Companies have spent billions on AI. Most have nothing to show for it. Gallup just found the missing piece, and it isn't the technology.
There's a number that should stop a board meeting cold.
Around $40 billion has gone into enterprise AI. An MIT study found 95% of organisations have seen zero measurable impact on profit. Not modest. Not slower than hoped. Zero.
So, naturally, everyone looked at the technology. Wrong model. Bad integration. We bought the wrong licences, too early, and someone in procurement is in trouble.
But the tools mostly work. That's the awkward part. The capability is real, it's sitting there largely unused, like a gym membership bought in January, and Gallup's State of the Global Workplace 2026 has the clearest answer I've seen as to why. In companies rolling out AI, employees who say their manager actively backs it are 8.7 times more likely to say AI has changed how work gets done.
Read that again. Not 8.7 percent. 8.7 times.
Gallup's CEO puts it more bluntly than I'd dare. "Even the most sophisticated neural network cannot overcome an indifferent team leader."
The bit everyone skipped
We've built brilliant machines and quietly forgotten the person standing between the machine and the team.
Only one in five employees say their manager actively supports their team's use of AI. So four out of five managers are somewhere between lukewarm and absent on the single thing that most decides whether any of this pays off.
Is that managers being difficult? Not even close. They're knackered and the data backs this up. We handed the most important job in the whole transformation to the most worn-out layer in businessess, then sent them a training module and wished them luck.
You can see how it happened. AI is technology, so it went to the technologists. The rollout looks like an IT migration. Licences. Access. A module. A launch email. All sensible, all necessary, and none of it goes anywhere near the question that actually matters, which is whether a team leader in Manchester or Mumbai believes this thing will make their team's work better. Or quietly suspects it's there to make their team smaller.
And no, another module won't fix it. A belief gap and a knowledge gap are different things, and we keep treating the first like the second.
You can't sell what you don't believe
Nobody champions a thing they don't believe in. And belief, sadly for the rollout plan, doesn't turn up in an inbox.
Experience tells us this. You can hand someone every fact about a new tool until you're blue in the face, and still nothing shifts. Why? Because facts don't move people. Feelings do. People act when they feel something, not when they've been informed of something.
I keep thinking about a job we did with Deloitte in the US. The brief was coaching skills. Getting managers to stop telling people what to do and start asking, listening, drawing them out. It's the hardest behaviour change there is, because every manager already reckons they're great at it.
The obvious move would've been a course on active listening. But lecturing people about listening is its own punchline, so we made three short films instead. The manager talks straight to camera, mid-conversation, and you're dropped right inside the moment with them. The awkward pause. The urge to jump in. The better question they didn't ask. Nobody tells you what good coaching feels like. You feel it, and then you catch yourself doing it the next morning in your own one-to-one.
That's the whole trick. Don't explain the behaviour. Build the moment where someone feels it, and decides for themselves.
Your managers don't need more information about AI. They've got plenty. What's missing is the thing that turns a tool into a conviction. A reason to walk into Monday's stand-up and mean it when they say here's how this makes our work better, and here's why it isn't coming for you.
That was never a technology problem. It's a story problem wearing a technology hat.
What it's costing?
There's a bill for all this, so let's name it.
The billions are gone, and for 95% of companies the return simply hasn't shown up. That's the obvious cost. The quieter one is the credibility you burn through. You announce a transformation to the board, to the market, to your own people, and eighteen months later the honest answer to "did it land?" is a shrug.
That's the cost that compounds because you only get to cry wolf on transformation so many times before people stop believing.
So here's my take. The fix isn't another platform. The capability is already bought and paid for, sitting right there. The fix is the layer between the technology and the team, and you reach that layer with something they feel, not something they're told.
Spend the next budget line there. Not on more model. On more belief.
And belief in the right thing. Not a belief that AI will do the thinking for your people, but that it's worth their expertise to make it useful. Get that distinction wrong and you haven't fixed adoption. You've just automated the mediocrity.
Fin.
