When we started rolling out AI tools across our technical org, I assumed the hard part would be technical. Access, permissions, integrations, the usual friction of getting a new tool into people’s hands.
I was wrong. The hard part was psychological, and it took me a few weeks to actually see that.
The Numbers Looked Fine. They Weren’t.
In the first two weeks, adoption numbers were okay, not great. People logged in, poked around, and a lot of them quietly went back to doing things the old way.
Nobody said “I don’t trust this” out loud. But that’s what was happening.
Once I recognized it as a trust problem instead of a training problem, everything about how I approached the rollout changed.
Leading With Permission, Not Features
I stopped leading with features. Nobody was struggling to find the button. I started leading with permission.
What does it actually look like to use this well? What’s fine to hand off, and what still needs a human?
People weren’t slow because they didn’t understand the tool. They were slow because nobody had told them it was okay to actually use it, and what “using it well” meant.
So we built that in. Not a policy document nobody reads, but real examples, real walkthroughs, real “here’s what good looks like” moments, embedded into the actual training instead of bolted onto a slide at the end.
We also tracked something different: not just who logged in, but how deep the engagement went, and where people were re-engaging versus dropping off.
What Actually Moved the Number
That shift got us to 96% active adoption, ahead of the company-wide number.
But honestly, the stat matters less to me than what got us there. It wasn’t a better tool demo. It was treating adoption like the change management problem it actually is, and remembering that behind every “low usage” number is a person who probably has a reasonable, unspoken reason for hesitating.
If you’re rolling out AI to your team right now and the numbers feel stuck, ask yourself this: are people confused about the tool, or are they unsure they’re allowed to trust it?
Those need completely different answers.
Happy Learning!
KP