Decision Brief: Fund the Curator pilot (3–5 coaches)?
Recommendation (the ask)
Approve a small, time-boxed pilot with 3–5 coaches whose sole purpose is to test two kill-or-continue hypotheses — reproducibility of analysis and real, paid-for operational pain — before any v1 build. Do not fund a full product yet.
Why (Minto: answer first, then support)
Curator's entire thesis rests on one unproven claim: that AI session analysis can be made reproducible and trustworthy enough for a coach to send a client. If that's true, the cumulative client record becomes a sticky system of record; if it's false, Curator is "another transcriber." A small pilot is the cheapest way to learn which world we're in.
Three supporting points
- The moat is unvalidated and testable. Reproducibility (F1 ≥ 0.70, stable across re-runs) can be measured on real sessions in weeks — see
10-metric-design-experimentation.md. This is the highest-leverage thing to learn. - The pain is asserted, not measured. Value rests on a single founder conversation + the ICF study. A diary study + interviews convert T5/T6 assertions into behavioral evidence cheaply — see
02-discovery-research.md. - The business model has an open keystone. Who pays — coach, client, or sponsor — is unresolved (
18-pricing-packaging.md). The pilot is also the cheapest WTP probe.
What we're NOT deciding now
Pricing, full build scope, channel spend, and v1 timeline — all deferred until the pilot returns signal. Committing to them now would be spending ahead of evidence.
Cost / risk of the ask
- Cost: founder time + a thin pilot build (the trustworthy session loop,
12/13). - Risk of approving: sunk pilot effort if hypotheses fail — but that is the cheap failure we want.
- Risk of not approving: continue building on unvalidated assumptions; discover the reproducibility wall after a full build.
Decision criteria (go/no-go after pilot)
- GO to v1: reproducibility ≥ target, coaches rate usefulness ≥ 8/10, pain confirmed, a payer emerges.
- PIVOT/STOP: reproducibility unattainable after 3 iterations, or no WTP signal.
Ask Summary
Approve the pilot scope and the two experiments; revisit all scaling decisions at the post-pilot gate.
Audience calibration: framed as a single reversible decision with explicit kill criteria — investor/founder-grade, not a feature pitch. Jargon compressed; the one number that matters (reproducibility gate) is foregrounded.