Unable to decide on the ICP
There are multiple ways we can go, but unable to decide on the icp.
Solutions: ICP is two tiers, grounded in research and locked decisions: TIER 1 — PLG Entry (Arena-first): 20–150 employee companies with a technical founder or Head of Eng/Ops actively building and shipping AI agents — in n8n, LangChain, or custom glue — with $500–$5k/mo in LLM spend. These teams feel cost pain at the margin and will self-serve through Arena's cost-revelation flow. The highest-signal cohort: teams that wired brittle LangChain harnesses they are now abandoning (Research Brief #4 — greenfield/brownfield is the wrong binary; stalled at the 1→5 agent scale wall is the right filter). Arena is the natural first touch — no sales required. Converts at $99/mo (Basic, up to 5 agents). TIER 2 — Commercial Expansion: 100–500 employee Series B/C SaaS companies with a VP Eng or Head of AI actively running agents in production, with $5k–$50k/mo in LLM spend. Decision maker is the technical leader, not procurement. Willing to pay $499/mo for a proper agent operating layer once the Arena aha moment proves ROI. Dust.tt (CTO Stanislas Polu, ex-OpenAI, Anthropic in stack) is the archetype — outreach already queued as Task #23. Ema AI, Artisan AI, Voiceflow, Hyperbound are Tier 2 candidates from the July 5 prospect scan (Brief #3). NOT the ICP right now: enterprises (17-person decision stacks, 6–12 month cycles — Decision #29); pre-seed teams (no meaningful spend to optimize); teams still "exploring AI" (no active agent runs). Rationale grounded in brain: - Decision #29 locked mid-market (50–500 emp) as baseline — this solution refines the lower bound to 20 emp to capture the n8n/solo-technical-founder segment that Arena can convert without sales. - Brief #4 established that the greenfield/brownfield binary is wrong — the hottest buyers are teams with brittle self-built harnesses stalled at the 1→5 agent scale wall. - Entry #23 (validation flag): LLM prices fell ~80% YoY — pure model-cost savings are a shrinking wedge. The ICP must be teams who feel total agent run cost pain, not just model switching cost. - Brief #2 (competitor analysis): Gateway race is lost. Moat is the operating layer — compounding, reliability, control. Tier 1 enters on cost, Tier 2 pays for the operating layer. - Pricing (Decision #46): $99/$499 BYOK tiers fit this two-tier ICP cleanly. Tier 1 enters at $99, Tier 2 at $499. No managed keys needed yet. - Motion (Decision #30): Hybrid PLG + light-touch technical sales. Tier 1 is pure PLG. Tier 2 is one 20-min technical call by Vishnu post-Arena aha. Action: Rebuild outreach list against Tier 1 signals (n8n community, job posts for "AI engineer" at 20–150 emp co's, GitHub repos with LangChain/CrewAI abandonment signals). Validate $99 price anchor with first 5 Tier 1 converts before committing to public pricing page.