Fifty-five per cent of British workers say they use AI. Thirty-five per cent of British firms say they use any at all. Eleven per cent have trained more than half their staff.

On 20 July the Office for National Statistics published its fullest account yet of how artificial intelligence has landed inside British firms. It is not the report I expected.

Two numbers sit near each other in it. Fifty-five per cent of employees say they use AI for work or education. Thirty-five per cent of businesses with ten or more employees say they use at least one AI technology.

The two do not measure the same thing, and the ONS says so. One survey asks individuals in Great Britain and counts study as well as work. The other asks firms across the UK. James sets out below why the size of that gap is arguable. What is not arguable, and this is the statisticians' reading rather than mine, is that a lot of people are using these tools without their employer recording it.

Then there is the training figure, which is the one that stopped me. Eleven per cent of businesses report that more than half their workforce has had any AI-related training. One company in nine.

Put those together and you do not get the picture in the speeches. Nothing has changed office work this much since the spreadsheet. And it is not being rolled out. It is seeping in. People are working it out at their own desks, often on free software, and the organisation employing them is a step behind and sometimes does not know.

I have been asking since June what people will do when the machines take the work. The honest answer this morning is that in most British firms the machines have not arrived in any organised way at all. They have arrived in people's hands, and nobody has decided what that means.

That is not reassurance. A change nobody is managing is still a change, and this one is happening without anyone deciding how the gains are shared, who gets taught, or what the person who saves two hours on a Tuesday does with them. Left alone, that question answers itself, and usually in the employer's favour.

James has the depth numbers below and an awkward question with them: every summary of this release said it proves AI is not costing British jobs, and in July we used an earlier wave to argue close to the opposite. Ines has what follows when the commonest way a firm gets AI is that somebody downloads something free. Elena asks who is supposed to be in charge.

The Playbook.

If you use AI at work and nobody has ever asked you about it, you are in the majority rather than cutting a corner. That is worth knowing before your next appraisal.

The date to watch is the next wave of this survey. The questions on how extensively AI is used were only introduced in June, and one wave is a reading, not a trend.

— M.

*This note is mine: the view, and the call to run it. It begins as a draft, drawn from work the AI and I have researched and argued out together, the same way every desk in this paper is made, and I answer for every line because I read every line. Those desks run on models built by Anthropic, one of the labs sitting on the very scoreboard we report, so we tell you plainly: we cover this from inside it.*