In November 2026 the CIPD, the UK professional body for people in HR, reported that 21% of UK employers expect AI to reduce their headcount over the next 12 months. Is that 21% evidence that AI is already cutting jobs? It is not. The figure is a forecast for the coming year given by HR decision-makers, 61% of them expect no change at all, and the shares expecting change in either direction are highest where AI is built deeply into the work.

01The 21% is a 12-month forecast from HR decision-makers, not a count of cuts

Start with what the number measures. The CIPD survey asked employers how they expect AI to change their headcount over the next 12 months. It did not ask how many people had already been let go because of AI, and it does not count them.

The answers fell into three groups. Some 21% expect AI to reduce headcount, 61% expect it to make no difference, and 4% expect an increase. The headlines I read carried the first number, not the majority answer that headcount will stay where it is.

The expectation of change was not spread evenly. Among employers that do not use AI, only 5% expect cuts. Among those that have built AI deeply into their work, the share rises to 44%, and 17% of that same group also expect headcount to grow.

Anyone who wants to use an AI-and-jobs figure for decisions at their own workplace needs to check two things. The first is whether the number is a forecast or a record of cuts that actually happened. The second is how deeply the employers who gave it are using AI.

02YouGov asked 2,051 HR leaders about expectations and counted no actual cuts

That the 21% is a forecast becomes clear once you look at who was asked, when, and about what. The CIPD publishes the survey as its Labour Market Outlook, and the autumn 2026 edition came out in November.

The fieldwork was done by the polling firm YouGov. Between 17 September and 11 October 2026, YouGov surveyed 2,051 senior HR professionals and people who make hiring decisions in UK organisations, online. The CIPD weighted the results by the size of employment, so larger employers count for more.

1

Who

2,051 senior HR professionals and hiring decision-makers in the UK.

2

When

Online, from 17 September to 11 October 2026.

3

Asked

The expected effect of AI on headcount over the next 12 months.

The design narrows the meaning of the figure in two ways. The people answering were not the workers whose jobs are at stake; they were the people who run HR. And they were describing what they think will happen over the next year, not what has already happened.

The press then chose a number of its own. People Management, a trade magazine for HR, reported 25%. That is the share among employers who know AI is being used in their organisation, a narrower group than all employers. The news site Resultsense also put "a quarter of employers" in its headline. One survey thus put two different figures into circulation: the CIPD's 21% for all employers and the 25% the press chose.

Figure 1 From the survey question to a 21% headline
reportedAsk for a12-month…YouGov,online2,051 HRleaders…Weightedresult:…61% expectno changeNewsheadlines21% or 25%No countof actual…reportedAsk for a 12-month expectationYouGov, online2,051 HR leaders answerWeighted result: 21%61% expect no changeNews headlines21% or 25%No count of actual cuts
The 21% is a weighted share of expectations, not a number of jobs cut. Reports headlined either 21% of all employers or 25% of employers aware of AI use.

Neither figure is a count of people who lost their jobs. Both are shares of HR leaders describing what they expect.

03Expected cuts split by size: 33% at large private firms, 12% at small ones

The 2,051 forecasts also divide sharply by size and sector. In the CIPD figures, 33% of large private-sector employers expect AI to reduce headcount. Among small and medium-sized private firms the share is 12%. The public sector sits at the overall level of 21%.

The large-firm figure is more than double the figure for smaller firms. The headline number combines large firms, smaller firms and the public sector, weighted by size of employment. A reader who works for a large company and a reader who works for a small one should be looking at different figures.

The previous year's survey gives a hint about which jobs employers have in mind. In autumn 2025, 62% of employers who expected cuts pointed to roles such as clerical work and junior management. That breakdown is from 2025; I did not find the same breakdown for autumn 2026 in what I read.

One caution for readers in the pharmaceutical industry. The survey's sector breakdown does not list pharmaceuticals as a separate category. There is no way to read from this survey what HR leaders at drug companies expect.

04Expectations of change rise with the depth of AI use

Large private employers were the most likely to expect cuts, and the next question is where that tilt comes from. The CIPD also cut the answers another way, by how far each employer has built AI into its work, in four stages.

Among employers that do not use AI, 84% expect no change in headcount. Among those still trialling it, 69% expect no change. The share expecting cuts is 5% among non-users and 15% at the trial stage. It climbs to 32% where AI is integrated into some of the work, and to 44% where it is deeply integrated.

Figure 2 Share expecting AI to cut headcount, by depth of AI use
Depth of AI useNot used: 5%84% expect no changeTrial stage: 15%69% expect no changePartly integrated: 32%Deeply integrated: 44%17% expect growthDepth of AI useNot used: 5%84% expect no changeTrial stage: 15%69% expect no changePartly integrated: 32%Deeply integrated: 44%17% expect growth
Each figure is the share of employers expecting AI to reduce headcount. Deeply integrated employers are also more than four times as likely as average to expect growth.

The deeply integrated group stands out in a second way. In that group, 17% also expect AI to increase their headcount. Across all employers the figure is 4%, so the deeply integrated group is more than four times as likely to expect growth.

The CIPD reads this pattern as reflecting plans to redesign work at least as much as plans to cut. On that reading, employers that use AI deeply are the ones thinking concretely about how tasks are allocated and how staff are retrained. That is the CIPD's interpretation. The survey lines up answers collected at a single point in time; it does not show that deeper use of AI came first and the expectation of cuts followed.

Whether depth of use explains the higher figure at large firms cannot be settled here, because the figures I read do not say whether large firms use AI more deeply than small ones. From these numbers alone, it is not possible to say whether size or depth of use is moving the answers.

05In a year, the share expecting cuts rose from 17% to 21%

Answers split by depth of use. The next step is to see how answers to the same question moved over a year. The CIPD asked a question with the same intent in autumn 2025, when 2,019 HR leaders responded.

MeasureAutumn 2025 (2,019 respondents)Autumn 2026 (2,051 respondents)
AI will reduce headcount17%21%
No difference49%61%
AI will increase headcount6%4%

Over the year, the share expecting cuts rose by four points and the share expecting no change by twelve. The share expecting growth fell by two points. Among employers aware of AI use in their organisation, the share expecting cuts rose from 21% to 25%.

The CIPD attaches a caveat to this comparison itself. In the autumn 2026 survey, the route by which respondents reached this question was changed. The CIPD therefore notes that comparing the two years directly carries some difficulty. Any statement that the share "rose" should be read together with that caveat.

There is one more number that should be kept apart from AI. In the same survey, 20% of employers plan to make redundancies in the next three months. That figure covers redundancies for any reason, not only AI. The expectation of AI-related cuts and the plan to make redundancies soon are answers to two different questions.

06Separate forecasts from outcomes, depth of use, and changes to the work itself

The share of employers expecting cuts went up by four points in a year, subject to the CIPD's caveat about the changed question routing. To avoid misjudging what that means, readers need to change the way they read such figures in three respects.

First, do not treat a headline "prediction" as a count of jobs lost. The CIPD question asks about expectations for the coming 12 months; it does not count anyone who has been made redundant. Reading the headline as "AI has cut a fifth of jobs" creates an answer to a question the survey never asked.

Second, read the answers by depth of AI use. Among deeply integrated employers, 44% expect cuts and 17% expect growth. Looking only at the overall figure hides the fact that, in the same group of employers, expectations of shrinking and of growing sit side by side. Which stage your own workplace is at determines which figure is relevant.

Third, look at the content of the work before the number of people. The CIPD writes that hiring patterns, job roles and the skills employers need may change before headcount does.

1

Hiring patterns

Who is hired, and at which career stage, may change before headcount does.

2

Job content

The job title may stay the same while the work it covers is reorganised.

3

Skills required

With the same number of people, the mix of skills the work needs may shift first.

Research on generative AI shows the output of the work changing, separately from any question of headcount. The economist Erik Brynjolfsson and his co-authors studied a workplace where 5,179 customer-support agents were given help from a generative AI tool. Among the least experienced agents, the number of issues resolved per hour rose by 34%. For experienced, highly skilled agents, the effect was minimal. The study divided workers by experience and skill, and it did not look at changes in headcount.

The three ways of reading share one purpose: not to apply the headline number to your own workplace as it stands. Once you add whether it is a forecast or an outcome, how deeply the employers use AI, and what part of the work is changing, the parts that do not apply to you come into view first.

07Whether forecasts became cuts can only be checked against figures outside this survey

How, then, can anyone check whether the headcount forecasts came true? The CIPD put a question with the same intent to employers in autumn 2025 and autumn 2026. If it asks again next autumn, there will be three years of expectations to compare.

Even so, this survey cannot tell us whether the expectations turned into actual cuts. Each round asks about the next 12 months, so each round produces another forecast. What actually happened has to be checked against employment statistics and against the staffing numbers companies announce.

Figure 3 Reading an AI-and-jobs figure
The headlinefigureForecastor count?Whoanswered?HR, executivesor workersSplit bydepth of useCheckagainst…Labourstatistics,…The headline figureForecast or count?Who answered?HR, executives or workersSplit by depth of useCheck against countsLabour statistics, company reports
Check the question's time frame and who answered, split the answers by depth of AI use, and only then compare them with actual counts.

Some things remain unverified. The deeply integrated employers expect growth as well as cuts, but the figures do not say in which jobs the new hires would come. Nor can the survey tell us how answers differ in individual sectors such as pharmaceuticals.

Here is my own view of what may happen next. If more employers build AI deeply into their work, more of them are likely to expect headcount to move in one direction or the other. This is only an inference from answers collected at one point in time. How it will show up in actual hiring and redundancies will not be known until next autumn's answers can be set beside outside statistics.

  1. The headline 21% is a 12-month forecast from HR decision-makers, not a count of jobs cut. A headline "prediction" should not be read as an outcome.
  2. Expected cuts range from 5% where AI is not used to 44% where it is deeply integrated, and 17% of the deeply integrated also expect growth. Read the figure by depth of use.
  3. The CIPD notes that hiring patterns and job roles may change before headcount does. In your own workplace, check the work itself before the numbers.
Closing

The finding that 21% of UK employers expect AI to cut headcount is not evidence that AI is already cutting jobs. It is a forecast for the coming year from HR decision-makers. Expectations of change are most common among employers that use AI deeply, and 17% of those same employers also expect growth. When an AI-and-jobs headline arrives, first ask whether it is a forecast or a count, then ask how deeply the employers behind it use AI, and only then bring it to decisions about your own workplace.

Sources & references
  1. CIPD (Cockett, J. and Ghazaryan, A.). Labour Market Outlook – autumn 2026. 2026-11. https://www.cipd.org/globalassets/media/knowledge/knowledge-hub/reports/2026-pdfs/9270-lmo-autumn-2026-report_final.pdf (21%, 61% and 4%; figures by size, sector and depth of AI use; survey method; note on the changed question routing)
  2. People Management. UK hiring intentions remain near record lows, CIPD finds. 2026-11-16. https://www.peoplemanagement.co.uk/article/1975304/uk-hiring-intentions-remain-near-record-lows-cipd-finds (the 25% among employers aware of AI use)
  3. CIPD. Labour Market Outlook – autumn 2025. 2025-11-10. https://www.cipd.org/globalassets/media/knowledge/knowledge-hub/reports/2025-pdfs/9024-lmo-autumn-2025-report-web.pdf (17%, 49% and 6% in the previous year; 2,019 respondents)
  4. Brynjolfsson, E., Li, D. and Raymond, L. (NBER). Generative AI at Work. Working Paper 31161. 2023-04. https://www.nber.org/papers/w31161 (5,179 customer-support agents; 34% more issues resolved per hour among novice and low-skilled agents)
  5. Resultsense. CIPD: quarter of employers expect AI headcount cut. 2026-11-16. https://www.resultsense.com/news/2026-11-16-cipd-quarter-of-employers-expect-ai-headcount-cut/ (coverage headlining the 25%)
  6. Staffing Industry Analysts. One in six UK employers expect AI to cut jobs in the next year. 2025-11-10. https://www.staffingindustry.com/news/global-daily-news/one-in-six-uk-employees-expect-ai-to-cut-jobs-in-the-next-year (roles employers expected to shrink in the autumn 2025 survey)