You don’t scale AI with a big launch. You scale it the way Kotter scales any change — with short-term wins that earn the next step. On a floor, the cleanest way to say it is: pull, not push.
Successful test cases manufacture demand.
We took our Production SOPs, used AI to help structure them, and built a step-by-step trainer plus a 6-week onboarding plan — all in one department. Now other departments are asking to join. I didn’t roll it out to them; the win made them want in. That’s pull. Starbucks was push — AI inventory scanning shoved to 11,300 stores with no pilot, dead in nine months. Push forces it where it isn’t wanted; pull lets demand draw it.
It scaled because it was packaged and structured.
It’s shareable in Copilot 365, so it grows two ways — add users, or add content. And the structural detail that made it navigable: descriptive document titles were the single greatest enabler for AI to find the right process and its supporting documents. Structure is what let it scale.
So what, now what.
So what: a big launch forces AI on people who never asked for it; a proven test case makes them ask. Now what: don’t plan the rollout. Land one visible win on a real problem, package it so others can run it, and let the next team pull it toward themselves.
Leave a comment