Your AI Strategy Is Betting on a Burnt-Out Layer

[Originally written in April 2026] Gallup's 2026 report says managers are the key to AI adoption. It also says managers are the most burnt-out layer in the building. Both are true. That's the whole problem.

Everyone quoted the wrong half

The 2026 Gallup State of the Global Workplace report has two findings in it. One got quoted everywhere. The other got buried, and it's the one that matters.

The quoted one: aside from technical integration, the strongest predictor of whether your people adopt AI is whether their manager actively champions it. Not the platform. Not the training budget. The manager. Gallup's CEO said it cleaner than I can.

"Even the most sophisticated neural network can't overcome an indifferent team leader."

That finding is true, and it launched a hundred takes about equipping your managers to lead the AI charge. Fine. But it only works as advice if the managers have anything left to lead with.

They don't. And that's a problem.

The lever is the person you watch

Start with why it's managers in the first place, because the mechanism matters for what comes next. Everett Rogers worked this out in 1962.

Innovations don't spread by broadcast they spread through proximity. You adopt what you watch the people near you adopt, and in a company the person you watch is your direct manager. Not because of the title. Because they're close enough to the work, and sometimes physical proximity, to observe. If your manager quietly uses AI to prep for a meeting, you read it as permission. If they never open the tool, or talk about the tool you read that too.

The size of that effect is not subtle. Employees whose managers actively support AI use are 8.7 times more likely to say it has transformed their work, and 7.4 times more likely to say it lets them do what they do best. That's the whole case for managers as the lever, and it's a good one.

I've watched it convert people. The skeptics didn't move because of a mandate or a training session. They moved the first time a draft that should have taken a day showed up in twenty minutes, and it was theirs. That's the mechanism. Someone you watch, producing something you can see.

The behavior travels. The policy memo doesn't.

It matters more here than in a normal rollout, too, because AI doesn't show up like a new expense tool. It shows up carrying fear. Eighteen percent of U.S. workers now think it's likely their job is gone inside five years. When perceived risk runs that high, people lean harder on the person they watch for a signal that it's safe to engage. The manager's visible behavior isn't just helpful. It's the all-clear. Which means the lever isn't optional. The whole adoption curve runs through it.

The lever is collapsing

Now the buried finding. That same report shows the manager layer in free fall. Manager engagement is down nine points since 2022, sitting at 22%, and dropping at roughly three times the rate of the people they manage. Managers used to carry an engagement premium, more bought-in than their teams almost by default. That premium is gone. They're now only as checked-out as everyone else, and heading lower.

The most powerful lever for the most important behavioral change in a generation is the one layer the organization has spent three years quietly hollowing out.

The usage nobody can see

But wait, it gets worse when you look at who's actually using the tools daily or weekly:

- Leaders use AI most, 67%
- Managers, 52%
- Individual contributors, 46%

But a leader's AI habit is invisible to a front-line employee. They're not in the room. The only usage an IC can actually observe is their manager's, and the manager is the one running on empty.

You've concentrated the behavior that needs to be seen in exactly the place it can't be, run by the people with the least left to model it.

Fewer than one in three employees at AI-adopting companies say their manager supports their use of it. Of course they don't. Support is a thing you do with energy you have.

A capacity problem in an AI costume

So the reality is you don't have a managers-won't-support-AI problem. That's a symptom. You have a manager-capacity problem wearing an AI costume. The diffusion network everyone's diagnosing isn't missing. It's exhausted. And you can't fix an exhausted node by handing it one more thing to evangelize.

Which is what makes most of the current advice backwards. The prescription going around is: train your managers on AI, make them champions, add it to the role. You're piling load onto the layer that's already collapsing under the load as is.

Reverse the order

The order of operations is reversed. You don't get AI adoption by putting AI on a manager's plate. You get it by taking enough off the plate that there's room to model anything at all. Re-engage the layer first. Adoption follows it, not the other way around.

So before the next readiness assessment, a cheaper and more uncomfortable question. It isn't "are your managers using AI." It's "is there anything left in your managers to lead with." If the answer is no, no enablement budget changes the outcome. Rogers told you the innovation spreads through the person people trust. He didn't promise that person would still be standing.