Nearly two-thirds of organisations have not yet begun scaling AI across the enterprise. McKinsey's research is clear on why: the biggest barrier is not employees - who are ready - but the leaders.

This isn't a technology story. It's a human one.

After 25 years working with teams through significant change, I've seen the same pattern play out across mergers, restructures, digital transformations, and strategic pivots. AI is not different in kind, but it is different in intensity. The speed of change is faster, the stakes are higher, and the human dynamics it surfaces are harder to ignore.

Which is why so many AI transformations are stalling. Not because the technology isn't working. Because the leadership team isn't.

A Pattern Worth Naming

Here's what it typically looks like.

The CEO is fully committed. The opportunity is real. The energy at the very top is genuine. This isn't for show - they believe in it, they've seen the case, they want to move.

But somewhere in the leadership team, the picture is more complicated.

There isn't any obvious opposition. Nothing dramatic enough to challenge openly. What you see instead is a steady stream of entirely reasonable concerns:

Each concern, taken on its own, makes sense. Of course you need governance. Of course alignment matters. Nobody is wrong to raise these things.

But together, they create drag. The kind of silent, subtle drag which makes it almost impossible to address directly. You can't have a conversation about something that hasn't been named. You can't challenge something that hasn't been stated.

The result is a gap between the commitment expressed in the room and what actually happens when the work begins. Months pass. Pilots stay pilots. The organisation keeps investing in AI without scaling it.

What Lies Beneath

When I look at what's actually driving this dynamic, it's rarely cynicism or bad faith. It's something more human than that.

For some leaders, it's uncertainty about their own position. Real questions about their security, their relevance, and how they will find their place in what comes next. AI is forcing a reckoning that previous technology changes didn't, because previous technology changes didn't make seasoned leaders personally question whether their expertise still mattered. This one does.

For others - particularly those carrying the burden of leadership - AI may not be arriving as a "once-in-a-generation opportunity to reimagine everything". It may be arriving as yet another complex, unpredictable thing landing on top of everything else they are already dealing with. Accountabilities don't shrink when a new strategic priority arrives. They accumulate.

Neither of these is an unreasonable response to the situation. Both of them are worth taking seriously before they quietly shape the outcome of an expensive programme.

The Questions That Surface It

In my experience, the most useful thing a CEO or whoever is leading this can do before launching an AI transformation is ask a small number of questions that almost never get asked.

Not questions about the technology. Not questions about the business case. Questions about the human landscape they're actually working with.

"Who on your leadership team has the most to lose if this goes well?"

This question reframes the conversation entirely. Every sponsor has thought about who might slow things down. Very few have thought about it in these terms - from the perspective of the person's interests, not their behaviour. The answer is usually more revealing than anything else you'll hear.

"Who would be the first person to raise concerns - and what would those concerns probably be about?"

Indirect questions often produce better answers than direct ones. This one creates just enough distance for a considered response and almost everyone can answer it. What they say is usually the most useful intelligence you'll collect before the work starts.

"Three years from now, what does failure look like - and is that picture shared by your whole leadership team?"

Most CEOs have a vivid, detailed picture of success. They've thought about it, refined it, used it to build the business case. But the failure picture is often underdeveloped and almost never genuinely shared. That asymmetry is where programmes go wrong. When the team doesn't have a shared understanding of what they're trying to avoid, they can't make coordinated decisions when things get hard.

The pauses before the answers to these questions are usually the most useful data you'll get.

What Good Looks Like

The organisations scaling AI successfully are not the ones that have avoided the dynamics described above. They're the ones that have faced them directly and early, before the work starts, while there's still something that can be done about it.

Practically, this means a few things.

Real listening before the work starts. Not a "communication cascade". An actual conversation with the full leadership team about what they're navigating - the concerns, the uncertainties, the questions they haven't said out loud yet. This is not optional. It is the single most important thing you can do before a programme opens.

Meet people where they are. The person who can't move at the pace being asked of them is not an obstacle to be managed. They are a design brief. What would need to be true for that person to feel genuinely invested in it - not just consulted?

Agree a plan for when progress stalls - before it does. Who calls it and what happens next. Without this, the whole thing drifts into polite fiction. Having the agreement in place signals confidence that you're prepared for difficulty, not just hoping to avoid it.