Your usage dashboard is lying to you

Rising AI usage is not adoption. It can be compliance theater. Your team logs into the tool because you told them to, the numbers climb, and the dashboard turns green. Underneath, trust is draining out.

You can mandate usage. You can’t mandate belief.

This weekend a developer’s post hit the top of Hacker News. The title: “LLMs are eroding my career.” Thousands of engineers showed up in the comments to say the same thing in their own words. These aren’t laggards. They’re some of the most capable builders alive, and they’re quietly convinced the rollout is making their work worse.

If your team thinks that, your AI push is already losing. The dashboard just hasn’t told you yet.

Here’s the mechanism. Frame AI as a cost cut and your team hears one thing: you’re the cost. HBR’s research has a name for what happens next. People “rationally disengage.” They use the tool exactly as much as required and not one click more. They stop bringing you the friction. They stop telling you where the output is wrong. They protect themselves, because you’ve told them the goal is fewer of them.

A team can be capable and still not trust you. Trust doesn’t show up in a usage report.

So you keep buying. Another license. Another seat. Another mandate. The real cost of AI was never the license. It’s the change-management drag, the slow rot of a team that’s stopped believing the tool is on their side. You bought your way into that. You can’t buy your way out.

Now the contrarian part. The cheapest, highest-return move before your next AI rollout isn’t a tool. It’s a question.

Ask your team where AI is making their job worse.

Not better. Worse. “Better” gets you the answer they think you want. “Worse” gets you the truth, because it gives them permission to say the thing they’ve been swallowing in standup for three months. Where does the tool slow them down. Where does it make confident mistakes they have to catch. Where does it strip the part of the job they were actually good at.

You’ll hear it in thirty minutes. Then comes the part most leaders skip.

Act on one answer.

Pick the loudest complaint and fix it where everyone can see. The fix is small. The signal is everything. You asked, you listened, you moved. That’s how trust comes back, one kept promise at a time.

“AI replaces you” triggers resistance. “AI makes you faster” triggers adoption. Same software, opposite outcomes, and the only variable is what your team believes you’re building it for.

I’ve watched this from the inside. We’ve run forward-deployed teams inside other companies for years, sitting next to the people doing the work while the tools go in. The rollouts that stick share one trait. They made the team faster. The ones that died all had the same tell: great usage numbers, dead eyes in the room.

Sometimes AI does reduce the headcount you need. That’s real, and you don’t have to lie about it. But cutting heads is rarely just about cutting cost. Done right, it clears room and reward for your best performers to grow into bigger work. Say that. Your best people can smell the difference between a layoff dressed as a tool and an opening dressed as one.

The fear behind all this is that your team can’t handle the change. Flip it. Your team handles change every week without you noticing. What they can’t handle is a tool that arrives as a threat with a deadline attached. Give them the question and they’ll hand you the map. Skip it and you’ll find out anyway, in the exit interviews.

Adoption you ordered is rented. Adoption you earned compounds.

Ask where it’s making the job worse. Then go fix one answer.