Official survey data show adoption climbing fastest in large firms and in tech. Where AI-using firms report any effect on staff numbers, cuts outnumber hires — but most report no change at all.
In late September 2026, 30% of UK businesses told the Office for National Statistics they were using at least one type of artificial intelligence. When the question was first asked, in September 2023, the figure was 9%.1 The ONS puts the rise at 20 percentage points; the published shares are rounded, which is why they appear to differ by 21.
The figures come from the Business Insights and Conditions Survey, a fortnightly voluntary survey of private-sector firms. The latest wave was in the field from 21 September to 4 October and drew 9,933 responses from 38,592 businesses sampled, a response rate of 25.7%.1 All-business adoption was broadly stable against June 2026, so the three-year climb has levelled off over the summer, at least on this measure.
Adoption tracks size. About 51% of firms with 100 or more employees use AI, as do 51% of those with 250 or more. Among firms with fewer than 10 employees — the overwhelming majority of UK businesses by count — the figure is 29%.1 Information and communication is the most AI-intensive industry surveyed, at 60%. In June, construction was among the lowest, at 13%.2
The tools are mostly generative. Among firms with 10 or more staff, 21% use large language models to generate text, up 16 points since 2023, and 19% use AI for visual content, up 15 points.1
The survey asks AI-using firms with 10 or more staff which roles AI has affected most. Administrative and clerical work tops the list at 38%, up from 35% in June and 29% when the question was first asked in December 2025. Creative and design roles come next at 24%, broadly unchanged.1
“Affected” is not the same as “cut”. Asked directly about headcount, 6% of AI-using firms with 10 or more staff said it had fallen as a result, and fewer than 1% said it had risen. The ONS says both shares have been broadly stable over the past year.1 In June, among firms using AI to improve their operations, 63% reported no change in headcount.2
“Use” is broad: one employee generating text with a chatbot counts the same as a firm rebuilding its operations around machine learning. Only about one in four sampled firms responds, and the ONS labels the survey “official statistics in development”. It excludes the public sector and several industries, including finance and insurance, agriculture and energy generation. The headcount question measures what firms say AI caused, not what happened to jobs across the economy — firms that never adopted AI, or that hire less without cutting, do not appear in it. The ONS publishes the full industry and size breakdown only in its datasets, which this piece did not use.