Something odd showed up in the data this week. Over half of UK businesses are now using AI, yet nine in ten of them say it has made no measurable difference to the size or shape of their workforce.

Either AI is the most oversold tool in a generation, or most businesses are using it for the wrong things.

The numbers

The British Chambers of Commerce published Powering Productivity, its biggest study yet of how UK firms are actually using AI. The headline: adoption has jumped from 23% in 2023 to 54% now. That sounds like progress. But 95% of those businesses report no change in headcount, and 86% say job roles have not changed.

For a business owner, that is the gap worth paying attention to. Plenty of firms are paying for tools. Very few have changed how the work actually gets done, which is where the productivity gain lives.

Stanford also released its 2026 AI Index this week. The headline finding: nearly 90% of executives say AI has had no measurable effect on productivity or employment in their organisation.

What this actually means

Capability has raced ahead. Deployment has not.

In plain terms, the tools are ready. Most businesses are not, because adoption without training or workflow redesign is just licensing by another name.

The pattern is familiar from previous technology shifts. Businesses bought email in the 1990s and spent a decade figuring out how to use it properly. The difference with AI is that the gap between "having the tool" and "changing the work" is where the entire return on investment sits.

A business that has rolled out Microsoft 365 Copilot to 50 people but changed zero workflows has spent money. A business that has trained 10 people to use it on the three tasks where it saves the most time has invested money. The distinction matters.

What to do about it

If your business is in the 54% that has adopted AI tools, ask one question this week: which job has actually changed because of them?

If the answer is "none yet", you are in the 95%. The fix is not more tools. It is training, workflow redesign, and a clear view of which tasks AI should touch first.

That is what separates adoption from progress.