.linkedin.com/in/waseemalshikh/recent-activity/all/ — post from ~14 hours before this run. Opens: "Every enterprise AI evaluation runs the same test. Same prompts, four models, score the outputs." He argues that test "has never once told a security/eval team what it needed to know," and lists five questions to put to every vendor in writing: (1) where content goes after the prompt, with retention terms in the contract, not a marketing claim; (2) published data on model behaviour on contested topics, not a policy; (3) the model's security report as distinct from the company's SOC 2; (4) who holds the encryption keys — "if the answer isn't 'we do,' you're renting a promise"; (5) a year later, can you prove which asset was AI-generated, with provenance carried on the output itself rather than a log. Company page: https://www.linkedin.com/company/getwriter/about/
Company size: 201-500 employees (LinkedIn company page, verified this run). WRITER is an end-to-end enterprise platform where teams "build, activate, and supervise" AI — agentic work orchestration in production at enterprise customers.
Pain points: Standard model-comparison evals produce nothing decision-useful for security/eval teams; vendor claims about data handling are unverifiable; no durable provenance on AI-generated output; supervision and auditability are afterthoughts.
Challenges: Making enterprise buyers able to verify — contractually and technically — what a model/agent did with their data, and proving it a year later.
Must-haves: Evidence-grade artifacts (retention terms, model security report, key ownership, output-level provenance) rather than dashboards or policies.
Nice-to-haves: Published behavioural reports; a repeatable vendor-evaluation framework.
ICP confidence: High — technical co-founder/CTO, 201-500 employees, AI-native, agents in production. Note: his stated pain is supervision/auditability/provenance rather than cost-per-run; that is an adjacent wedge to Alpha's control story, not the cost story.
Signal: 4 (technical co-founder shipping accounting agents in production; regulated-adjacent finance). Source: https://finance.yahoo.com/news/digits-launches-first-ai-agents-140000473.html ; https://tracxn.com/d/companies/digits . Company size: ~78 employees (Crunchbase/Tracxn, 2026); Series C $65M led by SoftBank Vision Fund at ~$565M valuation (Mar 2022), ~$100M total; founded 2018 by Wayne Chang and Jeff Seibert (previously co-founded Crashlytics, acq. by Twitter) — Wayne is the technical co-founder. Pain points (inferred from public product/positioning; no first-party quotes captured): AI Accounting Agents on the Autonomous General Ledger handle up to 95% of bookkeeping workflows — correctness/reliability essential because outputs are financial records; auditability of automated ledger actions; scaling agent autonomy in finance. Challenges: correctness/reliability of financial automation; auditability and trust; scaling. Must-haves: reliability, action-level auditability, accuracy. Nice-to-haves: cost-per-run visibility. ICP confidence: Medium (technical co-founder, agents in production, accounting/regulated-adjacent; ~78 emp just above the 50 floor; last priced round Series C dates to 2022 — stage/recency caveat; no explicit CTO title).
Will (Dongxu) Lu — VP of Engineering & Head of AI Strategy (Co-Founder & CTO, Orby AI), Uniphore (Orby AI)
Signal: 4 (senior technical AI leader building enterprise agents; co-founded/CTO of Orby AI — Large Action Models & agentic process automation — now VP Eng & Head of AI Strategy at acquirer Uniphore). Source: https://www.linkedin.com/in/will-lu-9b9b972b/ ; https://www.uniphore.com/orby-ai/ . Company size: Uniphore ~700-1000 employees (Orby was ~50-100, Series A $30M, acquired by Uniphore) (NOTE: Uniphore is late-stage, beyond A-C; headcount within range). Ex-Google (9 yrs enterprise AI), NVIDIA. Pain points: scaling enterprise agentic process automation reliably; reliability of action-model/agent execution; agentic process discovery. Challenges: neuro-symbolic reasoning for reliable automation; enterprise-grade reliability & observability. Must-haves: reliable action-based agents, control, observability. Nice-to-haves: cost efficiency per run. ICP confidence: Medium (strong senior technical AI leader shipping agents; company stage beyond A-C).
Signal: 4 (ICP technical co-founder at agent-native company shipping agents in production). Source: LinkedIn (title "Co-founder at Convey, Previously CTO"), a16z announcement, Wellfound team page. Company size: 106 employees (PitchBook, confirmed). Convey builds enterprise "digital teammates" (AI agents trained by demonstration to autonomously own end-to-end workflows); $38M Series A led by a16z; $50M ARR in 7 months; >1.1M hours automated; customers NBCUniversal, Samsara, Unity, Faire, ChargePoint. Harvey previously sold a company to project44 (p44); Stanford; technical co-founder alongside CEO Rohan Chopra and Diego Canales. Pain points: reliable autonomous teammates owning outcomes end-to-end; last-mile reliability at production quality; scaling agent fleets across enterprise workflows. Challenges: reliability/quality at volume. Must-haves: production reliability, monitoring. Nice-to-haves: cost-per-run visibility. ICP confidence: High (technical co-founder/prev CTO at 106-person agent-native company). Note: same company as Rohan Chopra (id above) — distinct individual, second exec at Convey.
Will Lawrence — Co-Founder & CEO, Greenlite AI (Bretton AI)
Signal: 4 (ICP co-founder/decision-maker at agent-native company shipping agents in production). Source: https://www.businesswire.com/news/home/20250521200064/en/Greenlite-AI-Raises-$15M-Series-A ; https://www.heyfuturenexus.com/greenlite-is-on-a-mission-to-revolutionize-banking-compliance/ ; https://tracxn.com/d/companies/greenlite/ . Company size: ~78 employees (Tracxn, Jan 2026). Series A / ~$90.5M raised. Will Lawrence is co-founder & CEO; prior led product for Facebook's AML platform and built compliance infrastructure at Paxos — deep domain buyer, not a pure eng leader. Building/shipping AI agents: production compliance agents (KYC/AML/sanctions/monitoring) for regulated banks and fintechs. Pain points (inferred from product domain + public statements, NOT a verbatim quote): needing agents trustworthy and auditable enough for regulators; proving automated compliance decisions are correct and explainable; scaling the AI "workforce" across customers. Challenges: reliability/trust in regulated environments; audit and control. Must-haves: reliability, auditability, control at scale. Nice-to-haves: cost/observability per agent run. ICP confidence: Medium (co-founder & CEO/economic buyer rather than technical decision-maker; company in band and shipping agents; pair with CTO Alex Jin for technical entry).
Will Lu — VP Engineering & Head of AI Strategy, Uniphore
Signal: 4 (ICP Head-of-AI persona at company shipping agentic AI). Source: https://willlu.com/ ; https://www.uniphore.com/our-team/. Company size: ~700–1,000 (Uniphore). NOTE: later-stage funding (beyond Series C) — headcount and agent focus fit ICP. Leads AI strategy + platform; co-founded Orby AI (agentic automation), acquired by Uniphore. Pain points (inferred): turning enterprise intelligence into a "compounding advantage"; moving agents from pilots to production-grade outcomes; reliability and cost of agentic automation. Challenges: building a durable agent platform; observability/eval; cost per run. Must-haves: reliability, observability, compounding intelligence. Nice-to-haves: cost efficiency across model routing. ICP confidence: Medium (strong Head-of-AI persona and agent builder; softened by later funding stage; LinkedIn URL unconfirmed — profile_url left empty rather than guessed).
willlu.com
Willem Delbare — Co-Founder, CTO & CEO, Aikido Security
Signal: 1 (ICP publicly discussing AI agents as an operational/attack surface). NOTE: LinkedIn/Chrome unavailable this run — LinkedIn URL surfaced in search results, not opened/verified.
Source: https://www.aikido.dev/team-members/willem-delbare ; https://www.globenewswire.com/news-release/2026/01/14/3218567/0/en/aikido-security-raises-60-million-series-b-at-1-billion-valuation-to-lead-software-security.html
Company size: ~118 at $13M ARR (Jan 2025 press signal); materially larger now post-Series B. Ghent, Belgium. Series B — $60M at $1B valuation, Jan 2026, led by DST Global with PSG Equity, Notion Capital, Singular. ~$85M total.
Agents in production: STRONG. Two named shipping agents — "AI AutoTriage" (three-layer dedup + reachability analysis, claimed 95% alert-noise reduction) and "AI AutoFix" (autonomously opens remediation PRs across code, dependencies, IaC, containers). 100,000+ teams. Named customers: Premier League, Revolut, SoundCloud, Niantic. Series B explicitly earmarked for "autonomous self-securing software."
Pain points: [evidenced] He argues companies must treat AI agents and developer-installed tooling as part of their attack surface — "without visibility into these layers, organizations are missing where attacks are actually happening" (paraphrase from a Unite.AI interview summary — VERIFY EXACT WORDING BEFORE QUOTING). He describes the autonomous loop as: test real attack paths on every code change → confirm exploitability → generate and apply a fix in-workflow → retest to confirm. That verify-then-retest design is a direct admission he does not trust raw agent output. [inferred] False-positive economics and cost per autonomous scan at 100k-team scale.
Challenges: Agent output trustworthiness at scale (he engineered an explicit verification loop rather than trusting the fix); agent observability as a security surface; noise/false-positive cost when agents auto-open PRs into customer repos.
Must-haves: Verification/validation layer around agent output; visibility into what agents and dev tooling are actually doing; per-scan cost control at 100k-team volume.
Nice-to-haves: Cross-customer agent behaviour benchmarking; drift detection on agent fix quality.
ICP confidence: High — technical co-founder with CTO title, Series B at $1B, two named autonomous agents shipping to 100k+ teams, and public commentary on agent visibility. NOTE: he is CEO as well as CTO, so he is also the commercial buyer-facing exec.
William Colen — Director of Artificial Intelligence, Blip
Signal: 3 (ICP engaging with competitor/adjacent agent infrastructure — named Microsoft Azure OpenAI customer story). NOTE: LinkedIn/Chrome not connected this run. LinkedIn URL not verified.
Source: https://www.microsoft.com/pt-br/customers/story/1748706491048194135-blip-azure-openai-service-professional-services-pt-brazil
Company size: ~1,000-1,800 employees (Microsoft customer profile band "1,000-9,999"; press reports ~1,500-1,800) — near the top of our band, worth re-verifying before outreach. Series C, $60M (Nov 2024, Warburg Pincus), $230M total. Conversational/agent platform, Brazil.
What he said (paraphrased from Portuguese): needed a partner flexible and agile enough that its backing gave robustness to GenAI at scale.
Pain points: trust and robustness of infrastructure when scaling GenAI copilots across many enterprise clients; governance and moderation across a large agent fleet.
Challenges: moving from a single high-volume WhatsApp chatbot to governed, always-on enterprise copilots (Blip Copilot).
Must-haves: governed cloud infra, moderation/toxicity filters, compliance layers.
Nice-to-haves: pre-built ecosystem integrations.
ICP confidence: Medium — Director of AI title is a clean ICP match and Blip runs agents at genuinely large scale in an under-covered geography (Brazil); downgraded from High because the quoted pain is infrastructure-partner framing rather than explicit cost/reliability pain, and headcount may exceed 2,000.
William Steenbergen — Co-Founder & CTO, Federato
Signal: 4 (ICP technical co-founder/CTO shipping agents in production). Source: https://www.federato.ai/about ; https://www.crunchbase.com/organization/federato ; Tracxn (May 2026 ~232) / PitchBook (~156) headcount ; Carrier Management (Oct 2025). Company size: ~156-232 (in range). Stage caveat: Series D ($100M led by Goldman Sachs; ~$180M total) — past the ICP's stated Series A-C band, but headcount in range and clearly agent-shipping. Actively shipping agents: Federato's "RiskOps"/AI-native underwriting platform ships agentic AI that generates fully-explained P&C insurance quotes in minutes and guides underwriter decisions. Steenbergen = technical co-founder & CTO; Will Ross = CEO. Pain points (inferred from product + regulated insurance domain, NOT verbatim quotes): explainability/auditability of underwriting-decision agents; reliability of agent-generated quotes; cost-per-quote at carrier scale. Challenges: trust with carrier risk/compliance teams; scaling agents across portfolios. Must-haves: explainability, control, cost-per-run visibility. Nice-to-haves: agents compounding on underwriter feedback. ICP confidence: Medium (strong technical-CTO + agent fit and in-range headcount, but Series D is past the ICP's stated stage band).
William Wang — Founder & CEO (also Professor of AI/NLP, UC Santa Barbara), ChipAgents (Alpha Design AI)
Signal: 4 (ICP technical founder/CEO at agent-native company shipping agents in production). NOTE: Found via web research (press + CEO interview); LinkedIn/Chrome NOT connected this run. Source: https://www.businesswire.com/news/home/20260729819576/en/ChipAgents-Expands-Series-A-Funding-to-$134-Million-as-Demand-Grows-for-Agentic-AI-in-Semiconductor-Design and https://semiwiki.com/eda/chipagents-ai/353206-ceo-interview-with-dr-william-wang-of-alpha-ai-design/ and https://www.bvp.com/news/meet-the-founder-and-ceo-of-chipagents-william-wang. Company size: ESTIMATED >=50 (NOT directly confirmed — but $134M raised, deployed to 120+ semiconductor companies incl. MediaTek & Micron, 6x ARR growth in H1 2026, founding engineering team of veterans from NVIDIA/Amazon/Microsoft/Meta/Snowflake/Salesforce). Stage: Series A / A2 ($134M total; backers B Capital, Bessemer, Micron, MediaTek, Ericsson, ScOp). Role: technical founder/CEO (UCSB AI/NLP professor). Actively shipping AI agents: agentic AI platform applying domain-specific agents to turn specs/code into production-ready RTL, verification assets, and automated root-cause analysis. Pain points (inferred, not verbatim): reliability/accuracy of agents on high-stakes chip design & verification; scaling agents across 120+ customers; automated root-cause at scale. Challenges: correctness/trust of agent outputs in EDA; production-readiness. Must-haves: reliability, correctness, production-readiness. Nice-to-haves: token/cost efficiency per agent run. ICP confidence: Medium (technical founder/CEO, agent-native, well-funded and shipping to 120+ customers; company headcount not directly confirmed but strong signals of in-band size).
Willie Yao — Head of Engineering, Clay
Signal: 4 — ICP writing publicly about building/shipping agents.
Source: https://www.linkedin.com/in/willieyao/recent-activity/all/ (found via https://www.linkedin.com/company/clay-run/people/?keywords=head%20of%20engineering)
Company size: 201-500 employees (LinkedIn company page, verified this run). Clay = GTM data/agent platform (Claygent), Series C.
Observed this run (verbatim from his post hosting "Clay Tech Talks: How Better Builders Create Better Agents"): "Sam Shulman and Ryan Chang will share the real systems they've built, from AI agents that learn from your team to the principles behind great agentic products." He also shared a Notion case study on Clay: "Clay had all their context in Notion... So they built 80+ Custom Agents in a week."
Pain points (observed/implied by his own content): going from a handful of agents to 80+ custom agents inside one company in a week — classic 1→N agent sprawl; distinguishing "great agentic products" from ones that merely demo well.
Challenges: agent quality and learned/team-specific behaviour at scale; internal agent proliferation with no obvious per-agent accounting.
Must-haves: repeatable principles/systems for building agents that hold up in production.
Nice-to-haves: visibility into what each of the 80+ agents costs and contributes.
ICP confidence: High — Head of Engineering (core ICP title) at a 201-500 employee Series C company shipping agents both as product and internally.
Signal: 4 (ICP technical leader at an agent company). NEW contact at an account already in the brain (Cresta). Found via web + LinkedIn verification. Source: https://www.linkedin.com/in/xiangru-chen-09132719/ | Company: Cresta — AI agents & conversation intelligence for contact centers; ~$100M ARR (2026), Series C. Company size: 300+. Pain points (inferred from role/product, not verbatim quotes): scaling real-time voice/chat agents across large contact-center deployments; latency and cost-per-interaction at high volume; production reliability/consistency. Challenges: cost-per-run at scale, observability, agent quality. Must-haves: production reliability, cost visibility, low latency. Nice-to-haves: benchmarking/eval tooling. ICP confidence: High — VP Engineering, 300+ employees, agent-native, Series C.
Signal: 4 (ICP shipping agents). NOTE: LinkedIn/Chrome NOT connected this run — found via web research. LinkedIn URL NOT captured (not fabricating); profile_url is the Crunchbase person profile where the title was verified. Source: https://www.crunchbase.com/person/xun-wang-6ce6 ; https://www.bloomreach.com/en/about-us/leadership-team ; agent launches https://www.bloomreach.com/en/news/2026/bloomreach-announces-the-general-availability-of-loomi-marketing-agent/ ; https://www.bloomreach.com/en/news/2026/bloomreach-transforms-customer-conversation-into-ecommerce-insight-with-conversational-agent/ ; https://finance.yahoo.com/technology/ai/articles/bloomreach-launches-multi-agent-ask-123000815.html. Company size: 997 employees worldwide as of Mar 2026 (Revelio Labs); 37.6% of workforce is Engineering; 119 open roles in 2026 (+67% YoY). Mountain View CA, founded 2009. Multi-agent shipping confirmed and recent: Loomi Marketing Agent GA June 2026; Loomi Conversational Agent enhancements 2026; multi-agent "Ask Me Anything" launched Aug 2026; Loomi Connect agent-building product. Wang leads R&D engineering and large-scale low-latency ML/NLP search products. Pain points (INFERRED from company-published material, NOT personal quotes): a multi-agent surface running at consumer ecommerce traffic volumes — per-query inference cost is a direct margin line; agents that "remember shopper preferences" and reason over multi-turn context drive token growth. Challenges: 5+ distinct agent products (marketing, conversational, search AMA, Connect) on shared infrastructure; consumer-scale QPS makes cost-per-run economics existential rather than optional. Must-haves: cost-per-run/per-query visibility across a multi-agent portfolio; latency and reliability at ecommerce scale. Nice-to-haves: per-tenant cost attribution for merchant customers; cheaper model routing for low-intent queries. ICP confidence: High — CTO title, 997 employees (in band), 4+ agents shipped to GA in 2026. UNVERIFIED: LinkedIn URL, personal pain statements.
Signal: 4 + 1 (ICP technical co-founder shipping AI agents; posts about it on LinkedIn in his own voice). Source: https://www.linkedin.com/posts/yakirsudry_today-buildots-enters-a-new-era-for-years-activity-7450161576487104512-xlNS ("Today Buildots enters a new era...", Construction Intelligence / Intelligence Lab launch) ; https://buildots.com/about-us/ ; https://tracxn.com/d/companies/buildots/ ; https://www.crunchbase.com/organization/buildots ; LinkedIn: https://www.linkedin.com/in/yakirsudry/ . Company size: 416 employees (Tracxn, 26 May 2026). $166M raised across 7 rounds. Construction progress-tracking / construction intelligence.
Pain points: INFERRED from product shape (no verbatim cost or reliability quote found - his launch post is celebratory, not a pain post). Buildots ingests continuous jobsite camera and sensor video and runs vision + LLM reasoning over it; that is an always-on, high-token-volume workload where per-run cost scales directly with number of sites, not number of users. Their product Dot (GenAI assistant, Oct 2025) plus Intelligence Lab (2026) is a widening agent surface.
Challenges: heavy multimodal inference at per-site volume; construction customers buy on a per-project P&L, so inference cost lands directly against project margin.
Must-haves: cost per site/per project run visibility; reliability on unattended continuous ingestion.
Nice-to-haves: control/audit over what the reasoning layer looked at before it flagged a delay.
ICP confidence: MEDIUM. Role, headcount and technical-cofounder profile all qualify cleanly, but the agent evidence is weaker than the rest of this run - "Dot" is framed as an assistant and Intelligence Lab as a research hub, not as a named production agent fleet. Also note Crunchbase still lists him as "Co-Founder and VP of R&D" while The Org and Startup Nation Finder list Co-Founder & CTO. Read his recent posts and confirm the agent story before outreach; do not open with a 5-agent-fleet assumption.
Yaniv Shachar — SVP of Research & Development, Nym Health
Signal: 4 (senior technical leader at a company shipping autonomous decisioning in production). Found via web research — Claude-in-Chrome/LinkedIn NOT connected this run.
Source: https://nym.health/about/ (official leadership page)
Company size: ~85–100 employees (Tracxn/CB Insights profiles cite 89 as of mid-2024; a second source cites 84 across Tel Aviv and New York). NOTE: no official 2026 figure found — size is an estimate near the ICP floor and should be reconfirmed. Series C; ~$94.5M total raised (Addition-led rounds; PSG in the investor list). HQ New York, R&D Tel Aviv.
LinkedIn URL: not verified
Pain points: Running an autonomous medical-coding decisioning engine "processing millions of patient encounters annually in more than 400 healthcare facilities across the US" — per-encounter compute cost directly compresses gross margin; no clean attribution of inference spend to individual automated runs; accuracy/reliability bar is regulatory-grade.
Challenges: Moving from a deterministic NLP/rules engine to LLM-and-agent-assisted workflows without breaking auditability; scaling R&D headcount and agent workloads in parallel. Nym's own page says he "leads the company's R&D organization and drives its AI-first engineering strategy." Previously SVP R&D at Verbit.ai; co-founded Upscale.guru, "an AI-driven cloud optimization platform for development teams" — so cost instrumentation is a personal thesis, not a new idea to him.
Must-haves: Cost-per-run attribution at encounter level; deterministic audit trail; HIPAA-grade data handling; no accuracy regression.
Nice-to-haves: Model/version A-B comparison in production; automated regression detection across coding outputs.
ICP confidence: Medium-High — non-founder SVP R&D at a Series C healthcare vertical AI company running autonomous decisioning in production with an explicit "AI-first engineering strategy" mandate and a cloud-cost-optimization background. Downgraded from High because headcount (~85–100) is only verified to 2024 and sits near the 50-employee floor.
Yann Jouanin — Director of Engineering Strategy and Transformation, TheFork
Signal: 3 (ICP engaging with competitor content — named in an Arize AI customer case study).
Source: https://arize.com/blog/how-thefork-leverages-online-evals-to-boost-conversions-with-arize-ax-on-aws/
Company size: ~926–965 employees (LeadIQ / RocketReach cross-check, 2026). TheFork, France — restaurant booking platform, a Tripadvisor company. Running LLM/agent features on the conversion-critical path.
Pain points: (1) no visibility into duplicated or wasted LLM calls sitting on a revenue-critical path; (2) unclear cost signal per agent run; (3) slow evaluation iteration loop.
Quote: "Arize helped us turn tracing into tangible wins: lower latency, clearer cost signals, and faster iteration."
Challenges: proving agent changes actually improve conversion rather than just latency; running online evals continuously in production rather than offline.
Must-haves: clear per-run cost signals; tracing that surfaces redundant/wasted calls; fast eval iteration tied to a business metric.
Nice-to-haves: automated regression detection on agent quality; cost/latency tradeoff dashboards.
ICP confidence: Medium-High — Director of Engineering level (clears the Director bar), headcount in range, and an explicit "clearer cost signals" pain that maps directly onto the wedge. Caveat: TheFork is a Tripadvisor subsidiary rather than an independent Series A–C company, so budget authority may sit upstream. LinkedIn URL not found in search results — not guessed.
Yehoshua 'Shuki' Cohen — VP Applied AI, AI21 Labs
Signal: 1/4 (ICP at AI-agent-orchestration company Maestro; surfaced via LinkedIn people search). Source: https://www.linkedin.com/in/shuki-cohen/ | Company size: ~70 employees mid-2026 (post ~60% restructuring; Series D; sole product = Maestro agent orchestration; customers Nebius, Wix). Pain points (inferred): delivering reliable, cost-efficient agent solutions to enterprise customers; per-run agent cost/latency; eval of agentic workflows. Challenges: applied-AI delivery of Maestro agents at production quality; cost/reliability tradeoffs. Must-haves: cost-per-run visibility, reliability/guardrails. Nice-to-haves: eval & observability tooling. ICP confidence: Medium — VP Applied AI core to the Maestro agent product; caveat: Series D, lean post-restructuring.
Yi Liu — VP of Engineering, Head of Search, Moveworks
Signal: 4 (ICP VP of Engineering at an agent company). Source: https://www.linkedin.com/in/yi-liu-60284970/ (found via LinkedIn people search 'Moveworks VP engineering AI agents'). Company size: Moveworks ~500 employees (enterprise agentic assistant / AI agents; acquired by ServiceNow in 2025 — operates as an agent-building unit, parent org larger). Pain points (inferred from role, not a verbatim quote): scaling enterprise agent + search infrastructure reliably; retrieval and agent-response quality vs cost at enterprise scale. Challenges: multi-agent reliability, latency, and cost across large enterprise deployments. Must-haves: production reliability and observability of agents; retrieval quality. Nice-to-haves: per-agent cost attribution. ICP confidence: Medium-High (clear VP Eng at an agent company; caveat: now part of ServiceNow, so parent headcount exceeds 2,000).
Signal: 1/4 (LinkedIn people search for ICP agentic-AI leaders; verified via web). Source: https://www.linkedin.com/search/results/people/?keywords=VP%20of%20AI%20agents | profile https://www.linkedin.com/in/yinyin-liu-ml. Company size: ~1,500 employees (late-stage/private; PAST A-C stage). Agent activity: Seismic ships 9+ "Aura" GTM agents (Meeting, Presentation, Analytics, Program, Page, Lesson, Admin, Roleplay, Search agents); 248 orgs using Aura. Pain points (INFERRED from role/company, not a verbatim quote): scaling many agents across GTM workflows reliably; cost as agent usage scales across the customer base; visibility into per-agent performance and cost. Challenges: reliability + cost control across a broad multi-agent product surface. Must-haves: per-agent cost/behavior visibility, reliability. Nice-to-haves: routing/efficiency levers. ICP confidence: Medium (right-size ~1,500 emp and clearly agent-active, but company is past the Series A-C stage band).
Signal: 4 (ICP shipping a multi-agent product) + adjacent to Signal 3 (cost-observability category — he thinks in cost-per-unit natively).
Source: https://www.businesswire.com/news/home/20260604534090/en/Finout-Launches-AI-Agent-Suite-for-Enterprise-FinOps ; https://www.finout.io/about/ ; https://theorg.com/org/finout/org-chart/yizhar-gilboa
LinkedIn: https://www.linkedin.com/in/yizhargilboa
Company size: ~113-120 employees (May 2026, Tracxn/CB Insights/Revelio). Tel Aviv. $85M raised over 4 rounds; ~$14.7M revenue 2026. Series B/C stage.
Agent evidence: Finout Agents launched 5 Jun 2026 — a 3-agent chain: Detector Agent (continuous anomaly watch), Investigator Agent (root-cause trace), Orchestrator Agent (drives the fix). Plus Billy AI co-pilot and the MegaBill data layer.
Pain points (INFERRED): the Detector Agent runs CONTINUOUSLY over enterprise cloud billing data — an always-on agent over huge datasets is the single most expensive agent shape there is; a 3-agent chain means compounding context between detect -> investigate -> remediate.
Challenges: the Orchestrator Agent takes remediation ACTION on customer cloud infrastructure — that is a high-consequence write path needing approval gates and rollback; also the irony risk of a cost-control vendor whose own agent spend is unmodelled.
Must-haves: per-run cost ceiling on an always-on agent; approval/rollback on the remediation agent.
Nice-to-haves: eval/drift tracking across the 3-agent chain.
ICP confidence: HIGH — technical co-founder/CTO, 120 employees (comfortably in band), 3 agents in production as of Jun 2026.
Positioning note: he sells cloud-cost observability. Pitch as complementary (agent-layer cost/control, not cloud-infra cost) — he will grasp the value prop faster than anyone, but will also benchmark us hard on it. Not a competitor: different layer.
Yoav Ben Azar — Head of Product, Wonderful
Signal: 1 (ICP author writing explicitly about cost-vs-capability tradeoffs in agent work).
Source: https://www.wonderful.ai/blog-articles/fast-iteration-as-a-strategy (byline on wonderful.ai). Title "Head of Product" from LinkedIn/ZoomInfo snippets — the blog byline itself carries no title, so treat title as medium confidence.
Company size: 350 employees as of 12 Mar 2026, scaling to ~900 by year-end — verbatim from Insight Partners Series B release. Series B, $150M at €1.7B.
Pain points: (verbatim) "Agent work is full of tradeoffs: autonomy vs. control, flexibility vs. compliance, cost vs. capability. Iteration gets faster when those tradeoffs are surfaced clearly, not discovered after deployment."
Challenges: Cost/capability tradeoffs are discovered post-deployment rather than at design time; slow iteration because the tradeoff surface is opaque.
Must-haves: Cost and capability tradeoffs made visible pre-deployment; fast feedback loop on what a config change costs.
Nice-to-haves: Side-by-side comparison of agent configurations on cost and quality.
ICP confidence: Medium — the pain statement is a direct match to Alpha's "know what a run costs before you ship it" thesis and headcount is verified from primary source, but the title is sourced only from third-party aggregators, not a Wonderful-owned page. Verify title before outreach.
Yochai Konig — VP, Machine Learning & AI, Ada
Signal: Bucket 4 (ICP leader building/shipping agents at a qualifying company). NOTE: LinkedIn/Chrome not connected this run; found via web research + primary-source verification.
Source: https://theorg.com/org/ada-cx/org-chart/yochai-konig ; https://www.crunchbase.com/person/yochai-konig ; https://www.ada.cx/posts/meet-adas-vp-of-machine-learning . VP, Machine Learning & AI at Ada since 2021; builds/leads the ML org and spearheaded Ada's generative AI customer-service agent. Prior VP of Genesys X; 30+ yrs ML, 100+ patents.
Company size: ~350–500 employees (Series C unicorn). Non-founder (founders: Mike Murchison, David Hariri — both already in library).
Ships agents: Ada's "agentic customer service" — generative AI agents that resolve customer inquiries end-to-end at high volume.
Pain points: cost/quality trade-offs of automated resolutions at scale; model selection/routing; token waste; reliability/accuracy of agent resolutions.
Challenges: keeping resolution quality high while controlling per-resolution cost across enterprise volume.
Must-haves: visibility into cost-per-resolution and model routing controls.
Nice-to-haves: automated evaluation/observability of agent quality in production.
ICP confidence: High — exact non-founder VP (ML/AI) title verified on company + third-party sources; AI-native support-agent vendor.
Yonatan Boguslavsky — Co-Founder & CTO, Port (port.io)
Signal: 1 and 4 (ICP writing prolifically about agent control, governance and scaling; highly active poster). Source: https://www.linkedin.com/in/yonatan-boguslavsky-36354b125/ (post feed read live 2026-08-30) and https://www.calcalistech.com/ctechnews/article/r1oppgdzzg. Company size: ~200 employees (CTech); Tel Aviv, Israel; internal developer platform / "Agentic SDLC Platform"; $100M Series C at $800M valuation. Customers include GitHub, BT, Visa, StubHub, zooplus. Pain points (his own verbatim posts, past 7 days): "I keep hearing the same thing from engineering leaders: 'we're moving from AI-assisted coding to a fully autonomous SDLC.' Then you actually map where their teams are, and most are two stages earlier than they think." On governing agents: "Skills and MCP servers are load-bearing in the playbook, but the moment a policy changes, someone has to approve the skill that encodes it. Someone also has to decide which MCP servers a team is even allowed to use. Versioning, ownership, access control, and proof the skill is actually firing." On measurement: "The playbook names a leading and a lagging metric per stage. Naming them is the easy half. Nobody ships you the dashboard." Challenges: knowing whose turn it is in a chain of agent handoffs; calculating blast radius before letting an agent run unattended; capturing intent at scale from Slack/Zendesk/Sentry/PagerDuty; proving which teams actually adopted the process vs "just nodded". Scale reference he cites: a customer running agents across "600 engineers, 80 teams, 5000 repos". Must-haves: a governed control plane over the agent layer (which agents, which MCP servers, whose approval), pipeline-wide visibility "not just the code", blast-radius calculation. Nice-to-haves: scorecards for adoption, paved-path migration for legacy repos. ICP confidence: High — co-founder CTO, right size and stage, agents-in-production is both the internal reality and the core product thesis; note he is BUILDING an adjacent control layer, so this is a partner/competitor-aware conversation as much as a buyer one.
Signal: 4 (ICP co-founder/CTO at agent-native company shipping agents in production). NOTE: Found via web research (exec profile + press); LinkedIn/Chrome NOT connected this run. Source: https://venturefizz.com/insights/meet-the-executive-7ais-co-founder-cto/ and https://www.globenewswire.com/news-release/2026/05/26/3301008/0/en/after-7-million-investigations-in-production-7ai-widens-the-gap-in-agentic-security-with-plaid-elite-and-100-new-ai-jobs-in-boston.html and https://blog.7ai.com/three-stars-thoughts-on-passing-the-100-employee-mark-at-7ai-and-whats-next. Company size: ~100 (publicly passed 100-employee mark May 2026; plans to roughly double to ~200 by end of 2026). Stage: Series A ($166M total; described as largest cybersecurity Series A in history; Boston; backed by Index Ventures). Co-founders: Lior Div (CEO, already in People Library) and Yonatan Striem-Amit (CTO). Actively shipping AI agents: agentic security platform with 'swarming' AI agents; 7M+ investigations in production; launched Federated SIEM + agentic investigation, threat hunting, response, workflow automation. Pain points (inferred/market, not verbatim): reliability + control of swarming security agents in production; cost/scale of running millions of agent investigations; visibility into agent actions/outcomes. Challenges: trustworthy autonomous investigation at enterprise scale; governing agent behavior. Must-haves: reliability, control, auditability. Nice-to-haves: cost-per-investigation/run visibility. ICP confidence: High (Co-Founder & CTO; ~100 emp in-band; Series A; agent-native with agents in production at scale).
Yoni Blumenfeld — Co-Founder & CTO, Sett
Signal: 4 (technical co-founder/CTO at a qualifying company shipping AI agents in production)
Source: https://www.calcalistech.com/ctechnews/article/rjxbyqisze | https://www.bvp.com/news/meet-the-founders-of-sett | https://techcrunch.com/2025/05/07/game-sett-funding-a-startup-building-ai-agents-for-game-development-emerges-from-stealth-with-27m/
Company size: ~50 employees (Series B $30M led by Greenfield; $57M total; customers Zynga, Playtika, Papaya)
Pain points (inferred from company positioning): runs fleets of marketing/creative AI agents end-to-end for game user-acquisition; positions on cost efficiency (claims ~25x cheaper / 15x faster than human teams), implying heavy focus on per-output agent cost and throughput.
Challenges: scaling agent output reliably across many gaming clients; keeping creative quality high while driving cost down; orchestrating multi-step agent workflows for player acquisition & growth.
Must-haves: cost-efficient agent execution at scale; reliable production output.
Nice-to-haves: per-campaign visibility into agent cost & quality.
ICP confidence: High (CTO, ~50 emp, agents in production — exact ICP fit).
Yotam Sela — Co-founder & CTO (ex-Oracle, Ravello Systems), Aligned
Signal: 4 (building/shipping autonomous B2B-sales agents at scale — found via web research, LinkedIn/Chrome unavailable this run; pain inferred from company positioning, no personal quote). Source: https://www.calcalistech.com/ctechnews/article/b1kmpvz7gx ; https://app.dealroom.co/news/feed/aligned-raises-60m-series-b-to-automate-complex-b2b-sales-with-ai-agents . Company size: 55 employees (plan to ~85 by year-end). Funding: Series B $60M (PeakSpan Capital); ~$68M total. Product: AI-powered B2B sales workspace; "AI agents for both sides that manage deals almost entirely on their own"; serves 70k salespeople / ~1M buyers monthly (Deel, Similarweb clients). Pain points (inferred): reliability & control of autonomous deal-managing agents at scale; visibility into agent behavior/cost. Challenges: reliability at scale, cost/visibility per run. Must-haves: reliability, control. Nice-to-haves: cost-per-run visibility. ICP confidence: Medium (CTO, ships agents; 55 emp — just above the 50 threshold).
Signal: 1/4 (CTO shipping medical AI documentation + agentic workflows at massive scale). Source: https://www.mobihealthnews.com/news/heidi-health-raises-65m-expand-global-reach-its-ai-medical-scribe-platform ; https://techcrunch.com/2025/10/05/heidi-health-raises-65m-series-b-led-by-steve-cohens-point72 ; heidihealth.com bio. Company size: ~200-300 (Series B $65M Oct 2025 led by Point72; $100M+ total; 2M+ weekly consultations; enterprise rollout at Beth Israel Lahey Health). Pain points: scaling AI clinical documentation to millions of consults reliably; inference cost at scale; production accuracy/reliability. Challenges: enterprise-scale reliability, building own models, cost control at 2M+ weekly consults. Must-haves: reliability + cost control at scale. Nice-to-haves: own-model efficiency, wearable hardware. ICP confidence: Medium-High (Series B AI-native healthcare shipping agents at scale). NOTE: LinkedIn vanity URL not verified (common name); Crunchbase profile used as profile_url.
Signal: 4 (ICP at a company with unusually well-documented autonomous agents live in a regulated production deployment). NOTE: LinkedIn/Chrome unavailable this run — found via company site + Japanese press. Japanese name: 宇波 雄平. Appointed CTO Oct 2023.
Source: https://about.karakuri.ai/ (board list: 執行役員 CTO 宇波 雄平) ; https://about.karakuri.ai/news/mitsui-direct-voice-agent
Company size: 57 (Japanese company databases — initial.inc / NIKKEI COMPASS / Baseconnect). RIGHT AT the 50-employee ICP floor. Tokyo, Japan. ¥1.56B (~$10M) cumulative — Series B equivalent. Selected for METI/NEDO GENIAC Phase 3 national LLM program.
Agents in production: VERY STRONG — arguably the best-documented agent deployment found this run. KARAKURI's voice agent went FULLY LIVE at Mitsui Direct General Insurance's customer centre in April 2026. Published two-month results: 52.7% of after-hours (post-18:00) inquiries resolved autonomously; ZERO customer complaints as of May 2026; elderly customers completing full AI conversations. 150+ enterprise customers incl. SBI Securities, Daiwa Securities, Hoshino Resorts. Also shipped KARAKURI VL, Japan's first Computer-Using Agent model. Eng team publishes on Zenn (zenn.dev/p/karakuri_blog).
Pain points: [evidenced, company-level — no first-person Uba quote found] Three concrete signals: (1) They sell on OUTCOME-BASED pricing (成果報酬型) explicitly because per-minute voicebot pricing misaligns vendor and customer incentives — i.e. they have already re-architected their business model around agent run economics. (2) They ground RAG on the customer's existing chatbot knowledge base specifically to STRUCTURALLY bound hallucination risk in a regulated insurance context. (3) They state all-hours autonomous resolution is still climbing toward a 60-70% goal — so autonomy rate is a tracked, unmet metric.
Challenges: Hallucination containment in a regulated insurance deployment; pushing autonomous-resolution rate from 52.7% toward 60-70%; speech-to-speech latency and interruption/recovery handling; outcome-based pricing means agent inefficiency directly destroys their own margin.
Must-haves: Cost-per-successful-outcome measurement (their pricing model demands it); hallucination guardrails provable to insurance regulators; autonomy-rate tracking per inquiry type.
Nice-to-haves: Per-conversation token/cost breakdown; regression testing across voice model upgrades.
ICP confidence: Medium-High. Title and production evidence are excellent and specific. Downgraded because company sits at the very bottom of the size band (57) and is modestly funded, and there is no first-person technical quote from Uba himself.
Yuki Matsumoto — Representative Director & CTO, LayerX
Signal: 4 (ICP writing publicly about building and shipping agents). Found via web research — LinkedIn/Chrome unavailable this run.
Source: https://note.com/y_matsuwitter/n/nf039c93ed60d?hl=en (first-person post, 23 Apr 2026)
Company size: 717 employees as of Mar 2026 (Revelio Labs; up from 377 in 2025; engineering = 44.7% of workforce). Series B, $100M (Sept 2025, TechCrunch). Tokyo, Japan. Also a board director of the Japan CTO Association.
Agent evidence: LayerX ships Bakuraku (back-office suite), Ai Workforce, and AgenticSec. Shipped AgenticSec Pentest — a fully automated LLM-agent penetration testing framework (patent pending) that takes only an IP address as input and gains internal access without human intervention, constructing attack scenarios from target config and prior diagnostic results rather than just running tools. Targets ¥100B ARR by FY2030 with roughly half from the AI agent business. 15,000+ enterprise customers.
Verbatim: "With the spread of AI-driven code generation, development speed itself has increased. On the other hand, the mechanisms to ensure that 'the code written is secure' have not yet caught up... As the speed of writing code increases, downstream security verification becomes a bottleneck."
Pain points: downstream verification is now the bottleneck, not code authoring; expanding attack surface as DX increases digital touchpoints; severe security talent shortage (~110,000 people short in Japan per METI); pentesting too expensive and skill-gated to run more than annually.
Challenges: making agent-run pentests reliable enough to replace expert human judgment; integrating agents with security tooling customers already own; scaling three product lines across an engineering org that nearly doubled in a year; keeping agent output trustworthy for 15,000+ enterprise customers.
Must-haves: agents that plan and execute multi-step adversarial workflows autonomously (not scripted automation); integration with the customer's existing security stack; frontier-model access; Japanese-enterprise-grade reliability and compliance.
Nice-to-haves: high-frequency (vs annual) diagnostic cadence; reuse of past diagnostic results as agent context; M&A-friendly tech integration (LayerX is acquiring to expand agent surface area).
ICP confidence: High — Series B, 717 employees, CTO personally shipping and writing about production agent products. Rare Japan-region ICP contact; library is US/EU-heavy.
Yunjing Ma — VP of AI Engineering, Research & Applied AI, Talkdesk
Signal: 4 (ICP technical leader shipping agents). NEW account for the brain (Talkdesk not previously present). Found via LinkedIn people search ('VP Engineering AI agents') + verification. Source: https://www.linkedin.com/in/yunjing-ma-99637819/ | Company: Talkdesk — CX platform shipping Voice AI Agents + an AI Agent Platform (Autopilot/Ascend); her profile explicitly lists 'Voice AI Agents, the AI Agent Platform'. Company size: ~1,300. NOTE: later-stage than the Series A-C target, but within the 50-2,000 size band and clearly shipping agents. Pain points (inferred from role/product, not verbatim quotes): scaling voice agents across enterprise contact centers; cost per interaction/run; production reliability; model benchmarking/eval at scale. Challenges: cost control across large agent fleets; observability; latency. Must-haves: cost-per-run visibility, reliability, eval/benchmarking. Nice-to-haves: model routing, drift monitoring. ICP confidence: Medium — size + agent-shipping confirmed; company is later-stage than the stated Series A-C preference (noted).
Signal: 4 (ICP writing publicly about how his company builds/ships AI agents in production). Source: https://www.linkedin.com/in/yunyu-lin/ , https://ramp.com/authors/yunyu-lin , LinkedIn post "How Ramp builds customer-first AI". Company size: ~1000-1500 (Ramp, fintech/spend management, late-stage, agents in production across capital planning, variance analysis, board reporting, financial close). Former co-founder/CEO of Cohere.io (acquired by Ramp) — technical. Pain points: Ramp's own finance org is heavily reliant on agents in production; getting agents to complete complex multi-step financial workflows accurately and safely; scaling agent usage across the company. Challenges: building customer-first, reliable AI agents on real financial data; controlling agent behavior with the controls finance teams need. Must-haves: reliability/safety of agent actions on money-adjacent workflows, accuracy at scale. Nice-to-haves: cost/token efficiency and per-run cost visibility (Ramp itself sells "token spend management"). ICP confidence: High (Head of Applied AI / technical decision-maker at in-band company shipping many agents in production).
Signal: 4 (ICP technical co-founder/CTO at agent-native company shipping production agents in regulated industries). Source: https://www.globenewswire.com/news-release/2026/03/25/3262061/0/en/Notch-Raises-30-Million-to-Bring-Production-Ready-AI-Agents-to-Regulated-Industries.html ; https://en.globes.co.il/en/article-ai-agents-for-regulated-industries-co-notch-raises-30m-1001538452 ; https://fintech.global/2026/03/26/notch-raises-30m-series-a-for-ai-agents-in-insurance/ . Company size: ~50 employees (dev centers in Israel + US). Stage: Series A ($30M led by Headline, Mar 2026; $45M total). Independent. Actively shipping agents: platform deploys AI agents that automate end-to-end operational workflows for insurance/finance — conversational broker/policyholder interactions plus back-office claims handling, underwriting support, document ingestion/classification/routing; markets "auditable, production-ready AI agents for regulated industries"; ARR grew 12x over 12 months with adoption at leading global insurers. Founders: CEO Rafael Broshi, CPO Elool Jacoby, CTO Yuval Peled. Pain points (INFERRED from positioning, NOT verbatim): making multi-step agents reliable and auditable in high-stakes regulated workflows; provable correctness/traceability of agent actions on claims/underwriting; scaling many agents while keeping compliance. Challenges: reliability + auditability of agents acting on regulated financial/insurance data. Must-haves: reliability, auditability/traceability, control. Nice-to-haves: per-run cost visibility, faster agent deployment. ICP confidence: High (named technical co-founder & CTO, 50 emp = at target floor, Series A, agents in production).
Signal: 1 (ICP publicly speaking about agent context cost and accuracy). NOTE: LinkedIn/Chrome not connected this run; found via published interview. LinkedIn URL not verified.
Source: https://diginomica.com/forget-ai-data-readiness-its-about-architecture-k2views-cto-mcp-context-and-how-get-enterprise-ai (diginomica interview)
Company size: ~191 employees (Tracxn, Apr 2026). Series A $28M (2020) + ~$15M (2025), $68.9M total. Israel. Data product platform, now shipping MCP-based agent context tooling.
What he said (verbatim, short): "Sometimes data that's 30 days old is still fresh."
Pain points: agents "not connected to anything"; over-fetching fresh context is wasted spend; ungoverned data hurting agent accuracy.
Challenges: moving agents from a hard-won ~80% accuracy to ~97% on discrete workflows; enforcing security boundaries LLMs cannot cross.
Must-haves: a semantic/modeling layer over source systems; MCP-based governed data access; context freshness policy tied to cost.
Nice-to-haves: smaller dedicated models routed to specific sub-tasks.
ICP confidence: High — CTO, headcount and stage in band, Israel (under-covered), and his public thesis is explicitly that context retrieval policy is a cost lever.
Zac Gottschall — Director of Engineering, EliseAI
Signal: 4 (ICP senior technical leader at a 50-2,000-employee company actively shipping AI agents; surfaced via LinkedIn People search after Signal 1-3 content searches returned mostly non-ICP consultants/vendors). Source: https://www.linkedin.com/in/zgottschall/. Company size: ~400+ (EliseAI; conversational AI leasing/healthcare agents in production). Pain points (inferred, NOT verbatim): scaling conversational-agent infrastructure reliably; per-conversation cost/latency; production monitoring. Challenges: reliability + cost at scale. Must-haves: production reliability + cost visibility per run. Nice-to-haves: fleet observability. ICP confidence: High (Director of Engineering).
Zach Lloyd — Founder & CEO (technical co-founder), Warp
Signal: 4 (technical founder shipping agents in production) + Signal 3-adjacent (dev-tooling vendor whose product runs many agents). Source: https://en.wikipedia.org/wiki/Warp_(terminal) ; https://www.thetwentyminutevc.com/zach-lloyd . Company size: ~102 employees (Tracxn, Apr 2026); Series B $50M led by Sequoia (Jun 2023), ~$73M total; founded 2020 by Zach Lloyd (ex-Google Principal Engineer, ex-interim CTO at Time). Pain points (inferred from public product/positioning; no first-party quotes captured): Warp 2.0 'Agentic Development Environment' deploys multiple long-running coding agents concurrently (editing code, running commands, managing workflows) — token/compute cost of many concurrent agents; reliability of autonomous edits/commands; visibility into concurrent agent activity. Publicly discusses adding ~$1M ARR/week and competing with OpenAI Codex / Claude Code (scaling pressure). Challenges: cost efficiency of running many concurrent agents; reliability of autonomous command execution; infra scale. Must-haves: per-agent-run cost visibility/control, reliability guardrails, observability across concurrent agents. Nice-to-haves: compounding agent quality over time. ICP confidence: High-Medium (technical founder-CEO, agents in production, Series B, ~102 emp; caveat: dev-tooling vendor — build-vs-buy/positioning review before outreach).
Zachary Lipton — Co-Founder & CTO, Abridge
Signal: 1 & 4 (ICP publicly writing/speaking about agentic transformation; ICP building & shipping agents). Source: https://www.bloomberg.com/news/audio/2026-03-26/vanguards-of-health-care-abridge-looking-beyond-the-scribe ; https://theorg.com/org/abridge/org-chart/zachary-lipton. Company size: ~605-635 employees (May-June 2026, Series E, ~$5.3B val). Pain points: moving beyond single-purpose ambient scribe to an integrated multi-agent clinical platform; dynamically composing multiple agent skills; trust/reliability of generative AI in high-stakes clinical workflows. Challenges: orchestrating end-to-end clinical agent workflows (visit prep, prior-auth checks, billing, care decisions) reliably; replacing fragmented point solutions with one governed platform. Must-haves: reliability & trust guardrails for clinical agents; visibility and control over multi-skill agent behavior. Nice-to-haves: dynamic skill composition; real-time guidance across documentation/billing. ICP confidence: High (Co-founder & CTO + CMU ML professor, ~620-employee Series E company explicitly pursuing 'agentic transformation' in 2026).
Zachary Tosh — Director of Engineering, Forethought
Signal: 4 (Director of Engineering at an in-band multi-agent AI company; identified via Forethought LinkedIn company People page filtered to engineering). Source: https://www.linkedin.com/in/zachary-tosh-1a014511a . Company size: 51-200 employees (LinkedIn; ~98 per Revelio 2026); Forethought ships a multi-agent omnichannel CX AI platform in production. Pain points (inferred from role; no direct post/quote read this run): owning engineering for production agents => reliability in prod, token/context waste, and no clean per-run cost attribution. Challenges: scaling from a few agents to a fleet without cost/latency blowout. Must-haves: cost-per-run visibility, reliability guardrails, observability. Nice-to-haves: model routing / caching to cut token spend. ICP confidence: High (Director of Engineering; in-band agent-native company; pains inferred not quoted).
Signal: 1 (frequent public speaker/podcast guest on building at scale with small team) + 4 (technical co-founder shipping agents). Source: https://www.crunchbase.com/person/zachary-ziegler-109f ; https://sequoiacap.com/founder/zachary-ziegler/ ; https://www.youtube.com/watch?v=sZO1_YJWD-4 (podcast). LinkedIn URL not confirmed in search — do not fabricate. Company size: ~119 employees (May 2026); raised ~$210M at $3.5B valuation; fastest-growing physician app, 40%+ of US physicians, 17M monthly clinical queries. Building agentic features: auto-drafting clinical notes/discharge summaries, generating prior-authorization letters with evidence. Pain points: reliability/accuracy in high-stakes point-of-care decisions; cost/inference scaling at 17M queries/month; small team operating enormous agent volume. Challenges: correctness and citation/grounding in medical answers; controlling cost as query volume compounds. Must-haves: reliable grounded output, cost control at scale. Nice-to-haves: per-query/per-agent cost visibility. ICP confidence: High — CTO/co-founder, ~119 emp growth-stage, agentic product at massive production scale; cost + reliability pain both present. HIGH PRIORITY.
Zack Lipton — CTO & Co-founder, Abridge
Signal: 1 (ICP publicly speaking/writing about shipping agents in production). Source: https://www.statnews.com/2026/02/10/abridge-ai-scribe-cto-talks-epic-microsoft-rebranding/ and Bloomberg Intelligence "Vanguards of Healthcare" podcast (Mar 2026). Company size: ~450 employees (per StatNews "450-person enterprise"), Series-stage AI-native healthcare, live across 200+ hospital systems. Pain points: moving agents from ambient scribing to full "agentic transformation of the workflow"; achieving production-grade reliability in regulated clinical environments; replacing fragmented point solutions with integrated agent platform. Challenges: agent reliability/accuracy across 200+ hospital systems and 100+ specialties; integrating agentic workflows into Epic/EHR; scaling multi-step reasoning agents at hospital scale. Must-haves: production-grade reliability and accuracy in clinical settings; visibility and control over agent behavior. Nice-to-haves: cost efficiency of multi-cycle agent reasoning. ICP confidence: High — technical co-founder/CTO and decision-maker at a ~450-person AI-native company actively shipping clinical agents in production. Profile URL left blank (not fabricating LinkedIn URL).
Zack Reneau-Wedeen — Head of Product, Sierra
Signal: 4 (AI-focused product leader publicly speaking/writing about building enterprise AI agents). Source: https://www.langchain.com/blog/why-the-best-agents-are-simpler-than-you-think-sierra-max-agency-podcast ; https://sierra.ai/author/zack-reneau-wedeen ; https://www.linkedin.com/in/zackrw/. Leads Sierra's enterprise AI agent product — Agent Studio 2.0, Agent Data Platform — focused on richer workflows, persistent memory, voice. Ex-Google (founding PM Google Lens/Podcasts), ex-Robinhood, ex-CoinTracker. Company size: ~500-600 (Sierra, $15B). Pain points: making enterprise agents reliable and simple at scale; persistent memory/context and high-quality voice; delivering measurable outcomes in production. Challenges: richer multi-step workflows without sacrificing reliability; agent data/context management. Must-haves: reliability, context/memory management, observability of outcomes. Nice-to-haves: simpler agent architectures, cost efficiency. ICP confidence: High (VP/Head of Product, AI-focused, at 50-2,000-emp agent-shipping company).
Signal: 4 (ICP technical leader at AI-native company shipping agents in production). Source: https://www.anterior.com/insights/anterior-raises-40m-series ; https://medcitynews.com/2026/02/anterior-ai-health-plan/ . Company size: est. ~60–120 (Series B, $40M round Feb 2026, $64M total raised; ~40% of team are clinicians). Actively building AI agents: runs "Florence," an AI agent that automates medical prior-authorization/utilization-management workflows for health plans covering 50M+ members. Pain points (inferred from domain + 2026 market signal, NOT a verbatim quote): production reliability and clinical accuracy of agents at scale; cost per authorization run rising as volume grows; auditability of agent decisions in a regulated payer setting. Challenges: scaling one agent across many payer workflows while keeping outputs compliant and explainable. Must-haves: reliability + explainability of agent decisions; visibility into cost/behavior per run. Nice-to-haves: agents that compound/improve over time. ICP confidence: High (technical co-founder/CTO, Series B AI-native agent company; only the exact headcount is an estimate to confirm).
Zayd Enam — Co-Founder & CEO, Cresta
Signal: 4 (ICP senior technical leader at a company actively shipping AI agents; found via LinkedIn people-search, not an observed post). Source: https://www.linkedin.com/in/zaydenam/ | Company size: ~350-500 (Series D; real-time AI agents & coaching for contact centers). Pain points (INFERRED from role + company product focus — no observed quote): scaling agents reliably in production; controlling per-run LLM/agent cost as volume grows; visibility into what agents cost and how they behave across customers. Challenges: keeping agent accuracy/guardrails consistent at scale; token/context efficiency; cost blowout as they move from pilots to broad production. Must-haves: production reliability + observability into agent cost and behavior. Nice-to-haves: per-run cost attribution, context/token optimization, spend guardrails. ICP confidence: High (Technical co-founder (Stanford AI PhD); agent-native company squarely in size range).
Zexia Zhang — Co-founder & CTO, Retell AI
Signal: 4 (ICP technical co-founder/CTO shipping voice agents in production at massive scale). Source: https://www.retellai.com/about-us ; Tracxn (headcount ~128 as of May 2026) ; https://www.retellai.com/blog/seed-announcement. Company size: ~128. Stage: seed-funded but at scale — reported ~$40-50M ARR, powering 50M+ real-time AI phone calls/month. Actively shipping agents: Retell is a voice-agent builder platform running production phone agents for contact centers at very high call volume. Pain points (inferred from product + domain, NOT verbatim quotes): per-call cost and latency of real-time voice agents at 50M+ calls/month; reliability of agents in production; visibility into what each agent run costs. Challenges: keeping voice agents reliable and cheap at extreme call volume; observability across agents. Must-haves: cost-per-run visibility, reliability, low latency. Nice-to-haves: efficiency gains that compound at scale. ICP confidence: Medium (technical co-founder/CTO; ~128 employees fits band; ships agents in production at scale; stage labelled seed but operating well beyond seed in substance).
[IGNORE - test row, safe to delete] — -, -
Accidental API-probe row created by the ICP Prospect Signal Scanner run on 2026-07-21 while discovering the profileUrl field name. Not a prospect. The API exposes no DELETE endpoint, so this row was blanked instead. Safe to remove manually.