2026-06-23
Becoming an AI-Era PM 04 | Judging "Should We Build It" Now Costs More Than "Can We Build It"
This is the fourth piece in the series Becoming an AI-Era PM. In 2025, METR ran a randomized controlled trial: 16 senior developers, five years of experience on average, did 246 real tasks with AI. Beforehand they expected to be 24% faster; afterward they still felt 20% faster; measured, they were 19% slower. Even the simplest judgment — "did AI make me faster" — got called backwards by the people who knew the work best. When building gets fast and cheap, "can we build it" stops filtering any idea out, and the expensive judgment moves to "should we build it." This piece gives you four things you can actually do: stop using difficulty as a gate, ask what happens if you don't build it, write down what becomes true before you start, and let AI lay out options but never trust "feels right."