UK AI Adoption Statistics: Why the Numbers Disagree

What percentage of UK businesses use AI? The Office for National Statistics says about 3 in 10, with 30% of all businesses reporting use in September 2026 and 38% of those with 10 or more employees. A Bank of England panel of firms with 10 or more employees, surveyed a few months earlier, puts it at 88%. Both are careful pieces of work, and they open with the same question, word for word.

This page sets the main UK figures side by side, explains what separates them and says which to use for what. The headline numbers disagree widely, but the surveys agree on something more useful, which is how few businesses have embedded AI in the way they work. That is the number we think a leadership team should compare itself against.

UK AI adoption statistics, side by side

These are the most recent results we could check at source, with the newest fieldwork first.

We have left out polls we could only find reported in press releases, because we could not check how they were asked or who answered.

Why do AI adoption surveys give different numbers?

The simplest reason is timing, because this number is moving quickly. The government's AI adoption survey, commissioned by the Department for Science, Innovation and Technology (DSIT), went from 16% in early 2025 to 32% a year later, so surveys run even a few months apart are hard to compare. Timing is not the whole story, though, since the ONS measure grew by only a little under half over much the same period, by our calculation.

Then there is who gets counted. The same ONS wave reads 30% for all businesses and 38% once firms with fewer than 10 employees are taken out. The UK Business Data Survey's 41% is a share of businesses that handle digitised data, and the same answers come to 35% of all businesses.

The authors of the paper using the Bank of England panel measured two more causes. The same responses give 88% weighted by employment and 83% weighted by the number of businesses, and chief executives and finance directors report 89% against 84% for other respondents. They conclude that size, weighting and seniority "each move measured adoption by between five and twelve percentage points".

What counts as AI matters as well. The UK Business Data Survey counts "free or paid for tools", including AI used in place of a search engine. DSIT counts piloting. ONS lists technologies with no tool names, so a firm whose staff use a chat assistant has to recognise it as "text generation using Large Language Models". Whether that counts as using AI is a fair question in its own right. As ONS puts it, "Lower estimates tend to capture direct, bottom-up use", while higher ones reflect "a broader, top-down view of whether firms use any AI technology across their operations".

The gap nobody has explained

Those causes do not close the gap, and the paper's authors say they "narrow the range of estimates in the literature without closing it". In mid-2026, ONS read 35% among firms with 10 or more employees. The Bank of England panel read 83% when weighted the same way, over overlapping dates and with the same opening question. Causes worth five to twelve points each do not obviously add up to 48.

Our guess is that part of the answer lies in who joins a panel, how the options are labelled and the "not sure" answer that ONS offers. That is our inference, and we have not seen anyone measure it.

Which figure to use

No survey here is the single correct one, so the choice depends on your question.

• For the whole UK economy, use ONS. It has the largest response of the surveys above, at 9,933 businesses, and it repeats regularly.
• For change among private firms with five or more staff, use DSIT, bearing in mind that its response rate fell from 9% to 5% between waves.
• For what senior leaders of larger firms report, use the Bank of England panel, remembering that the paper is a working paper and has not been peer reviewed.

Whichever you choose, quote the date and the size limit beside it. "32% of private businesses with five or more staff" is a fact, and "a third of UK businesses" is a looser claim than the survey supports.

How many mid-sized UK businesses use AI?

About half on official counts, and more than 80% on the Bank of England panel. DSIT found 47% of businesses with 50 to 249 employees. ONS data gives 46% for firms with 50 to 99 employees and 50% for those with 100 to 249, both our calculation from its dataset. The UK Business Data Survey puts medium firms at 57%, with a margin that runs from 50% to 64%. The panel reports 83% and 87% for those same two ONS size bands.

So the same size of firm can read as 50% or 87%, within a few months of each other. If you run a business of this size, neither number tells you whether you are behind.

How many UK businesses have properly embedded AI?

Very few, and on this the surveys stop arguing. They measure different things, so read these as a pattern and not a ranking.

• DSIT classes 2% of businesses as intensive adopters, meaning they have built bespoke AI or embedded it in their processes, rate it as highly important and use it constantly.
• Among firms with 100 to 249 employees that use AI or are unsure, 8.5% told ONS it was used extensively, by our reading of the dataset.
• The British Chambers of Commerce found that around one firm in ten is adopting bespoke AI, from an unweighted sample of its members.
• Make UK reports that "Only 2% of manufacturers say AI is widely embedded across their operations".

One figure seems to pull the other way. DSIT found that 57% of AI users use it constantly, rising to 69% of medium businesses. We read that as frequency and not depth, because a team can use an assistant all day without changing how a piece of work gets done. ONS makes a related point about its own headline, that "it treats all reported uses as equivalent and does not distinguish between light use" and heavier use, and we have written before about how shallow that use tends to be.

This is why we would stop comparing adoption. "Everyone is using AI" is true on some surveys and false on others, and it does not tell a leadership team what to do next. The more useful question is whether AI has changed how a specific piece of work gets done, with hours to show for it. When Designs4You automated its inbound RFQ processing, handling time fell from 20 minutes to under 90 seconds and more than 12 hours of weekly capacity came back. That is what depth looks like in one process.

Our AI maturity ladder sets out the steps between trying a tool and running on a system of your own, and the free AI roadmap shows you where you stand on it.

How we compiled these figures

We read every survey at source and not from press coverage, and we checked each quoted line against the published document. Percentages are as published, with three exceptions that come from our own calculations on the ONS dataset. Those are the 50 to 99 and 100 to 249 employee adoption figures, the extent-of-use figure and the growth between March 2025 and March 2026. For adoption, we take 100 minus "not sure" and "does not currently use", a method that reproduces the figures ONS publishes for other size bands.

ONS publishes new AI figures roughly every quarter, and we will update this page when it does. The address will stay the same, and the date at the top shows when it last changed. If you spot a mistake or know a survey we should include, write to us at hello@agenticise.co.uk.

Frequently asked questions

What percentage of UK businesses use AI?

The Office for National Statistics put it at 30% of all businesses in September 2026, and 38% of those with 10 or more employees. The government's own AI adoption survey found 32% of private businesses with five or more staff earlier in the year. A Bank of England panel of firms with 10 or more employees reports 88%, weighted by employment, so any single figure needs its date and its population beside it.

Why do AI adoption surveys give different numbers?

They are run at different times, count different sizes of firm, weight their answers differently, ask different people and define AI differently. Researchers working with Bank of England data found that firm size, weighting and the seniority of the person answering each move the result by five to twelve percentage points. Those factors narrow the gap without closing it.

How many mid-sized UK businesses use AI?

About half on official counts. The government's AI adoption survey found 47% of businesses with 50 to 249 employees, and ONS data gives about 50% for firms with 100 to 249 employees, by our calculation. The Bank of England panel reports 83% to 87% for firms with 50 to 249 employees, so the answer depends heavily on who is asked.

Which UK AI adoption statistic is the most reliable?

For the whole economy we would use the ONS figure, because it has the largest response of the surveys here, repeats regularly and publishes its method. No survey is the single correct one, though. Each measures something slightly different, so choose the one whose population matches your question and quote its date and size limit.

How many UK businesses have properly embedded AI?

Very few. The government's AI adoption survey classes 2% of businesses as intensive adopters, and Make UK reports that 2% of manufacturers have AI widely embedded across their operations. The British Chambers of Commerce puts bespoke AI at around one firm in ten. These measure different things, but they point the same way.