The Survey Is a Compass, Not a Report Card
Vohra's core move is asking users how disappointed they would be if the product disappeared. Forty percent saying "very disappointed" marks the threshold. That single question cuts through politeness and surfaces genuine dependency. For an AI-native founder, the same logic applies with higher stakes: if users would not miss your agent, they do not trust it. Measuring that dependency early tells you whether humans are actually handing off judgment or just tolerating a demo.
Segment for the People Who Already Love You
The method insists you stop averaging. Pull out the "very disappointed" cohort and study them — who they are, what job they hired the product for, what words they use. This is the unscalable work Vohra recommends and the same work that distribution thinking demands: hand-onboard the first fifty, listen with full attention, and let their language teach you why you matter. In an agent-driven product, this segment also tells you which workflows earned trust and which ones users quietly routed around.
Split the Roadmap — Double Down and Remove Friction
Once you know who loves you and why, Vohra prescribes two parallel moves: deepen the features that create love, and convert the "somewhat disappointed" group by removing their specific objections. The objections are almost always about confidence — users who want your product but do not yet trust it enough to rely on it. For agentic software, removing those objections means designing accountability and override into the product itself, not treating trust as a marketing problem.
PMF Is Dynamic When Agents Keep Changing
A static survey taken once is a photograph of a moment. AI-native products change faster than traditional software because the underlying models shift, capabilities expand, and the scope of what agents handle grows. Vohra's framework is most valuable not as a one-time diagnostic but as a recurring signal layer — a feedback loop that tells you whether each expansion of agent autonomy is widening the "very disappointed" cohort or shrinking it. That loop is how you know whether your judgment calls about what to automate are landing.
