Article ·
All insightsIn this article
The short version
Surveys published in 2026 put the share of businesses using AI at 32% (DSIT) and 30% (ONS). A third, the Cyber Security Breaches Survey, found 21% had adopted AI tools and 31% were using, adopting or considering them. The three measure different things in different ways, so they are not in conflict.
The 2026 AI adoption survey, commissioned by the Department for Science, Innovation and Technology (DSIT) and published on GOV.UK under the name of the Department for Business, Innovation, Science and Trade, found 32% of private-sector businesses with five or more employees using at least one AI technology, including piloting. We call it the DSIT survey for short. The Office for National Statistics (ONS) found 30% of all businesses, and 38% of those with 10 or more employees. The cyber survey's 31% includes businesses that are only considering AI.
The differences to check are whether piloting counts, which sizes and sectors are included, how the question is asked, and when and how the survey was run.
In our reading, use DSIT for the level among businesses with five or more employees (and its year-on-year change only with the caveats below), ONS for the movement across all businesses, and the cyber survey's 21% for adoption. Its 31% adds businesses that are adopting or only considering.
Three surveys, three numbers
Three UK government publications give a headline for AI use or adoption. The table sets out what each one asked, whom it covered and how it was run.
| Survey | Headline and question | Who is included | Fieldwork and method |
|---|---|---|---|
| DSIT AI adoption survey 2026 (published 29 October 2026) | 32% using at least one AI technology, up from 16% in 2025. The figure combines a statement question, much like the cyber survey's, with questions on five named technologies. Those ask whether the business is 'currently using it (including piloting)'. | Private-sector businesses with 5 or more employees. Public administration and defence, education, human health and social work, households as employers and extraterritorial bodies are excluded. | Telephone, 7 January to 21 April 2026. 3,503 interviews and a 5% response rate. No margin of error is stated. |
| ONS Business Insights and Conditions Survey, wave 165 (8 October 2026) | 30% of all businesses and 38% of those with 10 or more employees using at least one AI technology in September 2026. Asked: 'Which of the following artificial intelligence technologies, if any, does your business currently use?' | Private-sector businesses of all sizes, including the smallest (0 to 9 employees). Agriculture, oil and gas extraction, energy, public administration and defence, public education and health, and finance and insurance are excluded. | Online and voluntary, live 21 September to 4 October 2026. 9,933 of about 38,600 businesses responded (25.7%). Official statistics in development. |
| Cyber Security Breaches Survey 2025/26 (30 April 2026) | 31% were using AI, in the process of adopting it, or actively considering it. 21% had adopted some AI tools. Businesses chose which of several descriptions best fitted their organisation's use of AI tools. | Private businesses with more than one person on the payroll. Sole traders, public-sector organisations other than education, and businesses with no IT capacity or online presence are outside the scope. | Mostly telephone, 11 August to 12 December 2025. 2,112 businesses. Margin of error of about 1.6 to 2.6 percentage points. A new question this year. |
Sources: the three publications and their technical pages, read on 10 November 2026.
Why the figures differ
Five differences between the surveys could account for some of the gaps between the headlines. None of the publishers says how much.
The differences:
- Piloting. DSIT counts a business that is 'currently using it (including piloting)'. The ONS question asks which technologies a business 'currently' uses and does not mention piloting. The cyber survey separates 'We have adopted some AI tools in our organisation' from 'We are in the process of adopting AI into our organisation (including piloting)'.
- Size. DSIT leaves out businesses with fewer than five employees, oversamples large ones, and weights the results. ONS includes the smallest businesses and weights by count. ONS says businesses with 0 to 9 employees make up around nine in ten UK businesses, so they carry most of an all-business figure. DSIT found 27% of micro businesses (5 to 9 employees) using AI, against 52% of large ones. ONS found 29% of businesses with fewer than 10 employees.
- Sector. ONS leaves out finance and insurance, agriculture, energy and the public sector. DSIT includes finance, agriculture and energy, but leaves out public administration, education, and health and social work.
- Question design. ONS asks about specific technologies and then works out whether a business uses at least one. DSIT combines a statement question like the cyber survey's with questions on five named technologies. The cyber survey asks one question about how far the organisation has gone, from 'adopted' to 'not relevant'. Its 31% adds three answers: adopted (21%), in the process of adopting, including piloting (4%), and actively considering (6%). On DSIT's definition the first two, 25 points, are close to use, because the second includes piloting. The third, 6 points, is not.
- Timing and method. The cyber survey finished fieldwork in December 2025, DSIT in April 2026 and ONS in October 2026. DSIT and the cyber survey were mostly by telephone. The ONS survey is online and voluntary.
Which to use for what
This section is our reading, not the publishers' advice. No publication compares its AI figure with the others.
Use DSIT when you want the share of businesses with five or more employees using AI, and the barriers and effects they report. Treat the change since 2025 with the caveats in the next section. Use ONS when you want the movement across all businesses over time: it has asked a similar question since September 2023, when 9% reported use. Use the cyber survey's 21% as a single snapshot of adoption so far, since the question is new this year, and read its 31% as adoption, plus adoption in progress, plus active consideration.
Has adoption doubled?
DSIT says so: 'AI adoption doubled since 2025', from 16% to 32%. It says the second wave 'was designed to allow comparisons with the previous wave'. Both waves used the same population wording and the same commercial business database.
Some things changed. The contractor changed, from IFF Research and Technopolis Group to BMG Research. The average interview fell from 20 minutes to 15. The response rate fell from 9% to 5%. DSIT also says the 2026 questionnaire was adapted, with new and amended questions, and fieldwork ran in different months. DSIT does not say whether any of this affects the comparison, and neither report publishes its questionnaire.
The ONS series is the longer one. It rose from 9% of all businesses in September 2023 to 30% in September 2026. Its chart data give 23.5% for all businesses in September 2025 and 29.7% in September 2026, and 27.2% and 38.1% for businesses with 10 or more employees. ONS describes the figures as official statistics in development and advises caution.
Its July 2026 article, based on an earlier wave, gave 'around 35%' for businesses with 10 or more employees. The October bulletin gives 38% for the same group, up 3 percentage points from June 2026.
Small inconsistencies to check before you quote
The ONS bulletin of 8 October gives 51% for larger businesses. Its main points and the Figure 1 title say 250 or more employees. Its body text says 100 or more employees for the same figure. The chart data gives 50.9% for 250 or more and 50.4% for 100 to 249, so only the 250 or more band rounds to 51%.
DSIT's 2026 report gives real estate as 45% in its summary and in Figure 2, and as 42% in its body text. The 42% matches the combined finance and real estate row in DSIT's data tables. The report does not say which it meant.
Size, sector and region
By size, DSIT found 52% of large businesses (250 or more employees) using AI, 47% of medium businesses (50 to 249), 34% of small businesses (10 to 49) and 27% of micro businesses (5 to 9). ONS found 38% of businesses with 10 or more employees and 29% of those with fewer than 10.
By sector, DSIT found the highest use in information and communication (61%), real estate (45%) and business services and administration (39%). ONS found 60% in information and communication, the highest of any industry it surveyed.
By region, DSIT says: 'The regional variation seen in 2025 was hardly evident in 2026, with no significant differences in the proportion who had adopted AI.' Its data table gives 30% for the West Midlands against 32% overall, from 298 interviews in the region. Scotland was lowest at 26%.
ONS publishes no weighted regional AI estimate. Its wave 165 dataset does include unweighted regional percentages for each technology. For the share of businesses not currently using any, it gives 35.8% for the West Midlands and 40.1% for the UK. These are unweighted, so they describe the businesses that responded, not all businesses in the region. The cyber survey publishes no AI breakdown by region.
How to quote them
When you quote one of these figures:
- name the survey, the date and who it covers, for example '32% of private-sector businesses with five or more employees (DSIT, 2026)'
- do not put the cyber survey's 31% beside the 32% and 30% as if they measure the same thing
- say 'including piloting' when you use DSIT's figure
- avoid 'doubled' without saying that DSIT does not explain the effect of its changes of contractor, method and questions
- check which size band an ONS figure refers to
Only the cyber survey states a margin of error, about 1.6 to 2.6 percentage points depending on the result. DSIT states none. ONS publishes confidence intervals for individual technologies, and we found none for its 'at least one' figure in the files we read.
What we could not establish
Whether the change of contractor, interview length, response rate or questions affects DSIT's comparison of 16% with 32%. DSIT does not say.
How DSIT routes respondents between its statement question and its technology questions. The questionnaire is not published.
Any margin of error for DSIT headline figures, in either year. None is stated.
A confidence interval for ONS's 30% and 38% figures. None is published in the files we read.
Which size band the ONS body text means by 51%: 100 or more employees, or 250 or more.
Whether the ONS question counts piloting as use. The question does not say.
Any official explanation of why the DSIT and ONS figures differ. We found none.
Any weighted ONS AI figure for the West Midlands, or any cyber survey AI figure for the region. ONS publishes only unweighted shares of responding businesses, and the cyber survey publishes none.
How it was researched
We read each publication and its technical or methodology pages, and downloaded the official files, on 10 November 2026. We compared what each says about its question, population, fieldwork and weighting. We did not average, adjust or reconcile the figures ourselves, and we used no news coverage for any number. We drafted this article with AI assistance. A second AI reviewer, working separately, then checked each claim against the sources on 10 November 2026, and we changed the article where it found errors. BespokeWorks sells AI consulting and training, so we have an interest in this topic. Our suggestions on which figure to use are our reading, and we mark them.
Questions for your team
- Which survey is this figure from, and who does it cover?
- Does it count businesses that are only piloting or considering AI?
- Is it all businesses, or only those above a size threshold?
- When was the fieldwork done?
- Does the source give a margin of error?
Sources
Links open the source. Page dates are those shown on each page. All were read on 10 November 2026.
The DSIT AI adoption surveys
- AI adoption survey report 2026: executive summary (29 October 2026)
- Artificial intelligence adoption research 2026: full report (October 2026)
- AI adoption survey 2025 (published 28 January 2026)
ONS Business Insights and Conditions Survey
- ONS, Business insights and impact on the UK economy, wave 165 (8 October 2026)
- ONS, BICS questions, 21 September to 4 October 2026 (5 October 2026)
- ONS, Artificial intelligence in UK businesses: 2023 to 2026 (20 July 2026)
- ONS, Business Insights and Conditions Survey quality and methodology information
- ONS, BICS preliminary weighted results (14 September 2020)
