More UK CFOs expect AI to cut headcount than to lift margins, Deloitte's poll of 66 shows

Both questions cover the next twelve months, so Deloitte's poll measures a gap in expectations and cannot say which event arrives first ^[2][4]. The gap is 52% against a third, about 19 points ^[17]: roughly 34 of the 66 CFOs expecting cuts ^[15] and 22 expecting a margin lift ^[16]. The difference is about 12 respondents ^[20].

The two accounts of the margin question also differ. Bloomberg says a third of CFOs believe AI will help increase margins ^[4]; Crypto Briefing says a third believe it will significantly lift them ^[5]. If Crypto Briefing has the wording right, the margin question sets a higher bar than the headcount question, and some of the 19 points would be wording.

The base is thin. Treated as a simple random sample (an assumption, since none of the three reports describes how respondents were chosen), 52% of 66 carries a margin of about 12 points either way, a range of roughly 40% to 64% ^[24]. Of the 66, 12 were from FTSE 100 companies and 25 from FTSE 250 companies ^[7], which puts 37 in the two indices and 29 outside them ^[18].

We think the poll fits a cost-first reading of AI, and it fits a second reading as well. Cost reduction rose 15 points as a pressing priority ^[21], and Bloomberg reports that AI adoption and high employment costs are encouraging firms to expand with fewer roles ^[9]. If the wage bill is the driver, a CFO can expect a smaller payroll without expecting AI to lift margins, because the saving has to cover what the technology costs first.

The cost side is visible in a separate report. Multiverse, a London training platform founded by Euan Blair, published findings this week that firms are spending more on AI tools without training their workers to use them ^[13]. "Almost none of them know what their people can do with AI, and that is where their spend is going to waste," Blair said ^[14]. Multiverse sells training, so the finding comes from a party with something to sell. The direction is still plain: payroll falling while tool spend rises nets to a smaller gain than either line suggests. With about 75% of the long-run change still ahead on the CFOs' own account ^[19], most of that spend is yet to come.

Two results would show the thesis wrong. A repeat poll in which the margin share rises toward 52% while the headcount share holds would mean the gap was wording or timing. And Crypto Briefing cautions that expectations in a survey do not always translate into action ^[23], so cuts that never arrive would break it too.