If your agents can iterate product weekly, the bottleneck isn't velocity anymore — it's knowing precisely who you're building for and what would break their heart to lose.
Rahul Vohra's method, published in First Round Review, turns product-market fit from a feeling into a number. You survey users with a single brutally simple question — how would you feel if you could no longer use this product — and you treat the "very disappointed" percentage as your north star. Forty percent is the threshold that correlates with real traction. More importantly, Vohra's insight is that you don't chase the median user; you isolate your most enthusiastic cohort, study what they love and why, then use the lukewarm crowd's objections not to water down the product but to find the shortest bridge between them and the believers.
Agent-native products compound this problem because the surface area of the product keeps shifting underneath users — agents add capabilities, change behavior, produce different outputs run to run. That makes the "very disappointed" signal even more load-bearing, because you need an anchor that isn't tied to any single feature. Vohra's segmentation logic maps cleanly: your power users are the ones who've learned to direct agents well, and their disappointment tells you which judgment-plus-automation loops are genuinely irreplaceable. Build the roadmap around protecting those loops first, then examine why the persuadables hesitate — usually it's trust, not capability — and address that specifically without dulling what the devotees love.
- Survey early and often, treating the very-disappointed rate as a hard metric, not a vibe check
- let your most enthusiastic users define the product's core rather than averaging them into invisibility
- when agents change behavior, re-run the survey — PMF is a moving target that needs continuous triangulation, not a milestone you pass once.
