A new CIPD survey finds that 25% of UK employers using artificial intelligence expect to reduce their workforce within the next year. But official statistics reveal a more complicated picture of AI-related job losses, occupational risks and the future of employment in Britain.
Artificial intelligence is becoming a factor in British employers' decisions about how many people they need. For some businesses, the technology offers a way to complete existing work with fewer employees. Others expect AI to change the responsibilities of their workforce without reducing its size.
The distinction matters. A company can automate administrative work without making anyone redundant. It can reduce recruitment, leave vacant positions unfilled or transfer employees into different roles. Each decision affects the labour market differently.
The Chartered Institute of Personnel and Development (CIPD) published its Autumn 2026 Labour Market Outlook on 16 November, reporting that 25% of employers already using AI expect the technology to reduce their headcount over the following 12 months.
That proportion has increased from 21% in autumn 2025. Meanwhile, 59% of AI-using employers still expect no change to their workforce numbers.
The findings raise a question that extends far beyond technology companies: is Britain approaching a period of AI-related job losses, or are employers reconsidering how work is organised before making substantial changes to employment?
What the CIPD survey actually tells us
The CIPD surveyed 2,051 UK employers for its Autumn 2026 Labour Market Outlook. Of these, 1,641 reported that AI was being used within their organisations.
Among employers using AI, the survey found: 25% expect AI to reduce headcount over the next 12 months, 59% anticipate no change, 5% expect AI to increase headcount. The remaining respondents were uncertain about the effect.
These figures represent employers' expectations. They do not establish how many employees have already lost their jobs, how many redundancies will occur or whether the anticipated reductions will eventually materialise.
An employer expecting to reduce headcount may be considering redundancies, natural staff turnover or slower recruitment. The survey's headline percentage cannot be converted into a national estimate of jobs at risk without additional information about workforce size and the scale of planned reductions.
The CIPD also reports that 21% of all surveyed employers, including organisations not currently using AI, expect the technology to reduce headcount. That distinction explains why the one-in-four figure applies specifically to employers already using AI.
There is another limitation when comparing the results with 2025: the CIPD changed how respondents were routed to the questions, which may slightly affect year-on-year comparability.
Nevertheless, the survey provides evidence that workforce reduction has become a more prominent expectation among organisations adopting AI.
Is AI causing unemployment in the UK?
There is evidence that AI has already contributed to workforce reductions in some British businesses. There is not yet sufficient evidence to determine how much of the UK's overall unemployment is attributable to the technology.
The Office for National Statistics (ONS) provides a useful comparison.
In September 2026, 6% of UK businesses with at least ten employees that were using AI reported a reduction in headcount as a result of adopting the technology. Fewer than 1% reported an increase.
The ONS figure measures businesses reporting an effect that has already occurred. The CIPD figure measures employers anticipating future reductions.
The two surveys therefore answer different questions. Their populations and methodologies also differ, so the gap between 6% and 25% cannot be interpreted as a forecast of additional redundancies.
The wider UK labour market offers further context.
According to the ONS figures released on 17 November 2026, the unemployment rate was estimated at 4.6% for July to September 2026. The rate was 0.3 percentage points lower than both the previous quarter and the corresponding period a year earlier.
At the same time, the number of available jobs had weakened. UK vacancies stood at approximately 690,000 in August to October 2026, below their pre-pandemic level.
A reduction in vacancies can make finding employment harder even when existing employees are not being dismissed. Businesses may freeze recruitment because of weak customer demand, rising labour costs or uncertainty over investment.
AI adds another possible reason for employers to reconsider vacancies. If existing staff can complete more work using automated tools, a company may decide that replacing a departing employee is unnecessary.
Such a decision would not necessarily appear as an AI-related redundancy. Its consequences would instead emerge through fewer opportunities for jobseekers.
The evidence therefore supports a narrower conclusion than the claim that AI is driving British unemployment: some employers already report AI-related reductions, and more expect them, but the national employment effect remains difficult to isolate from broader economic conditions.
Which jobs are most at risk from AI in 2026?
The occupations most exposed to AI are not necessarily those facing the largest immediate losses of employment.
Exposure measures the extent to which technology can perform or assist with tasks within an occupation. Whether that leads to fewer workers depends on the decisions employers make, the cost of implementing AI and the demand for the services those workers provide.
The ONS offers evidence of where British businesses are observing the greatest disruption.
Among AI-using businesses with at least ten employees, 38% identified administrative and clerical roles as the roles most affected by AI adoption in September 2026. Creative and design roles followed at 24%.
These percentages measure businesses identifying affected occupational groups, not the proportion of employees made redundant.
Administrative and clerical work
Administrative occupations contain tasks that can often be standardised: processing documents, extracting information, preparing routine correspondence, maintaining records and producing summaries.
AI can help perform these activities without requiring an employee to complete every step manually.
That creates a potential economic incentive to reduce the number of hours devoted to routine administration.
However, workplace administration also includes exception handling, confidential decisions, communication and responsibility for accurate records. Automating individual tasks does not necessarily remove the entire occupation.
The more immediate employment effect may arise when an employer combines previously separate responsibilities into a smaller number of positions.
Creative and design occupations
Creative professionals face a different form of exposure.
Generative AI can produce preliminary images, layouts, marketing copy and variations of existing content. This may reduce the time spent on some production tasks.
But a faster first draft does not establish that the final result meets a client's requirements, complies with intellectual property rules or can be published without additional review.
The economic question is whether automation reduces the total cost of delivering acceptable work.
For some employers, AI may reduce demand for routine production. Others may use the same technology to increase output without reducing their creative workforce.
The ONS findings establish that these occupations are experiencing change. They do not demonstrate that creative employment is disappearing.
Software development and technical work
Software development presents a particularly complicated case.
AI can generate code, identify errors, prepare tests and assist with technical documentation. Yet software engineering also involves requirements analysis, security, architecture, deployment and responsibility for the operation of systems.
The UK government's 2026 assessment of AI capabilities and the labour market identifies evidence of weaker recruitment in some highly AI-exposed occupations.
However, falling recruitment cannot automatically be attributed to AI. Investment conditions, the correction following pandemic-era technology hiring and changes in business demand may contribute.
This is particularly relevant for junior developers, whose work may include tasks that AI tools can now assist with.
A company may reduce recruitment at entry level while continuing to require experienced engineers for technical oversight. That would create a different long-term problem: fewer opportunities for workers to acquire the experience required for senior positions.
The relationship between AI and junior software engineering recruitment is examined separately in Prime Economist's existing coverage. The wider CIPD findings provide another reason to monitor the issue, but they do not establish that junior developers are the occupation most likely to disappear.
Which five jobs are most likely to survive AI?
Searches about which jobs will survive AI reflect a reasonable concern about the security of future employment. However, no credible occupational assessment can guarantee that a particular profession will remain unaffected.
A more useful distinction concerns the types of work that AI currently struggles to replace entirely.
Five occupational groups illustrate why some jobs may be less exposed to complete automation.
1. Electricians
Electrical installation and maintenance require physical work in buildings, diagnosis of faults and compliance with safety requirements.
AI may assist with diagnostics, documentation and work planning. It cannot, by itself, replace the physical installation and inspection involved in many electrical jobs.
2. Plumbers and heating engineers
These occupations involve work in unpredictable physical environments, often requiring direct access to buildings and equipment.
Automated diagnostics and scheduling may improve efficiency, but repairs and installation continue to depend heavily on practical skills.
3. Registered nurses
AI can assist with clinical information, documentation and aspects of decision support. Nursing also involves patient assessment, hands-on care and professional accountability.
Technology may change how nurses work, but the presence of automatable administrative tasks does not mean the nursing profession can be fully replaced.
4. Social care workers
Personal care frequently requires physical assistance, communication and responses to changing individual circumstances.
AI may support record-keeping and care coordination. It cannot currently provide the full range of physical and interpersonal assistance delivered by human carers.
5. Equipment maintenance technicians
Maintaining industrial and commercial equipment involves physical inspection, troubleshooting and repair.
AI-supported monitoring can identify potential faults, but many interventions still require trained technicians working directly with machinery.
These examples are based on the nature of occupational tasks, not on a verified ranking of Britain's five safest professions.
The distinction is significant. A job can be difficult to automate completely while still experiencing substantial technological change. Moreover, demand for workers depends on economic conditions as well as automation.
The better question is therefore not which three, five or ten jobs will remain untouched, but which occupations contain tasks that require human physical intervention, professional judgement or direct responsibility, and how much of their work AI can realistically perform.
Which jobs could grow as AI becomes more widely adopted?
AI adoption may reduce demand for certain tasks while increasing demand for people responsible for implementing, operating and verifying automated systems.
The UK's Skills England annual skills report 2026 identifies substantial uncertainty about the employment effects of AI. It also points to changes in the skills workers need and the potential for new opportunities.
In the digital and technologies sector, its associated assessment identifies demand for software development, information technology infrastructure, quality assurance and related technical occupations.
For example, Skills England projects demand for an additional 69,000 programmers and software development professionals within the digital and technologies sector between 2025 and 2035. The projection is based on sector growth assumptions and should not be interpreted as a forecast of jobs created specifically by AI.
It also highlights how automation is changing the distribution of technical responsibilities. Workers may spend less time producing routine outputs and more time reviewing, validating and managing technology-enabled processes.
This suggests potential demand for specialist capabilities in AI implementation, cybersecurity, data systems, software assurance and technology governance.
However, identifying an occupation as strategically important does not establish that its employment will increase because of AI. Some of these professions may themselves become more productive through automation.
The outcome will depend partly on whether businesses expand their activities as costs decline or retain the savings without increasing output.
Why do large UK employers expect more workforce reductions?
One of the most revealing findings in the CIPD survey concerns differences between large employers and smaller businesses.
Among private sector organisations, 33% of large employers expect AI to reduce headcount over the next year. The equivalent figure for small and medium-sized enterprises is only 12%.
A further 77% of private sector SMEs expect no change in headcount, compared with 49% of large private employers.
The size distinction follows the CIPD's categories: private SMEs employ fewer than 250 people, while large private employers have at least 250.
AI adoption also differs.
According to the CIPD, 29% of private sector SMEs report no AI use, compared with just 6% of large private employers. Deep integration of AI is also more common among larger organisations.
These differences suggest a plausible relationship between AI adoption and anticipated workforce restructuring.
Large organisations may have more standardised processes across multiple departments. They may also have the budgets and technical resources required to implement automation on a substantial scale.
For smaller companies, replacing labour with technology may be less straightforward.
A business with fifteen employees may rely on individuals who combine several responsibilities. Automating one part of a person's work does not necessarily remove the need for that employee.
There may also be limited financial capacity to introduce new systems, integrate them with existing operations and manage the resulting risks.
The CIPD findings do not establish which of these mechanisms explains the difference. They do, however, show that AI-related workforce reductions are not expected uniformly across British businesses.
That matters when discussing the economic consequences of automation. A national headline can conceal substantially different experiences among employers of different sizes.
Does higher AI productivity necessarily mean fewer workers?
The CIPD reports that 62% of employers using AI believe it has improved job performance.
But the degree of improvement deserves attention.
Of those employers, 52% report a small improvement and 10% report a substantial improvement. A further 28% report no difference.
These are assessments of job performance rather than audited financial outcomes.
An employee who completes a task faster does not automatically generate an equivalent reduction in company expenditure.
For example, if a five-person team completes its existing workload in fewer hours, management has several options. It can reduce staffing, assign additional work, increase output or allow the team to spend more time on activities that were previously neglected.
Each option produces a different economic outcome.
Reducing staffing could lower salary expenditure, although the company must consider any restructuring costs and whether the remaining employees can maintain service quality.
Increasing output could generate additional revenue, but only if sufficient customer demand exists.
Alternatively, the company may retain its workforce while improving service, reducing errors or shortening delivery times.
AI adoption also carries costs, including software licensing, integration, employee training and human verification of automated work.
The relationship between productivity and headcount therefore depends on business decisions rather than technological capability alone.
There is an additional finding in the CIPD survey: among employers with deeply integrated AI, 44% expect workforce reductions, while 17% expect increases.
This is consistent with technology affecting organisations differently even when adoption is advanced. Deep integration does not imply that every employer intends to cut jobs.
For the British economy, the distinction between productivity and profitability is consequential. Productivity improvements create the potential for economic gains, but whether those gains result in employment growth, higher wages, lower prices or larger profits remains an open question.
What happens to young workers and entry-level jobs?
Workforce reductions are only one way technological change can affect employment.
A company might retain its experienced staff while cancelling planned junior vacancies. Its headcount may remain stable even as opportunities for new entrants decline.
This creates a particular concern for graduates and younger workers who depend on entry-level positions to acquire professional experience.
The 2026 Skills England annual report identifies evidence of pressure on graduate recruitment but also warns against attributing the decline entirely to AI.
It cites research indicating that employment reductions may be concentrated in junior roles among AI-exposed firms. Other evidence points to a wider slowdown in graduate recruitment that predates the current wave of AI adoption.
The UK government's assessment of AI and employment similarly acknowledges changes in recruitment for highly exposed occupations while identifying broader economic explanations.
This matters because a slowdown in entry-level recruitment can have consequences that are not immediately visible in unemployment statistics.
Graduates may accept positions outside their intended profession, remain in temporary employment or take longer to establish their careers.
Employers may also face longer-term difficulties if fewer workers obtain the experience required for advanced responsibilities.
The problem is particularly relevant to technology companies, where progression from junior to senior positions depends on practical exposure to complex projects and professional supervision.
But AI is unlikely to be the only factor behind changes in entry-level opportunities. Recruitment budgets, business investment, demand for services and labour costs must also be considered.
The available evidence supports concern about access to early-career employment. It does not establish that AI is the dominant cause of Britain's graduate recruitment difficulties.
Will AI create more UK jobs than it eliminates?
At present, there is no reliable basis for concluding that AI will create more jobs than it eliminates in the UK, or that the reverse is inevitable. Some technological improvements reduce the labour required for particular activities. Others lower production costs, improve products and increase demand. The employment outcome depends on how these effects interact. If AI enables a business to deliver services more cheaply and attract additional customers, employment may increase despite higher productivity per worker. If demand remains unchanged, the same productivity improvement may reduce the amount of labour required.
Businesses may also reorganise their workforce, reducing some positions while creating others.
The CIPD survey reflects this uncertainty. Most employers using AI still anticipate no headcount change over the next year, even as the proportion expecting reductions has risen. The ONS data add evidence that some workforce reductions have already occurred, but the reported prevalence remains substantially lower than the share of employers expecting future cuts.
At the same time, UK labour market conditions cannot be explained by AI adoption alone. Changes in vacancies, investment, operating costs and consumer demand continue to influence employers' decisions.
These factors may reinforce each other. A company facing rising costs might adopt AI to improve efficiency, but it could also reduce recruitment regardless of whether the technology delivers measurable savings.
Establishing the specific contribution of AI requires evidence that goes beyond employer expectations, including workforce data over time, recruitment records and information about how organisations use technology.
What Britain's AI employment figures reveal
The headline finding from the CIPD is serious: one in four UK employers using AI expects to reduce headcount within twelve months.
But the wider evidence does not support the conclusion that a quarter of British jobs are about to disappear.
Most employers surveyed do not expect AI to change their workforce numbers. Smaller businesses are less likely than large employers to anticipate reductions, and official ONS data indicate that a minority of AI-using businesses have so far reported actual headcount cuts attributable to the technology.
The occupational evidence also resists simple predictions. Administrative and creative roles are already experiencing changes in how tasks are performed, while technical and hands-on professions face different forms of exposure.
For people entering the labour market, reduced recruitment may prove as consequential as redundancies. For businesses, the economic results of automation depend on how productivity improvements translate into staffing decisions, output and demand.
The unresolved question for Britain is how many employers' expectations will become measurable employment changes over the next year, and whether businesses will use AI primarily to reduce labour costs or to expand what their existing workforce can achieve.
The next stage of evidence will come from observed changes in recruitment, employment and business performance, rather than forecasts alone.
Sources and methodology
1. CIPD (2026). Labour Market Outlook: Autumn 2026. Primary source for employer expectations, organisational AI adoption, workforce size comparisons and reported job performance improvements. See particularly Figures 12–18. Survey of 2,051 UK employers, including 1,641 reporting AI use.
2. Office for National Statistics (2026). Business insights and impact on the UK economy, 8 October 2026. Business Insights and Conditions Survey, Wave 165. Used for AI adoption, affected occupational groups and reported changes to headcount.
3. Office for National Statistics (2026). Employment in the UK: November 2026. Published 17 November 2026. Used for UK unemployment and employment indicators.
4. Office for National Statistics (2026). Vacancies and jobs in the UK: November 2026. Used for the latest UK vacancy estimates.
5. Department for Science, Innovation and Technology (2026). Assessment of AI capabilities and the impact on the UK labour market. Used for occupational exposure, recruitment patterns and limitations of causal attribution.
6. Skills England (2026). Annual skills report 2026. Used for employment exposure, skills requirements and graduate recruitment uncertainty.
7. Skills England (2026). Sector skills needs assessment: Digital and technologies. Used for occupational demand projections and changes to technical roles.
Methodological note: CIPD measures employers' stated expectations; ONS business surveys report employers' observations of effects that have already occurred. The occupational examples are analytical illustrations, not a statistically established ranking of jobs guaranteed to survive AI. The article does not claim that Prime Economist has conducted original employer interviews or independently measured AI-related redundancies.