Two numbers from the same fortnight
Deloitte’s third-quarter poll of 66 finance chiefs at large UK companies, published on 14 October, found 52% expecting AI to cut their own payroll within twelve months. Six days earlier the Office for National Statistics (ONS) published its September reading of the same question asked the other way round: of businesses with ten or more employees that already use AI, 6% said it had reduced their headcount, fewer than 1% said it had increased it, and both figures had been broadly stable for a year. Our news report covers the Deloitte release itself. This piece sets its headline figure against what the UK’s adoption surveys say has actually been delivered, and asks what fills the gap between the two.
Strategic Insight: The CFO figure is a forecast and the ONS figure is an outcome, and the two are not in conflict. What is new is the size of the gap between them. In July, Deloitte’s CFOs were optimistic about AI’s effect on performance. By October a majority expect it to show up in their own payroll within a year, while the official measure of realised headcount effects has not moved.
Why expectation and delivery are being measured by different people
The Deloitte survey asks finance chiefs at large UK companies, 37 of its 66 respondents at FTSE 100 and FTSE 250 firms, what they expect. The ONS business survey asks a weighted sample of UK businesses what has happened. Both are legitimate readings, but they answer different questions, and a headline percentage does not say which one it is.
What the quarter-done finding tells us
The most useful line in Deloitte’s report is not the headcount figure. It is the answer to a question the survey asked for the first time this quarter, which the report labels the “CFO’s assessment of how far their business has progressed towards the level of change from AI that they ultimately expect”. The full distribution is stark. Some 18% said no change had yet been achieved, 67% put themselves at a quarter of the expected change, 13% at half, 3% at three-quarters, and nobody at all of it. Deloitte’s own reading is that “the vast majority are still in the early stages of adoption”.
Read together, the two findings say that boards expect the workforce effect to arrive well before the transformation does. Deloitte frames the headcount result as a net balance: taken together, its CFOs foresee a smaller payroll and lower operating costs, with margins and revenue improving, across the coming year. That ordering, payroll first and transformation later, is the pattern this piece is about, and the official data describe a UK business population still at the start of it.
The ONS put UK adoption at 30% of all businesses in September 2026, 38% of those with ten or more employees, and 51% of those employing 100 or more. Its July analysis of the June wave of the same series described adoption to date as “relatively shallow”: the average adopting business used about 1.6 AI technologies, up from 1.4 in late 2023, and only 10% of adopters reported using AI extensively.
| Metric | Value | Source | What it measures |
|---|---|---|---|
| CFOs expecting AI to reduce headcount within 12 months | 52% | Deloitte, Q3 2026, 66 CFOs | Expectation |
| CFOs placing their firm at a quarter of the AI change they expect | 67% | Deloitte, Q3 2026 | Self-assessed progress |
| CFOs placing their firm at three-quarters or more | 3% | Deloitte, Q3 2026 | Self-assessed progress |
| Businesses with 10+ employees using any AI | 38% | ONS, September 2026 | Adoption breadth |
| AI-using businesses (10+ employees) reporting reduced headcount | 6% | ONS, September 2026 | Delivered outcome |
| Adopters reporting extensive AI use | 10% | ONS, June 2026 wave | Adoption depth |
| Housing providers where over half of staff are regular AI users | 14% | Housemark, 66 responses | Depth, one sector |
Strategic Reality: Who you ask changes the answer. The ONS notes that a National Bureau of Economic Research working paper built on polling of senior executives found around 71% of UK firms using any AI technology, against the ONS’s own figure of around 35% at the time. The ONS says estimates vary substantially with the measurement approach, and the surveys gathered here bear that out: Lloyds’ Business Barometer reports 61% of firms using AI, against the ONS’s 30% for all businesses in September.
What the adoption surveys actually show
Access is widening faster than use
Four other UK surveys, each covering a different slice of the economy, describe the same shape. Housemark’s survey, drawing 66 responses across housing associations, councils and arm’s-length management organisations, found 59% saying all or most staff had access to AI tools, while just 14% put regular use above half of their staff, and 11% had a formal strategy for AI. Crowe’s latest Manufacturing Outlook Report, as reported by the Greater Birmingham Chambers of Commerce, found 80% of manufacturers using or trialling AI, up from 52% in spring 2025; the published summary says nothing about depth of use.
The ONS data give the national version of the same picture. Among businesses with ten or more employees, 15% report “that more than half of their employees use AI as part of their daily work”. The ONS added that question only recently and calls the results early indications, but the figure sits close to Housemark’s 14% for one sector.
Critical Context: The ONS also found that employees report more AI use than their employers do. Between May and June 2026, 55% of employees said they used AI for work or study, against 35% of businesses using any AI. Some of the adoption CFOs are counting on may already be happening informally, outside anything the business measures or governs.
Where the gains are going
FRP Advisory’s survey of 250 senior executives from mid-market firms across the UK, reported by Insider Media, found 53% of Scottish respondents saying AI efficiencies had shown up mainly as faster customer delivery, 47% reporting improved margins, and 53% saying the gains had been offset by cost rises elsewhere. The Lloyds Business Barometer points the opposite way from the CFOs on jobs: 54% of UK businesses in its sample said AI had created new jobs, and 21% were creating AI-specific roles.
None of these surveys measures the same thing as Deloitte’s, and none is a rebuttal of it. Together they describe a business population that has bought access, is seeing speed rather than savings, and in the Lloyds sample reports AI creating jobs far more often than the ONS sample reports it removing them.
Why the surveys disagree
- They ask different populations. Deloitte’s 66 respondents include 12 FTSE 100 and 25 FTSE 250 finance chiefs. The ONS weights across all UK businesses, and finds adoption at 51% for firms employing 100 or more against 29% for those with fewer than ten. Large-company expectations and whole-economy outcomes will diverge even if both are accurate.
- They ask about different time frames. Deloitte asks about the next twelve months. The ONS asks adopters what has already happened to headcount, and the Bank of England, through its Decision Maker Panel as cited by the ONS, found that for most firms AI had no material effect on employee numbers over the previous three years.
- They ask different questions. Expectation, access, use, depth and outcome are five separate measures. A survey that reports 80% adoption may be counting trials; a survey that reports 6% headcount reduction is counting only businesses that have already adopted.
The expectation has been building for a quarter
Deloitte’s own second-quarter survey shows an earlier form of the expectation. In July, 73% of CFOs reported improved optimism about AI’s effect on their business over the previous year, and when asked what was dampening graduate hiring, AI ranked second behind cost control, at a net 47%. In Deloitte’s surveys, the expectation that AI would reduce hiring appeared first as a question about graduates. By October it had become an expectation about headcount overall, in a survey where cost reduction had also risen to a strong priority for 68% of CFOs, from 53% in the second quarter.
Hidden Cost: FRP’s parallel poll of 251 lenders and investors in the UK found 43% of Scottish respondents “not confident borrowers fully capture the cost of AI adoption in their business plans”, and 43% saying weak delivery of anticipated benefits could feed into the price or margin of a loan. A headcount saving that is forecast but not delivered is something 43% of the Scottish lenders and investors FRP polled say could affect the price or margin of a loan.
Who sits in the gap
Entry-level recruits, administrative staff and the finance teams making the forecast
The groups the surveys name as affected are not the finance chiefs making the forecast. Deloitte’s second-quarter survey asked about AI’s effect on graduate hiring before the third asked about headcount overall. The ONS September bulletin found that, among AI-using businesses with ten or more employees, 38% said administrative and clerical staff were the most affected, up nine percentage points from when the question began in December 2025, with creative and design roles next at 24%.
Finance teams themselves are affected in a second way. If a headcount saving is written into a 2027 budget on the strength of an expectation, the FRP lender findings suggest that the people reviewing that budget from outside will want to see how the saving is to be delivered, what it will cost, and what happens if it arrives late.
| Group | What the surveys say | What is still unknown |
|---|---|---|
| Graduates and entry-level recruits | Deloitte Q2: AI a net 47% dampener on graduate hiring over the next year | Whether the dampening is realised or absorbed by the cost-control motive ranked above it |
| Administrative and clerical staff | ONS: the role most often named as affected, 38% of AI-using businesses with 10+ employees | Whether “affected” means reduced, retrained or reorganised; the ONS does not break this down |
| Finance teams and lenders | FRP: 43% of Scottish lenders and investors not confident borrowers capture AI costs | How many 2027 budgets already book AI headcount savings |
| Housing providers and other public-interest sectors | Housemark: 59% access, 14% majority use, 11% with a formal strategy | What adoption looks like once strategies now in development are in place |
What this means for boards booking the saving
The surveys do not say CFOs are wrong. They say the saving most CFOs expect has, so far, been realised by a small minority of adopting businesses, and that where effects on roles are reported they fall most often on administrative and clerical staff. A board that treats the 52% as a market signal about what peers are doing may be reading an expectation as a delivered result. The ONS finding that only 10% of adopters use AI extensively, and the CFOs’ own placing of themselves at a quarter of the change they expect, both point the same direction: the workforce effect is being planned before the operating change that would produce it.
What remains unresolved
The two measures of adoption are moving apart
The ONS itself flags that executive surveys and business-population surveys return very different adoption rates for the UK. As long as boards read the executive figures and statisticians publish the population figures, the two groups will keep describing different economies. Neither is wrong, but a workforce plan built on the first will be tested against the second.
”A quarter of the way” has no shared definition
Deloitte does not say what the full expected change would look like, so the 67% figure records a feeling of distance rather than a measured one. It is still informative: the same respondents who expect payroll effects within a year describe themselves as far from the destination that would justify them, and only 3% say they are most of the way there.
Job creation and job loss are being counted separately
Lloyds finds a majority of businesses saying AI has created jobs. Deloitte finds a majority of CFOs expecting it to remove them. Both can be true in the same firm, as our analysis of HSBC’s wealth cuts alongside banks’ AI hiring argued. None of the surveys gathered here reports a net figure, and the ONS publishes its headcount result as reduced or increased, not by how much.
Reality Check: The ONS summarises its own evidence plainly: most businesses report that AI has not changed their overall headcount, and the Bank of England’s panel found no material employment effect for most firms over three years. Any claim that the UK has already seen AI-driven job losses at scale is running ahead of the official record, whichever way it is intended.
What to watch
Four readings over the coming months will show whether the expectation is converting or fading.
- The Budget later this month. Deloitte’s CFOs rated domestic economic and political uncertainty the top constraint on UK competitiveness, at 78 out of 100, and its chief economist tied that to near-constant speculation over tax. The report notes that John Healey, now Chancellor, delivers his first Budget later this month; gov.uk gives the date as 28 October. If the uncertainty score falls in the next survey, the cost-reduction mood that sits behind the headcount expectation may ease with it.
- The ONS headcount line. The ONS business survey asks AI-using businesses what adoption has done to their headcount, and its latest bulletin covers September 2026: 6% reduced and under 1% increased. A move above 10% by spring 2027 would be the first official sign that CFO expectations are being delivered. A flat line would say the expectation has moved further than the economy.
- Deloitte’s next CFO survey. If the 52% holds or rises while the progress distribution does not shift, the gap between planned and delivered becomes the story in Deloitte’s own numbers. A fall in the headcount expectation would suggest the Q3 reading was partly a cost-reduction mood, which rose from 53% to 68% in the same survey.
- Housemark’s full report. The social housing survey found 42 of its 66 respondents developing or planning an AI strategy. The published report, with its sector benchmarks and its findings on whether providers can show a return on investment, will show whether a sector with wide access and thin use closes the gap once those strategies are in place.
Boards and finance teams have a short window in which both figures are public and neither has been tested. The expectation is now on the record, with a date attached. The ONS publishes the delivery data in its business bulletins. By this time next year it will be clear which of the two was describing the UK economy.
Strategic Insight: The useful question for a board is not whether AI will reduce headcount but which of the five measures, expectation, access, use, depth or outcome, its own organisation is actually tracking. The ONS found that larger businesses in particular were often unsure how extensively AI is used in their own operations, which suggests many can answer the first two and far fewer the last.
Source: Deloitte CFO Survey Q3: Economic and political uncertainty seen as the greatest constraint on UK competitiveness (Deloitte UK, 14 October 2026) and the full Deloitte CFO Survey Q3 2026 report (PDF). Comparison data from the ONS Business insights and impact on the UK economy: 8 October 2026 and Artificial intelligence in UK businesses: 2023 to 2026; the Deloitte CFO Survey Q2 2026; the Lloyds Business Barometer; Housemark’s survey as reported by Housing Digital; FRP Advisory’s survey as reported by Insider Media; and Crowe’s Manufacturing Outlook Report as reported by the Greater Birmingham Chambers of Commerce.
This strategic analysis was written by Resultsense, a UK-focused AI news and analysis publication. We will be tracking the ONS headcount line and Deloitte’s next quarterly reading to see whether the expectation converts. Read more analysis at Insights, or get in touch.