Most enterprises measure AI success the way they'd measure a software rollout, by adoption. When we surveyed our organization, the teams with the highest adoption weren't the ones capturing the most value, and none hit the 10x everyone talks about. I have led many transformations — private equity, commercial turnarounds, and major software rollouts. AI is the first where classic playbook doesn't work: you can't build a roadmap backward from a target state that keeps moving. An honest account from inside an enterprise — including what didn't work: what to measure instead of adoption, how to plan without a fixed target state, and how to build a system that turns experiments into real value.
