$250 Billion Spent on AI. Ninety Percent of Companies Have Nothing to Show for It.
$250 Billion Spent on AI. Ninety Percent of Companies Have Nothing to Show for It.
That's the headline finding from a study published this month by the National Bureau of Economic Research (NBER), one of the most respected economics research organisations in the US. They surveyed nearly 6,000 CFOs, CEOs, and senior executives across the US, UK, Germany, and Australia. More than 90 percent said AI had no impact on employment at their organisation over the past three years. 89 percent saw no change in productivity.
Two-thirds of those firms say they're "using AI." But when you look at what that actually means, the average employee spends about 90 minutes a week with AI tools. That's not a transformation. That's a coffee break. One almond croissant on a Tuesday and back to business as usual.
The Playbook That Doesn't Work
Nobody would buy a gym membership, go once a week for twenty minutes, and then blame the equipment for not getting fitter. Yet that is essentially what most organisations have done with AI. They've bought the membership, sent the welcome email, and mistaken the purchase for the workout.
We've seen this up close working with organisations across industries. The pattern is almost always the same: buy the tools, send the "exciting news" email, maybe run a lunch-and-learn, and then wait for magic to happen. But nobody redesigns the workflows. Nobody rethinks how decisions actually get made. Nobody asks the harder question of what work should look like now that the tools have changed.
It reminds me of something the economist Robert Solow said back in 1987: "You can see the computer age everywhere but in the productivity statistics." At the time, businesses had been investing in computers for over a decade with almost nothing to show for it in the data. It took another 15 years before productivity gains materialised. Not because the technology was bad, but because organisations had to learn how to reorganise around it. New roles emerged. New processes replaced old ones. The resistance to changing how things were done had to be overcome, one team at a time.
We're in the Awkward Middle
AI is in that same phase right now. The tools work. They're improving at breakneck speed. But the systems around them haven't caught up yet. Most organisations are layering AI on top of processes that were designed for a world without it.
The real productivity gains don't come from the tool itself. They come from rethinking the work. Which decisions can be made faster with better information? Which manual steps can be eliminated entirely? Which roles need to shift from execution to oversight? These aren't technical questions. They're organisational ones. And they require leadership that's willing to do more than just approve a software license.
Access Is Not Adoption
Giving people access to AI tools isn't the same as changing how they work. A Copilot seat doesn't redesign a procurement workflow. A ChatGPT subscription doesn't fix a broken approval chain. Confusing availability with adoption is like confusing owning a cookbook with knowing how to cook.
The NBER data reinforces what we keep seeing on the ground: the gap between having AI and benefiting from AI is enormous. And it's not closing on its own.
So What Actually Works?
The organisations seeing results aren't the ones with the biggest AI budgets. They're the ones willing to do the uncomfortable work: mapping how decisions actually flow through their business, identifying where AI can genuinely change outcomes, and then committing to making those changes stick.
That means redesigning workflows, not just adding tools to existing ones. It means measuring impact in real terms, not in pilot demos. And it means treating AI adoption as an organisational change programme, not an IT rollout.
The 90% stat isn't a failure of AI. It's a signal that most organisations haven't started the real work yet. And until they do, the productivity statistics will keep telling the same story.