Why AI Adoption Stalls on Managers, Not Employees

Almost every organisation now says AI is a strategic priority. Far fewer have worked out whether their managers are actually allowed to act like it.

The number that explains a lot of stalled pilots

Gallup's latest workplace research, published 17 August 2026, found that 99 per cent of HR leaders consider AI important to their organisation's strategy. Only half trust their managers to actually guide a team through using it. That gap, between what leadership says and what frontline managers are trusted to do, goes a long way toward explaining why so many AI adoption efforts stall at the pilot stage, long after the tool itself works fine. It also echoes something McKinsey Global Institute pointed out in late 2025: most human skills don't disappear under AI, they just get applied differently. But only if the system around people is actually redesigned to let that happen, something we've written about when the operating model, not the tooling, is the real constraint.

And the effect here isn't subtle. Where a manager actively champions AI with their team, a third of employees say it transformed how they work. Where a manager doesn't, that number drops to one in twenty five. Same tools, same company, an eightfold swing, and it comes down to one layer of the org chart.

Why the training budget usually goes to the wrong place first

The default playbook for an AI rollout goes: train the end users, measure adoption, then wonder why it's patchy. That approach assumes the bottleneck is employee skill. Gallup's data points somewhere else entirely: upstream, to whether the manager standing between strategy and the team actually has the standing, and the trust, to make the case for the tool day to day.

This tracks with what we keep seeing inside organisations further along in an AI rollout. Licensing the tool and running the demo takes weeks. What takes months, and often just doesn't happen, is a manager deciding it's on them to use the tool themselves in front of their team, defend the time it costs to learn it during an already busy quarter, and absorb the mistakes a team makes while it gets up to speed. Nobody handed them that job. Most companies never built it into what a manager is actually measured on.

The evaluation gap makes the trust gap durable

Only a quarter of companies build AI use into a manager's job expectations or performance goals, and 12 per cent haven't rolled out any formal AI training at all. Put those two together and Gallup's 50 per cent trust gap stops looking like a temporary growing pain. It's structural: managers aren't trained for this part of the job, aren't evaluated on it, and yet the ones who end up championing AI anyway, on their own initiative, get dramatically better results than the ones who don't.

Two researchers quoted in the same reporting get at this from different angles. Mark Muro describes "bring-your-own-AI" adoption, employees quietly picking up tools on their own without any organisational backing, as producing individual "freelancing" rather than real company-wide gains. The upside never compounds, because nobody's actually coordinating it. Anton Dahbura's point lands closer to the management layer itself: AI adoption "needs to be done very thoughtfully and very carefully," with the real customisation work happening at the team level, which happens to be exactly the level nobody's formally supporting right now.

What this means for how capability gets built

For institutions and employers building AI capability programmes, the implication isn't comfortable, but it is clarifying: a training programme aimed only at end users is pulling the smaller lever. Trust, modelling, time-protection, all of it lives at the manager layer, and that's currently the least supported part of most organisations' AI adoption plans.

So the question worth asking before the next AI training budget gets signed off isn't "how do we get more employees using the tool." It's "do our managers actually have the training, the time, and the mandate to make that happen." Because the data says that's the one variable driving the eightfold gap between teams that transform and teams that stall.

‍