## ALIGNMENT FLAGS
**Critical tension — Task 20 vs Thesis 2:** Task 20 says "rewrite positioning: NOT a gateway/cost tool — lead with reliability + compounding harness." Thesis 2 says "cost optimization is the wedge; compounding is the moat — sell the painkiller." These directly contradict. No copy can be written, no landing page built, until this is resolved. One positioning. Commit.
**Duplicate task bloat:** Tasks 11/14 (Arena 3-step aha), 12/15 (reposition Arena copy), 13/16 (expansion revenue tiers) are exact duplicates — 6 tasks that are really 3. One set should be closed.
**Open question vs closed decision:** Question #1 ("What ICP to target?") is still status:open. Decision entry #29 locked ICP on July 5. Close the question.
**Challenge #1 due tomorrow (July 7)** with zero people assigned — yet ICP was locked in entry #29. Update challenge to closed or document why it stays open.
## OVERDUE & UNEXPLAINED
No tasks are past due today. Challenge #1 expires tomorrow — action required today.
## VALIDATION FINDINGS
**1. Pricing ($99/$499/Enterprise BYOK tiers) — CAUTION FLAG:** BYOK market moved toward free/cheap in 2026. Portkey went Apache 2.0 open-source (March 2026), LiteLLM MIT-licensed with 40K+ stars, Helicone absorbed into Mintlify. Entry BYOK tooling is now $0–$79. Alpha's $99 entry tier is 3-5x more expensive than alternatives with no proven differentiation yet. Pricing is untested. Do not anchor publicly until 3 prospect conversations validate willingness to pay.
**2. ICP (50–500 employees, actively shipping agents) — PARTIALLY VALIDATED:** Only 42% of 50–499 employee firms have deployed any AI at all (vs 83% of large enterprise). Mid-market has the highest agent abandonment rates. The "actively shipping" qualifier narrows the addressable universe significantly — smaller TAM than thesis assumes. Pain is real but segment is thin. Outreach must qualify hard for active production agents, not "exploring AI."
**3. Gateway competition — CONFIRMED COMMODITIZED:** Portkey open-source March 2026, LiteLLM 40K stars, Helicone acquired and dissolving into Mintlify. Standalone gateway/observability is not a defensible category. Validates the harness+compounding positioning — but only after Flag #1 tension is resolved.
## WHO TO CONTACT
Challenge #1 (ICP, due tomorrow): Ravi Sindri at Qualizeal (ravikiran@qualizeal.com) — he is actively implementing agentic solutions for clients and IS the ICP. One direct buyer conversation validates or kills the mid-market thesis faster than any research. Ask: "Would you pay for a harness that compounds agent quality over time?"
Outreach tasks 23/24 (Polu at Dust.tt, Raghunathan at Hyperbound): both high-priority, both undated — they will drift. Assign due dates today.
## PATTERNS TO FIX
**Content at zero.** Three days post-AIBoomi. Zero LinkedIn posts shipped. Flagship POV (task 7) untouched. PLG engine is 100% dependent on content driving Arena traffic — nothing is moving.
**Undated high-priority tasks.** Tasks 23 and 24 (first prospect outreaches) are marked high priority with no due date. Undated priorities don't execute.
**$0 ARR, $10M clock running.** No prospect conversations, no content, no product shipped post-AIBoomi. Operations, Partnerships, and Content pillars show zero forward motion since July 4.
## TOP 3 NEXT ACTIONS
**Vishnu (tied to $10M):**
1. DM Stanislas Polu (CTO @ Dust.tt) today — first qualified prospect conversation unlocks real pricing and ICP signal
2. Post flagship POV on LinkedIn — "agent costs are going out of control" — break the content silence
3. Resolve Thesis 2 vs Task 20 tension in writing — one positioning, update brain
**Anu (tied to $10M):**
1. Load July 5 prospect list into CRM today and set due dates on outreach tasks 23/24
2. Schedule a call with Ravi Sindri (Qualizeal) — ICP validation from a live buyer
3. Park task 9 (partnership tracks) — confirmed misaligned with 12-month PLG thesis; revisit at $1M ARR
Experiment #1 interim update — LLM cost pain signal
Hypothesis: Everyone building agents wants to reduce their LLM costs, move to open source where possible, but don't know how.
INTERIM VERDICT: Hypothesis is partially true but needs significant refinement. The pain is real; the mechanism is narrower than stated.
WHAT THE RESEARCH SUPPORTS:
1. Cost pain is confirmed real. Research Brief #1 (LLM cost brief) validated that agent builders do experience meaningful cost pressure. Thesis entry #3 and content entry #10 ("agent costs are going out of control") reflect observed market sentiment.
2. The pain is shifting from model cost to total run cost. Validation flag (Entry #23, July 5): LLM API prices fell ~80% from early-2025 to early-2026. Inference is now only 30–45% of total agent run cost (was 70–80%). Pure model-switching savings are a shrinking wedge. The real pain is total agent operating cost — routing, caching, eval, human review combined — not just model tokens.
3. "Move to open source" motivation is weakening. The 80% API price drop means proprietary models (GPT-4o, Claude) are now cheap enough that open-source migration is not the obvious move it was 18 months ago. The compelling reason to switch is control and compounding, not raw model cost.
4. "Don't know how" is partially true but being eroded. Free tools exist: Netflix Headroom (OSS, launched Jan 2026) delivers 60–95% token reduction for free; Helicone is free up to 10K req/mo; Langfuse is self-hostable free. These cover the "don't know how to reduce model cost" gap. Alpha's edge must be the operating layer above the cost meter — control, reliability, compounding — not a cheaper version of free tools.
DISTRIBUTION SIGNAL (17 Arena runs targeting n8n builders, cost-saving messaging):
No conversion data yet in the brain. The runs establish that the n8n builder segment is reachable and that cost-saving messaging is being served. Absence of conversion data is not confirmation of failure — it is the expected state at 17 runs with no aha-moment funnel yet built. The Arena 3-step aha flow (Decision #21) is not shipped; distribution without the product hook underweights the hypothesis test.
WHAT THIS MEANS FOR THE EXPERIMENT:
The hypothesis should be refined: the real pain is not "reduce LLM costs via open source" but "regain control over total agent run cost and prevent cost blowout at scale." Prospect scan (Brief #3) found 6 high-confidence ICP fits (Dust.tt, Artisan AI, Ema AI, Voiceflow, Hyperbound, Lindy AI) — all spending meaningfully on agent infra — but none were reached via cost messaging directly. The signal from the ICP decision (Decision #29) and the greenfield/brownfield research (Brief #4) suggests the stickiest buyers are teams stalled at the 1→5 agent scale wall, where cost blowout and operational chaos happen simultaneously. Cost is the entry point; control is the reason they stay.
RECOMMENDED NEXT STATE: Keep experiment running. Ship Arena's 3-step aha flow (baseline → optimized → routed cost) as the minimum viable test of whether cost-revelation converts browsers to $99/mo subscribers. Only then can the hypothesis be properly evaluated against actual user behavior.
Pricing: BYOK subscription — $99/mo (up to 5 agents), $499/mo (up to 15 agents), Enterprise (above 15)
Locked pricing tiers for thealpha.ai. All plans are BYOK (Bring Your Own Key) only — no hosted LLM spend billed through Alpha.
Tiers:
- Basic: $99/mo — up to 5 agents
- Professional: $499/mo — up to 15 agents
- Enterprise: custom pricing — above 15 agents, sold via direct sales
No per-seat pricing. No usage-based billing. Subscription is pure agent-count. BYOK means the customer brings their own API keys; Alpha never marks up LLM costs.
Research: Greenfield vs brownfield — who is Alpha's ICP?
## Question
Is Alpha's ICP greenfield AI-native companies, or brownfield incumbents onboarding AI? Vishnu's thesis: greenfield firms may already have built agent harnesses, so the incumbents "figuring it out" are the ones that need help.
## Verdict
Partially validated. The instinct that struggling teams need the harness is correct, but the greenfield/brownfield binary is the wrong segmentation axis and, taken literally, would steer Alpha toward the worst-fit buyers (slow legacy enterprises) while writing off some of its best PLG buyers (AI-forward mid-market teams with brittle homegrown scaffolding).
## Evidence
1) INCUMBENTS ARE GENUINELY STUCK — supports the thesis.
- 86-88% of enterprise agent pilots never reach production; ~60% of enterprises stall specifically in the jump from one pilot to 5-20 production agents. Failures cluster on governance, data-readiness and observability, not model quality. Sources: https://agentmarketcap.ai/blog/2026/04/11/enterprise-agent-deployment-maturity-model-2026 , https://www.institutepm.com/knowledge-hub/why-enterprise-ai-pilots-fail , https://www.digitalapplied.com/blog/ai-agent-adoption-2026-enterprise-data-points
- ~60% of AI leaders cite legacy-system integration as their #1 agentic blocker; 82% struggle with data standardization/compatibility. Brownfield AI "lands on top of legacy," so integration refactoring — not model selection — is the real blocker. Sources: https://medium.com/@manjeerachandarao/why-brownfield-integration-is-the-hard-part-of-ai-adoption-179fbfd87915 , https://www.v2soft.com/blogs/modernize-legacy-applications-ai
2) BUT "GREENFIELD ALREADY BUILT A HARNESS = NOT A CUSTOMER" IS LARGELY WRONG.
- Only the most serious AI-natives built durable internal harnesses (LangGraph/MCP orchestration, cron/heartbeat/sub-agent tooling; e.g. Context Studios runs 16 production cron agents). That is a minority. Sources: https://www.contextstudios.ai/guides/ai-agents-business-automation-2026 , https://viston.tech/ai-agent-orchestration-in-2026-moving-from-pilots-to-enterprise-wide-execution/
- Most teams wired brittle LangChain/LlamaIndex glue that is now being abandoned: better native tool-calling + MCP standardization removed the reason for heavyweight frameworks, and hidden run/maintenance costs exceed license fees within ~6 months. ~90% of enterprise use cases now favor BUY over build. Sources: https://www.mindstudio.ai/blog/llm-frameworks-replaced-by-agent-sdks , https://www.oreilly.com/radar/the-ai-agents-stack-2026-edition/ , https://aisera.com/blog/build-vs-buy-ai/ , https://composio.dev/content/build-vs-buy-ai-agent-integrations
- Greenfield/AI-native teams are the FASTEST adopters and highest-WTP buyers of exactly this category: Braintrust raised $80M Series B at $800M (Feb 2026); Respan/Keywords AI serves 100+ AI startups (2T+ tokens/mo). Sources: https://www.getmaxim.ai/articles/5-ai-observability-platforms-compared-maxim-ai-arize-helicone-braintrust-langfuse/ , https://www.landbase.com/blog/fastest-growing-observability-platforms
3) THE COST/OBSERVABILITY WEDGE IS REAL AND MID-MARKET-SHAPED — validates Alpha's positioning.
- Eval/observability is the #1 production blocker for 64% of teams and the hottest budget line of 2026. Mid-market spends ~$310k/yr on eval+observability (vs $2.4M Fortune 500). Classic surprise: a $1,000/mo estimate arrives as a ~$3,800 invoice (planning overhead, 18-44% tool-call retry rates, memory writes). Sources: https://guptadeepak.com/ai-agent-observability-evaluation-governance-the-2026-market-reality-check/ , https://firstpagesage.com/reports/agentic-ai-adoption-statistics/ , https://ranksquire.com/2026/05/04/what-are-ai-agents-in-2026/
- Comparable tools price at ~$300-1,200/mo (Helicone/LangSmith), so Alpha's ~$250/mo BYOK wedge sits at the low, self-serve end of an established willingness-to-pay band. Sources: https://tokenmix.ai/blog/langsmith-vs-helicone-vs-braintrust-observability-2026 , https://www.openhelm.ai/blog/langsmith-vs-helicone-vs-braintrust-llm-observability
4) CONTRARIAN / DISCONFIRMING EVIDENCE.
- Large brownfield enterprises have the most acute pain but are the WORST PLG fit: slow procurement, security review, zero-trust/audit gaps, "integration-refactoring-first" adoption — an Enterprise-tier sales motion, not $250/mo self-serve. A Forbes contrarian argues much agentic tooling targets "enterprises that don't exist." Source: https://www.forbes.com/councils/forbestechcouncil/2026/04/01/agentic-ai-is-being-built-for-enterprises-that-dont-exist/
## Implications for Alpha
- The productive axis is production-maturity + team-capability, not greenfield vs brownfield. The buyer is defined by "actively shipping agents, stalled scaling them, no platform team to build a harness."
- Sweet spot = "brownfield-lite" mid-market (the locked 50-500-employee ICP): past prototype, hitting the 1->5-20 agent wall, feeling cost/observability pain, without a dedicated agent-infra team. This aligns cleanly with the cost wedge + compounding moat theses.
- Pure greenfield harness-builders: small, hard to displace — deprioritize as a primary target (but reachable via the cost wedge when their homegrown stack gets expensive).
- Legacy giants: Enterprise-tier, sales-led, later — do not let them define the PLG ICP.
## Recommended actions
- Refine the ICP pillar: replace greenfield/brownfield framing with a maturity+capability definition ("50-500 employees, shipping agents in production, stalled at scale, no dedicated agent-platform team").
- Build GTM content around the cost-shock and 64% observability-blocker stats (Anu's money-saved lane).
- Consider an outbound list of teams abandoning homegrown LangChain harnesses (build->buy switchers).
agent reviewdaily-review-agent · 6 Jul 2026
Daily Brain Review — July 6 2026
## ALIGNMENT FLAGS
**Critical tension — Task 20 vs Thesis 2:** Task 20 says "rewrite positioning: NOT a gateway/cost tool — lead with reliability + compounding harness." Thesis 2 says "cost optimization is the wedge; compounding is the moat — sell the painkiller." These directly contradict. No copy can be written, no landing page built, until this is resolved. One positioning. Commit.
**Duplicate task bloat:** Tasks 11/14 (Arena 3-step aha), 12/15 (reposition Arena copy), 13/16 (expansion revenue tiers) are exact duplicates — 6 tasks that are really 3. One set should be closed.
**Open question vs closed decision:** Question #1 ("What ICP to target?") is still status:open. Decision entry #29 locked ICP on July 5. Close the question.
**Challenge #1 due tomorrow (July 7)** with zero people assigned — yet ICP was locked in entry #29. Update challenge to closed or document why it stays open.
## OVERDUE & UNEXPLAINED
No tasks are past due today. Challenge #1 expires tomorrow — action required today.
## VALIDATION FINDINGS
**1. Pricing ($99/$499/Enterprise BYOK tiers) — CAUTION FLAG:** BYOK market moved toward free/cheap in 2026. Portkey went Apache 2.0 open-source (March 2026), LiteLLM MIT-licensed with 40K+ stars, Helicone absorbed into Mintlify. Entry BYOK tooling is now $0–$79. Alpha's $99 entry tier is 3-5x more expensive than alternatives with no proven differentiation yet. Pricing is untested. Do not anchor publicly until 3 prospect conversations validate willingness to pay.
**2. ICP (50–500 employees, actively shipping agents) — PARTIALLY VALIDATED:** Only 42% of 50–499 employee firms have deployed any AI at all (vs 83% of large enterprise). Mid-market has the highest agent abandonment rates. The "actively shipping" qualifier narrows the addressable universe significantly — smaller TAM than thesis assumes. Pain is real but segment is thin. Outreach must qualify hard for active production agents, not "exploring AI."
**3. Gateway competition — CONFIRMED COMMODITIZED:** Portkey open-source March 2026, LiteLLM 40K stars, Helicone acquired and dissolving into Mintlify. Standalone gateway/observability is not a defensible category. Validates the harness+compounding positioning — but only after Flag #1 tension is resolved.
## WHO TO CONTACT
Challenge #1 (ICP, due tomorrow): Ravi Sindri at Qualizeal (ravikiran@qualizeal.com) — he is actively implementing agentic solutions for clients and IS the ICP. One direct buyer conversation validates or kills the mid-market thesis faster than any research. Ask: "Would you pay for a harness that compounds agent quality over time?"
Outreach tasks 23/24 (Polu at Dust.tt, Raghunathan at Hyperbound): both high-priority, both undated — they will drift. Assign due dates today.
## PATTERNS TO FIX
**Content at zero.** Three days post-AIBoomi. Zero LinkedIn posts shipped. Flagship POV (task 7) untouched. PLG engine is 100% dependent on content driving Arena traffic — nothing is moving.
**Undated high-priority tasks.** Tasks 23 and 24 (first prospect outreaches) are marked high priority with no due date. Undated priorities don't execute.
**$0 ARR, $10M clock running.** No prospect conversations, no content, no product shipped post-AIBoomi. Operations, Partnerships, and Content pillars show zero forward motion since July 4.
## TOP 3 NEXT ACTIONS
**Vishnu (tied to $10M):**
1. DM Stanislas Polu (CTO @ Dust.tt) today — first qualified prospect conversation unlocks real pricing and ICP signal
2. Post flagship POV on LinkedIn — "agent costs are going out of control" — break the content silence
3. Resolve Thesis 2 vs Task 20 tension in writing — one positioning, update brain
**Anu (tied to $10M):**
1. Load July 5 prospect list into CRM today and set due dates on outreach tasks 23/24
2. Schedule a call with Ravi Sindri (Qualizeal) — ICP validation from a live buyer
3. Park task 9 (partnership tracks) — confirmed misaligned with 12-month PLG thesis; revisit at $1M ARR
## Methodology
Scanned for Series A/B companies (50-500 employees) actively shipping AI agents. Signals: job postings for LLM/AI agent roles, funding data, engineering content, Crunchbase confirmation. Zero overlap with existing Alpha Brain accounts.
## 1. Artisan AI — Confidence 5/5
Headcount: ~168 | Stage: Series A, $46M (Glade Brook, YC, HubSpot Ventures, April 2025)
Agent use case: AI BDR automation. Ava (AI BDR) used by 250+ orgs; expanding to Aaron (Inbound SDR) and Aria (Meeting Assistant). Autonomous AI employees for sales teams.
Decision maker: CTO Ming Li (ex-Deel, Rippling, Google); CEO Jaspar Carmichael-Jack
LinkedIn slugs: ming-li (CTO), jaspar-carmichael-jack (CEO)
Why now: Just closed Series A. LLM inference is direct COGS — cost optimization is a core margin lever at their volume.
## 2. Dust.tt — Confidence 5/5
Headcount: ~144 | Stage: Series B, $61.5M (Abstract + Sequoia, May 2026)
Agent use case: Enterprise multi-agent collaboration platform. 3,000+ orgs, 300K+ deployed agents, 240% NRR. Customers: Alan, Qonto, Payfit. Deep Anthropic Claude integration.
Decision maker: CTO/Co-founder Stanislas Polu (ex-OpenAI researcher, ex-Stripe); CEO Gabriel Hubert
LinkedIn slugs: stanpolu (CTO), gabhubert (CEO)
Why now: Just closed $40M Series B. Anthropic already in stack. Budget available, scale exploding.
## 3. Ema AI — Confidence 5/5
Headcount: ~228 | Stage: Series A, $61M (Accel + Section 32; KPMG strategic minority)
Agent use case: Universal AI employees for enterprise — AI agents for HR, IT helpdesk, CS, sales ops. On-prem deployment. KPMG partnership for Fortune 500 distribution.
Decision maker: CTO/Co-founder Souvik Sen (ex-Okta VP Eng, ex-Google ML); CEO Surojit Chatterjee (ex-Coinbase CPO)
LinkedIn slugs: souvik-sen (CTO), surojitchatterjee (CEO)
Why now: Expanding enterprise + KPMG distribution = scaling fast. Multi-agent workflows = high LLM cost exposure.
## 4. Voiceflow — Confidence 4/5
Headcount: ~88 | Stage: Series A, $39.8M (OpenView Venture Partners, August 2023)
Agent use case: Enterprise AI agent builder platform. Multi-model support (OpenAI, Anthropic Claude, Google). 100K+ developer community. Redesigned around AI credits pricing in April 2025.
Decision maker: CEO Braden Ream (co-founder); CTO Tyler Han (co-founder)
LinkedIn slugs: braden-ream (CEO), tyler-han (CTO)
Why now: Credits-based pricing = LLM cost is their core business variable. Enterprise scale deployment.
## 5. Hyperbound — Confidence 4/5
Headcount: ~51 | Stage: Series A, $18M (Peak XV, September 2025; YC S23)
Agent use case: AI sales roleplay agents. AI buyer simulation agents for sales training. 7,000+ customers across SaaS, financial services, logistics.
Decision maker: CEO Sriharsha Guduguntla; CTO Atul Raghunathan (LLM researcher, ex-enterprise ML)
LinkedIn slugs: sguduguntla (CEO), atul-raghunathan (CTO)
Why now: Recently closed Series A. CTO is hands-on LLM researcher = high receptivity to optimization tools.
## 6. Lindy AI — Confidence 4/5
Headcount: ~52 | Stage: Series B, ~$54M
Agent use case: Personal AI workflow agents — email triage, scheduling, meeting notes, task delegation. Always-on AI chief of staff.
Decision maker: CEO/Founder Flo Crivello (ex-Uber PM, YC)
LinkedIn slug: florentcrivello (CEO)
Why now: Series B PMF signals strong. LLM inference is primary COGS. Founder active on LinkedIn/podcasts — reachable via content.
## Priority Outreach Order
1. Dust.tt (CTO Stanislas Polu) — Anthropic already in stack, 300K+ agents, fresh $40M raise
2. Artisan AI (CTO Ming Li) — highest LLM volume, fresh Series A
3. Hyperbound (CTO Atul Raghunathan) — LLM researcher, small team, ideal technical champion
4. Voiceflow (CTO Tyler Han) — credits-based business = direct LLM cost pressure
5. Ema AI (CTO Souvik Sen) — larger sale but KPMG partnership = scale
6. Lindy AI (CEO Flo Crivello) — reachable via content engagement
Next scan: July 12 2026. Watch: Ema AI Series B signals; Artisan AI LLM job postings; Voiceflow enterprise announcements.
## Methodology
Scanned for Series A/B companies (50-500 employees) actively shipping AI agents. Signals: job postings for LLM/AI agent roles, funding data, engineering content, Crunchbase confirmation. Zero overlap with existing Alpha Brain accounts.
## 1. Artisan AI — Confidence 5/5
Headcount: ~168 | Stage: Series A, $46M (Glade Brook, YC, HubSpot Ventures, April 2025)
Agent use case: AI BDR automation. Ava (AI BDR) used by 250+ orgs; expanding to Aaron (Inbound SDR) and Aria (Meeting Assistant). Autonomous AI employees for sales teams.
Decision maker: CTO Ming Li (ex-Deel, Rippling, Google); CEO Jaspar Carmichael-Jack
LinkedIn slugs: ming-li (CTO), jaspar-carmichael-jack (CEO)
Why now: Just closed Series A. LLM inference is direct COGS — cost optimization is a core margin lever at their volume.
## 2. Dust.tt — Confidence 5/5
Headcount: ~144 | Stage: Series B, $61.5M (Abstract + Sequoia, May 2026)
Agent use case: Enterprise multi-agent collaboration platform. 3,000+ orgs, 300K+ deployed agents, 240% NRR. Customers: Alan, Qonto, Payfit. Deep Anthropic Claude integration.
Decision maker: CTO/Co-founder Stanislas Polu (ex-OpenAI researcher, ex-Stripe); CEO Gabriel Hubert
LinkedIn slugs: stanpolu (CTO), gabhubert (CEO)
Why now: Just closed $40M Series B. Anthropic already in stack. Budget available, scale exploding.
## 3. Ema AI — Confidence 5/5
Headcount: ~228 | Stage: Series A, $61M (Accel + Section 32; KPMG strategic minority)
Agent use case: Universal AI employees for enterprise — AI agents for HR, IT helpdesk, CS, sales ops. On-prem deployment. KPMG partnership for Fortune 500 distribution.
Decision maker: CTO/Co-founder Souvik Sen (ex-Okta VP Eng, ex-Google ML); CEO Surojit Chatterjee (ex-Coinbase CPO)
LinkedIn slugs: souvik-sen (CTO), surojitchatterjee (CEO)
Why now: Expanding enterprise + KPMG distribution = scaling fast. Multi-agent workflows = high LLM cost exposure.
## 4. Voiceflow — Confidence 4/5
Headcount: ~88 | Stage: Series A, $39.8M (OpenView Venture Partners, August 2023)
Agent use case: Enterprise AI agent builder platform. Multi-model support (OpenAI, Anthropic Claude, Google). 100K+ developer community. Redesigned around AI credits pricing in April 2025.
Decision maker: CEO Braden Ream (co-founder); CTO Tyler Han (co-founder)
LinkedIn slugs: braden-ream (CEO), tyler-han (CTO)
Why now: Credits-based pricing = LLM cost is their core business variable. Enterprise scale deployment.
## 5. Hyperbound — Confidence 4/5
Headcount: ~51 | Stage: Series A, $18M (Peak XV, September 2025; YC S23)
Agent use case: AI sales roleplay agents. AI buyer simulation agents for sales training. 7,000+ customers across SaaS, financial services, logistics.
Decision maker: CEO Sriharsha Guduguntla; CTO Atul Raghunathan (LLM researcher, ex-enterprise ML)
LinkedIn slugs: sguduguntla (CEO), atul-raghunathan (CTO)
Why now: Recently closed Series A. CTO is hands-on LLM researcher = high receptivity to optimization tools.
## 6. Lindy AI — Confidence 4/5
Headcount: ~52 | Stage: Series B, ~$54M
Agent use case: Personal AI workflow agents — email triage, scheduling, meeting notes, task delegation. Always-on AI chief of staff.
Decision maker: CEO/Founder Flo Crivello (ex-Uber PM, YC)
LinkedIn slug: florentcrivello (CEO)
Why now: Series B PMF signals strong. LLM inference is primary COGS. Founder active on LinkedIn/podcasts — reachable via content.
## Priority Outreach Order
1. Dust.tt (CTO Stanislas Polu) — Anthropic already in stack, 300K+ agents, fresh $40M raise
2. Artisan AI (CTO Ming Li) — highest LLM volume, fresh Series A
3. Hyperbound (CTO Atul Raghunathan) — LLM researcher, small team, ideal technical champion
4. Voiceflow (CTO Tyler Han) — credits-based business = direct LLM cost pressure
5. Ema AI (CTO Souvik Sen) — larger sale but KPMG partnership = scale
6. Lindy AI (CEO Flo Crivello) — reachable via content engagement
Next scan: July 12 2026. Watch: Ema AI Series B signals; Artisan AI LLM job postings; Voiceflow enterprise announcements.
## Methodology
Scanned for Series A/B companies (50–500 employees) actively shipping AI agents. Signals weighted to last 90 days: job postings for LLM/AI agent roles, funding announcements, public engineering content, Crunchbase/Tracxn confirmation. Cross-checked against existing Alpha Brain accounts — zero overlap, clean slate.
---
## 1. Artisan AI — Confidence 5/5
- Headcount: ~168 employees | Funding: Series A, $46M total (Glade Brook Capital, YC, HubSpot Ventures — April 2025)
- Agent use case: AI Business Development Representatives. "Artisans" are autonomous AI employees handling outbound prospecting, email sequencing, lead qualification. Flagship Ava (AI BDR) used by 250+ orgs. Expanding to Aaron (Inbound SDR) and Aria (Meeting Assistant).
- Decision maker: CTO Ming Li (ex-Deel, Rippling, TikTok, Google) · LinkedIn: ming-li; CEO: Jaspar Carmichael-Jack
- Why now: Just closed Series A, scaling agent workforce product. LLM inference is their direct COGS — cost optimization is a core margin lever at their volume.
## 2. Dust.tt — Confidence 5/5
- Headcount: ~144 employees | Funding: Series B, $61.5M total ($40M Series B Abstract + Sequoia May 2026)
- Agent use case: Enterprise multi-agent collaboration platform — fleets of specialized agents connected to internal data (Notion, Slack, Drive). 3,000+ orgs, 300K+ deployed agents, 240% NRR. Customers: Alan, Qonto, Payfit.
- Decision maker: CTO/Co-founder Stanislas Polu (ex-OpenAI researcher, ex-Stripe) · LinkedIn: stanpolu; CEO: Gabriel Hubert · LinkedIn: gabhubert
- Why now: Just closed $40M Series B. Anthropic Claude already integrated in platform. Budget available, scale exploding.
## 3. Ema AI — Confidence 5/5
- Headcount: ~228 employees | Funding: Series A, $61M total (Accel + Section 32 led; KPMG strategic minority)
- Agent use case: Universal AI employees for enterprise — pre-built AI agents for HR, IT helpdesk, CS, sales ops. On-prem deployment. KPMG partnership for Fortune 500 distribution.
- Decision maker: CTO/Co-founder Souvik Sen (ex-Okta VP Eng, ex-Google ML) · LinkedIn: souvik-sen; CEO: Surojit Chatterjee (ex-Coinbase CPO) · LinkedIn: surojitchatterjee
- Why now: Expanding into on-prem enterprise, KPMG distribution = scaling fast. Multi-agent enterprise workflows = LLM cost optimization critical.
## 4. Voiceflow — Confidence 4/5
- Headcount: ~88 employees | Funding: Series A, $39.8M total (OpenView Venture Partners — August 2023)
- Agent use case: Enterprise AI agent builder platform — teams design, test, deploy AI agents for customer support. Model-agnostic: OpenAI, Anthropic Claude, Google. Developer community 100K+. Restructured pricing around AI credits April 2025.
- Decision maker: CEO/Co-founder Braden Ream · LinkedIn: braden-ream; CTO/Co-founder Tyler Han
- Why now: Credits-based pricing model means LLM cost is their core business variable. Multi-model agent pipelines at enterprise scale.
## 5. Hyperbound — Confidence 4/5
- Headcount: ~51 employees | Funding: Series A, $18M total (Peak XV led — September 2025; YC S23)
- Agent use case: AI sales roleplay agents — platform builds AI buyer simulation agents mimicking real ICP personas. 7,000+ customers across SaaS, financial services, logistics.
- Decision maker: CEO/Co-founder Sriharsha Guduguntla · LinkedIn: sguduguntla; CTO/Co-founder Atul Raghunathan (LLM researcher, ex-enterprise ML)
- Why now: Recently closed Series A, CTO is hands-on LLM researcher = high receptivity to optimization tools. Small team, CTO approachable.
## 6. Lindy AI — Confidence 4/5
- Headcount: ~52 employees | Funding: Series B, ~$54M total
- Agent use case: Personal AI workflow agents — AI agents handling email triage, scheduling, meeting notes, task delegation. Always-on autonomous AI "chief of staff."
- Decision maker: CEO/Founder Flo Crivello (ex-Uber PM, ex-Cruise/YC) · LinkedIn: florentcrivello
- Why now: Series B signals strong PMF. LLM inference is primary COGS. Founder is active on LinkedIn/podcasts — reachable via content engagement.
---
## Priority Outreach Order
1. Dust.tt (Stanislas Polu) — Sequoia-backed, Anthropic already a vendor, 300K+ agents deployed
2. Artisan AI (Ming Li) — highest LLM volume, Series A just closed, budget available
3. Hyperbound (Atul Raghunathan) — LLM researcher CTO, small team, high technical champion potential
4. Voiceflow (Tyler Han) — credits-based business = direct LLM cost pressure
5. Ema AI (Souvik Sen) — larger, more complex sale but KPMG partnership = scale
6. Lindy AI (Flo Crivello) — CEO as DM, reachable via content
## Next Scan
Refresh July 12 2026. Watch: Ema AI Series B signals (KPMG deal suggests imminent upgrade round); Artisan AI LLM engineer job postings; Voiceflow new enterprise customer announcements.
Agents: GET /api/brain to read everything · POST /api/agent to write. Header x-api-key. Actions: entry, task, update_task, question, answer_question, challenge, update_challenge, person, mission, metric.