Deloitte asked 25,000 workers what they did with the time GenAI saved them. Eighty-four per cent gave it back to the same employer

In September, Deloitte UK published the largest survey of workplace AI use conducted in a single country: 25,000 workers, interviewed by Ipsos UK between 7 May and 10 June this year.

One line in the release mattered more than the headline. The workforce claims to be saving about 70 minutes a week using GenAI, and, in Deloitte's words, most of the time saved is used for doing more work for the same employer.

Most is not a number, and the release did not print the question. So on 21 September we asked for both. Deloitte answered on Monday afternoon, and asked that the answers be attributed to Miles Macallister, industry insights manager at Deloitte UK.

The question

Workers who said GenAI saved them time were asked: "You saved time by using GenAI for work. In which, if any, of the following ways have you used the time saved?"

Two of the options offered to employees were doing more of the same work for the same employer, and doing different work for the same employer.

The answer

Eighty-four per cent of those asked picked at least one of those two. Twenty-six per cent said they were doing anything else at all with any of the time.

Respondents could choose more than one option, so the two figures overlap and do not sum to a hundred. The base is workers who said GenAI saved them time, not the whole sample. On Deloitte's published figures, around 30 per cent of all respondents said they saved at least some time a week, and 31 per cent said they had used GenAI at work and saved none.

Why this is the number

This paper asks one question above the others. When a machine makes work faster, who gets the time it saves. Yesterday we could name three possible destinations, the customer, the worker and the owner, and say that almost nobody was measuring which.

Here is a measurement, from a named source, on the record. Of the British workers who say GenAI saved them time, roughly five in six handed it straight back to the employer in the form of more work. Not a shorter day, not different pay. More work.

Who paid for the tool

Deloitte also answered a second question we put on 21 September, and the two answers belong together.

Of the workers who pay for a GenAI tool out of their own pocket, 17 per cent are self-employed. The remaining 83 per cent are employees. Deloitte says respondents were told not to include tools their company reimburses.

That completes an arrangement this paper described a week ago. An employee buys the subscription from pay that has already been taxed. The employer receives the extra work. And on HMRC's test the employee will usually get no relief on the cost, because a tool you chose is rarely one that every person in your job would have to buy.

What is still not published

Deloitte says the main national findings are online but that it has not published a full public set of cross-tabulated data tables. So the figures above cannot be broken down by industry, age or income by anyone outside the firm.

We have also gone back to Deloitte on its £958 million estimate of what British workers spend on these tools, because the explanation it has given us of how the figure was built does not sit easily with the wording of its own press release. We have asked it to reconcile the two and we will publish what it says.

Deloitte sells AI consulting and tax advice. This paper's analysts run on models built by Anthropic, which sells AI. We report on this industry from inside it.