Thawing the frozen middle: overcoming resistance to AI adoption

The frozen middle is the layer of middle managers who quietly stall AI adoption. Not by refusing it out loud, but by deprioritising it: the pilot that never gets staffed, the training slot that keeps slipping, the tool nobody in the team is told to use. Senior leaders set the direction and the front line is often keen to try things. In between sits a group who can slow everything to a stop without ever saying no.

It matters because this is usually where AI programmes die. The strategy is signed off, the budget exists, and then nothing moves, because the people who run the day-to-day work were never given a reason or a way to change how their teams operate.

Why middle managers resist

Most of it is rational. Understand the reasons and you can deal with them.

Fear is the obvious one. If AI does part of your team’s job, what happens to your team, and to you? A manager who suspects the answer is “fewer people” is not going to champion the thing that gets them there.

Then there is workload. Middle managers are already carrying delivery targets. Adoption asks them to run their normal job and learn a new way of working and bring their people along, usually with no let-up on the numbers they are measured against.

The bigger two are structural. Many have no clear mandate: they have heard the company is “doing AI” but not what that means for their function this quarter, so they wait. And the incentives point the wrong way. When a manager is paid on this quarter’s output, an investment that pays back next year is someone else’s problem.

The gap this opens up

You can see the cost of the frozen middle in the data. In aibl’s State of UK AI Adoption Survey 2026 (755 UK mid-market leaders), only 24 per cent of companies report full management alignment on AI priorities, and those companies report 79 per cent measurable ROI. Where senior colleagues mostly disagree on the priorities, that falls to around 30 per cent. Full alignment is the single biggest step up in the whole survey.

There is a second signal, and it is easy to misread. In the survey, 55 per cent of leaders say unapproved or personal AI use, the “shadow AI” of people quietly using their own tools, is common or very common. It is tempting to read that as defiance and clamp down. But 53 per cent say people reach for unapproved tools mainly to move faster than the approval process allows. That is a velocity signal, not a discipline problem. Your people want to work quicker than your process lets them. Freeze the middle and that pressure does not disappear, it just goes underground.

How to bring the middle with you

Involve them early. The managers who will run the change should help design it, not receive it as a finished plan. People defend what they helped build.

Give a clear mandate. Vague enthusiasm from the top is not direction. Tell each manager what their function is expected to do with AI this quarter and how it will be judged, so waiting is no longer the safe option.

Reskill them, and protect the time to do it. A quarter of leaders in the survey name skills gaps as their single biggest source of AI friction. You cannot ask someone to lead a change they do not understand. Book the training into the calendar and take something else off their plate to make room, rather than adding it on top.

Fix the incentives. If managers are measured only on short-term output, reward the adoption work explicitly, or it will always lose to this week’s targets.

Show quick wins. A pilot in one team that visibly saves hours does more than any all-hands. Let managers see a peer’s result, not a vendor’s slide.

The job senior leaders can’t delegate

The middle cannot thaw itself. Senior leaders have to name the change, say plainly what it means for jobs, and keep saying it. Honest communication beats reassurance here. If roles will change, say so and say how you will handle it, because managers can tell the difference between a real answer and a soothing one, and the vague version is what breeds the quiet resistance in the first place.

Two things are worth building in early. Alignment at the top comes first: if the executive team does not agree on the AI priorities, the middle will read the disagreement and stall, and the survey shows how much return rides on that agreement. And treat change management as a critical-path dependency, not an afterthought. It starts before the pilots, not after them. Tell people what is coming and why from the moment the plan exists, so the first pilot lands on prepared ground.

Resistance or legitimate caution?

Not all pushback is the frozen middle, and treating caution as obstruction will cost you your best managers. The difference is specifics. Legitimate caution comes with a named concern: this tool has a data problem, this workflow has a compliance step, this pilot has no success measure. It tends to offer a condition, not a refusal, and the person will move once the concern is met.

Frozen resistance is vaguer and more permanent. It agrees in the meeting and does nothing after. It cites no fixable objection because the real objection is unspoken. The test is simple: ask what would need to be true for them to back it. A cautious manager gives you a list. A frozen one cannot, because the answer is “nothing”, and that is the conversation to have next.

Frequently asked questions

How do we spot the frozen middle?

Look for the gap between agreement and action. The managers who nod in the meeting and whose teams show no change a month later are the tell. Track adoption by team, not company-wide: a healthy average can hide two or three functions that have not moved at all, and those are your frozen pockets.

What if the blockers are senior leaders, not middle managers?

Then fix that first, because it is worse. The survey shows return depends heavily on whether the executive team agrees on the AI priorities, and only 24 per cent do. If two directors quietly disagree, every manager below them sees it and waits. Get genuine alignment at the top, in writing, before you ask the middle to move.

How long does it take to shift the culture?

Longer than a pilot and shorter than people fear. Plan on several quarters, not weeks. Visible quick wins can shift attitudes in a team within a quarter, but changing how a whole layer of management works is a year or more, and pretending otherwise sets you up to look like you have failed when you are actually on track.