AI Hits UK Clerical Roles First—Headcount Cuts Remain at 6%

In its 8 October 2026 business bulletin, the Office for National Statistics found that 38% of AI-using UK businesses with at least 10 employees named administrative and clerical roles as most affected, while 6% reported an AI-related reduction in overall headcount. Creative or design roles came next at 24%; fewer than 1% of the same AI-using group reported an increase in headcount. The contrast is between two kinds of employer response, not between two shares of workers.

The figures come from a voluntary business survey conducted from 21 September to 4 October. Employers identified where the AI technologies they had adopted had the greatest effect and separately described the impact on the size of their workforce. The questions do not count how many clerical employees had their tasks changed, nor how many people left jobs at firms reporting lower headcount. They show where change is most often reported among adopters.

Clerical roles top employers’ responses

Administrative and clerical work now leads the survey’s ranking of affected functions. The share of AI-using firms naming it rose by three percentage points from June and by nine points from December 2025, when the role question was introduced. Creative or design work ranks second, with its share broadly stable over the latter period. The rise in the clerical response shows a growing concentration of reported effects within that function; it does not measure the number of posts eliminated there.

The unit of observation matters. Each business contributes a response about its roles, regardless of whether it employs a small administrative team or a large one. Nor does “most affected” specify whether the change concerns the volume of routine work, the tasks assigned to staff, the speed of a process or the number of employees. The survey can identify which functions adopters point to most often, but it cannot turn that ranking into a count of affected people.

Only a minority report a smaller workforce

On 9 October, Grade One Press reported that 6% of businesses using AI had cut staff. That is a firm-level result: a business reporting a reduction is counted once, whether the change involves one post or many. The published percentage cannot be multiplied by the national workforce to estimate job losses, and it does not identify which occupations accounted for any reductions.

The headcount question concerns a firm's overall staffing, whereas the role question concerns where its adopted AI has had an effect. A firm could report altered clerical work while retaining its employees, or reduce staffing in another function. The survey does not pair a clerical-role response with a clerical headcount change. Its headcount proportions have been broadly stable over the past year even as the clerical share in the role question has risen.

This also limits what can be inferred about cause. Respondents attribute the workforce change to AI, but the survey is a record of their assessments, not an experiment that separates AI from shifts in demand, hiring plans or reorganisation. It captures an employer's reported outcome at the time of the survey. It does not show whether a lower headcount came through layoffs, departures left unfilled or another route.

Adoption is uneven across UK firms

The workforce figures describe businesses that already use AI, a narrower group than all UK firms. Across the survey, three in ten businesses reported using at least one AI technology in September. Among firms with at least ten employees, the adoption share was higher, at 38%, while it was 29% among those with fewer than ten. These figures measure whether a business uses a listed technology, not how extensively it has reorganised work around it.

Firm size therefore changes both the likelihood that AI use is reported and the pool of employers answering the role and headcount questions. The survey recorded text generation with large language models and visual content creation as the most commonly used AI technologies among businesses with at least ten employees. Those categories cover different activities. A business reporting any use can appear alongside one relying on AI more routinely, so adoption alone cannot explain the strength of a reported role effect.

The technology changes which work is affected

An earlier ONS analysis of June responses found that 53% of businesses using machine-learning image processing reported effects on administrative or clerical roles and on creative or design roles. Among those using machine-learning data processing, 41% reported an effect on administrative or clerical roles and 39% on data analysis roles. More than half of businesses using AI for visual content creation identified effects on creative or design work.

These are technology-specific groups of adopters, not separate slices of the entire business population. Firms could use more than one technology and name more than one affected role. That explains why the broad clerical ranking in the latest survey need not describe the effect of every AI application equally. Image processing, data processing and visual content tools are associated with different patterns of reported role change, even when they are all counted as AI adoption.

The earlier analysis also found that training or retraining existing staff was the most commonly reported way businesses integrated AI skills. That response is consistent with employers altering the work done inside existing roles, though the survey does not trace an individual employee's tasks before and after adoption. It helps explain why a role may be affected without an accompanying fall in the firm's total workforce.

The business survey is voluntary and its estimates are official statistics in development, so movements across waves warrant care. The next scheduled business bulletin is due on 22 October; whether it repeats the same AI questions will determine when the distinction between role effects and headcount changes can next be measured on the same basis.