Your Employees Aren’t Resisting AI. They’re Telling You Something.

The COO called me after their AI rollout had stalled. Three months in, adoption was low, workarounds were everywhere, and the team that was supposed to be using the new system had quietly gone back to their spreadsheets. “They’re just resistant to change,” she told me. “We need to push harder.”

I asked her one question: did anyone talk to the employees before you chose the tool?

There was a pause. “We did demos.”

Demos aren’t conversations. Demos are presentations. And the difference between those two things is usually the difference between an AI implementation that works and one that costs you six figures and six months and ends up as a cautionary tale in your next all-hands.

When employees resist an AI tool, they’re not being difficult. They’re telling you something specific, and if you listen to it instead of pushing through it, you’ll usually find one of three things.

First: the tool doesn’t actually fit how they work. It was designed for an idealized version of their job — the way someone in a conference room imagined the job — not the real version with all its workarounds, exceptions, and tribal knowledge. Employees who’ve been doing the job for three years know things about the process that never made it into the requirements doc. When the tool ignores all of that, they ignore the tool.

Second: they don’t understand what happens to them if they use it well. If an AI tool makes someone twice as productive, the reasonable question is: does that mean they keep their job, or does it mean the company needs half as many of them? Most organizations roll out efficiency tools without ever answering that question out loud. The silence reads as confirmation of the worst-case scenario, and adoption dies before it starts.

Third: the experience of using the tool is worse than the experience of not using it. This sounds obvious but it’s the thing most technology evaluations miss entirely. You can run a hundred demos and still not know what it feels like to use a tool under pressure, on a deadline, with a frustrated customer on the other end of the interaction. Employees know. They’ve tried it. And if it’s clunkier than what it replaced, they’ll find a way around it every time.

The fix for all three of these is the same: involve employees before you decide, not after. Map what they actually do, not what the process document says they do. Ask them what’s broken in the current workflow. Ask them what would make their job better. Then evaluate tools against that reality instead of against a vendor’s use case.

When employees feel like the tool was built for them rather than imposed on them, adoption takes care of itself. When they feel like it was built without them, you’ll be having the same conversation in six months that COO was having with me — wondering why nobody’s using the thing you spent so much money on.

AI doesn’t replace employees who feel genuinely useful. It threatens employees who’ve been given no reason to believe it won’t. That’s not a technology problem. It’s a leadership problem, and it’s solvable before the rollout, not after.