You picked the tool. You wrote the rollout plan. You sent one email to everyone — and then adoption stalled. Not because the tool was wrong, but because you spoke to ten thousand people as if they were one person. They aren't.
On any question about AI, a workforce doesn't move as a block. It splits — into at least four groups, each of which hears the same announcement completely differently, and each of which needs a different thing from you before it will move. Treat them as one audience and you design for none of them.
The same announcement, four ways it lands
Enthusiasts
“Let me try everything.”
Already ahead of you. Give them guardrails, not brakes — somewhere to run fast without breaking what matters.
Pragmatists
“Show me it works first.”
They'll move — on proof. Not a vendor demo. Evidence from people like them, doing work like theirs.
Sceptics
“Seen the hype before.”
Won't be sold. They'll respect honesty about the limits far more than a pitch about the promise.
The anxious
“What happens to my job?”
The question underneath the silence. They need safety and involvement before they need features.
The common mistakeYou built the whole thing for the enthusiasts.
Your champions are the enthusiasts — and they were always going to adopt. Building your rollout around the people who need no convincing is the most natural mistake in the world, and it leaves the majority precisely where they started. The pragmatists, the sceptics and the anxious are the other ninety-eight per cent, and they are where adoption is actually won or lost.
Your champions are the enthusiasts. What about the other 9,800?
The lever most people skipUse it yourself, where they can see.
The single strongest signal isn't a training budget or a mandate. It's a leader seen using the thing — in their own work, in the open. When the person at the top visibly uses AI, the value their team gets from it rises sharply. Adoption is modelled from the front, not announced from it. A memo asks people to change; a boss who has quietly rebuilt their own Monday around it shows them it's safe to.
So the work isn't a better email. It's four conversations. Give the enthusiasts a sanctioned place to run and the guardrails to run safely. Give the pragmatists proof from their own peers. Give the sceptics an honest account of the limits — they'll trust you more for it. Give the anxious safety and a real role in what comes next, before you give them a tool. And whatever else you do, use it yourself, visibly.
Adoption isn't a tooling problem you can procure your way out of. It's a leadership problem — four groups, four needs, one leader willing to go first.
David Kolb advises leaders on adopting AI across a whole organisation without leaving most of it behind. This note is written to be forwarded — if it's useful to a colleague, send it on.