Middle managers are blocking AI progress. They often have good reasons

There’s a pattern that keeps repeating across organisations. Senior leadership gives speeches about AI and the future. Employees have already started on their own, using tools they’ve chosen themselves. And yet, actual adoption moves painfully slowly. Somewhere between the vision and everyday work, things get stuck.

That somewhere has a name: the middle manager. It’s her decision whether the new tool actually gets used in the department or gets filed under "we’ll look into it." Research on technological shifts is clear on this point: the real power over adoption rarely sits with the leadership team. It sits with the managers in the middle.

The gatekeeper isn’t wrong

Here, the story often turns unkind toward the middle manager: she’s holding things back because AI threatens her relevance, she who built her position on knowing things that a machine now knows too. There’s some truth to that. Researchers have documented how new technology is resisted from within by those who have the most to lose, and the paradox is toxic: the people who best know where AI could make the biggest difference have the strongest reasons to stay quiet about it.

But take a moment to consider the middle manager’s actual objections, because they usually sound like this. What happens to the customer data that gets pasted in, where does it end up? Who’s responsible when the AI gets it wrong and the answer has already gone to the customer? What does this actually cost when the whole department starts using it? And how do I explain to the union and the health and safety rep what the tool does?

This isn’t technophobia. These are the questions a responsible manager should ask, and most AI tools don’t have answers for them. In that case, hitting the brakes is the only responsible move. The gatekeeper is guarding because the gate has no lock.

Give the gate a lock instead of sawing off the gatekeeper

We built EuroWork so that the middle manager’s questions have answers she can point to, while the work is happening, not in some policy no one’s read.

Where does the data end up? Inside the EU: the database and application run in Frankfurt, and it’s openly declared in the service. Who’s responsible? A human: every AI response is labelled as an AI response, and anything an AI agent does is shown as a plan, waiting for approval, and leaves a receipt, so there’s always a name behind what gets sent. What does it cost? It’s displayed next to every answer in kronor and öre, and the manager has a budget with a cap, not a surprise afterwards. What does the employer see of employees’ work? The number of uses and costs, never the content, and even that is clearly stated.

With those answers in place, the middle manager changes roles, from the one who has to hit the brakes to the one who can lead. It’s the same person, with the same sense of responsibility. What’s changed is that the responsibility is now possible to take on.

The middle is smaller in smaller organisations

One more thing, and it’s hopeful for most of you reading this. The bigger the organisation, the more layers of gatekeepers sit between decision and action. A small or medium-sized business might only have one. That’s why we’re already seeing small organisations outpace giants with ten times the resources: they have a shorter path from "we should" to "we are," as long as they give their one gatekeeper good answers instead of good reasons to wait.

Your employees are already using AI, no matter what’s stuck in the middle. Do you know what they’re sharing with it?