AI adoption statistics, reconciled
Ask what percentage of businesses use AI and the honest answer is a range, not a number: 16% (DSIT), 25% (ONS) and 35% (British Chambers of Commerce), three UK surveys fielded across roughly the same year. All three are correct. They sampled different businesses, by different methods, and asked a different question about what counts as using AI.
By Sunny Patel published figures last checked
The honest answer, and why it moves
Ask what percentage of businesses use AI and you can produce almost any figure between one in six and one in three, and be correct each time. Three UK surveys, fielded across roughly the same period, put adoption at 16%, 25% and 35%. None of them are wrong. They are not measuring the same population, and they are not asking the same question.
Start with the lowest number, because it explains the method problem most clearly. DSIT's figure, 16% of UK businesses using at least one AI technology, comes from a telephone survey of 3,500 businesses with five or more employees. That sampling frame excludes businesses under five employees, which is most UK businesses by count. DSIT's own breakdown shows why the exclusion matters: micro businesses sit at 14%, large businesses at 36%. A survey that starts by cutting out the smallest, least-adopting firms should, if anything, read higher than one that does not. It still comes out lowest of the three, which tells you something about how much weight the other two surveys are putting on breadth of sample rather than size of business.
The UK's three numbers
Line the three up and the disagreement stops looking like noise and starts looking like three different instruments pointed at three different things.
| Survey | Figure | Sample | Who was asked | Fieldwork | Why it differs |
|---|---|---|---|---|---|
| DSIT 2025 | Around 1 in 6 UK businesses (16%) currently use at least one AI technology: micro 14%, mid-sized 23%, large 36% | 3,500 businesses with 5+ employees, plus 100 qualitative interviews | UK businesses with 5 or more employees, by telephone, plus 100 qualitative interviews | 12 February to 2 May 2025 | Excludes businesses under 5 employees, which is most UK businesses by count. The narrowest sample of the three, and the lowest figure. |
| ONS 2025 | Approximately a quarter (25%) of businesses report currently using some form of AI, up 15 percentage points since the question was introduced in late September 2023. 44% among businesses with 250+ employees | the BICS business panel | the BICS business panel, a national statistics survey | 15 to 28 December 2025 | A wider panel than DSIT's, fielded seven months later, over a period ONS's own trend line shows adoption climbing. |
| British Chambers of Commerce 2025 | Over a third of SMEs (35%) say they are actively using AI technology, up from 25% in 2024. 46% of B2B service firms, against 26% of B2C firms and manufacturers. Only 11% use it to a great extent | 1,500+ business leaders | British Chambers of Commerce members, surveyed online with Intuit QuickBooks | June to July 2025 | Surveys only Chamber members, who the release itself describes as more AI forward than UK businesses generally. The least representative sample, and the highest figure. |
Compiled by implementai.today from the sources named. We ran no survey. Machine-readable version: /data/ai-failure-rates.json.
The pattern holds even inside the two national statistics surveys that are supposedly measuring the same thing. ONS's 25% is not simply more than DSIT's 16%: it comes from a wider panel, fielded seven months later, over a period ONS's own trend line shows climbing. And the British Chambers of Commerce's 35%, the largest figure of the three, comes from the least representative sample of the three: its own members, whom the release itself describes as more AI forward than UK businesses generally.
The size gradient, which is the finding everyone skips
Every one of these surveys, when it breaks the figure out by firm size, finds the same shape: large firms adopt at roughly double the rate of small ones. That is a more useful number than any headline above, because it is the one figure that survives translation between surveys.
| Survey | Smallest firms measured | Largest firms measured |
|---|---|---|
| DSIT 2025 | Micro: 14% | Large: 36% |
| ONS 2025 | not broken out below 250+ employees in the published bulletin | 250+ employees: 44% |
| US Census Bureau 2026 | Four or fewer employees: under 20% | 250+ employees: 37% |
Read across the DSIT and Census rows and the doubling is close enough to trust: this is not a quirk of one survey design, it turns up on both sides of the Atlantic, using three different methods and two different currencies of business size. If you only remember one number from this page, it should not be 16%, 25% or 35%. It should be that whichever survey you pick, the largest firms in it are adopting AI at roughly double the rate of the smallest.
Adoption is not value
Using AI and getting anything from it are different questions, and most headlines only answer the first one. McKinsey's 2025 global survey found 88% of organisations report regular AI use in at least one business function, and 72% use generative AI specifically. Read past the headline and the same survey finds only 7% have fully scaled AI, and just 39% report any enterprise-level EBIT impact.
BCG's 2025 executive survey draws the same gap from the other direction: 5% of companies are achieving AI value at scale, its top maturity band, while 60% are achieving no material value at all. And IBM's CEO study found that CEOs themselves rate only 25% of AI initiatives as having delivered the ROI they expected, with just 16% scaled enterprise wide. Adoption is the easy number to report. This is the one that actually matters, and it is not the one that gets quoted.
How these numbers get misquoted
Three specific errors recur often enough to be worth naming, each checked against the primary text.
| What circulates | What the source actually says | Source |
|---|---|---|
| "UK AI adoption rose from 9% to 25%" | ONS states a rise of 15 percentage points to 25%, which puts the baseline at 10%, not 9%. A small error, and it is in almost every retelling. | ONS 2025 |
| "Only 4% generate substantial value" and "26% have capabilities to move beyond experimentation" | The 4% is BCG's 2024 figure; the 2025 report puts future-built firms at 5%. No 26% figure appears anywhere in the 2025 report. It is not a BCG category. | BCG 2025 |
| "88% of companies use AI", cited as evidence that AI adoption is solved | The same survey reports that only 7% have fully scaled AI and nearly two thirds have not begun scaling at all. Regular use in one function is a low bar. Adoption and value are different measurements. | McKinsey 2025 |
The pattern is the same in all three: a hedge, a hard-won maturity band or a low-bar usage question gets flattened into a plain measurement, and the flattened version is the one that keeps circulating. ONS's baseline drops a percentage point in the retelling. BCG's 2025 report gets credited with a 2024 number and a category it never used. McKinsey's low bar for "use" gets read as evidence the whole subject is solved, in the same document that says two thirds of companies have not begun scaling.
How to quote an adoption figure honestly
A bare percentage is not a citation. Before repeating any adoption figure, and before implementai.today does, name four things.
- The survey and the year it was fielded, not just the organisation that publicised it.
- The sample: who was asked, and how many.
- The fieldwork window, since a survey run in December 2025 and one run in May 2025 are not describing the same market.
- The literal question asked. "Uses at least one AI technology" (DSIT), "currently using some form of AI" (ONS) and "actively using AI" (BCC) are three different questions wearing the same word.
None of this tells you whether the AI in question does anything useful, which is a separate and much less quotable question, covered in why AI projects fail and reconciled the same way, source by source, in the AI project failure rate.
Questions, answered from the sources
What percentage of businesses use AI?
There is no single figure. UK surveys taken across roughly the same year report 16% (DSIT), 25% (ONS) and 35% (British Chambers of Commerce). The US Census Bureau's Business Trends and Outlook Survey put overall usage between 17% and 20% from December 2025 to May 2026. All of these can be correct at once, because none of them measure the same population or ask the same question.
What is UK AI adoption?
Depends which survey you read. DSIT's telephone survey of 3,500 businesses with 5 or more employees found 16% use at least one AI technology. ONS's Business Insights and Conditions Survey, fieldwork 15 to 28 December 2025, put it at 25%, rising to 44% among businesses with 250+ employees. The British Chambers of Commerce, surveying its own members with Intuit QuickBooks, found 35% of SMEs actively using AI.
Why do the numbers differ so much?
Different samples and different questions. DSIT excludes businesses under 5 employees, which is most UK businesses, and asks about any AI technology. ONS surveys a wider national panel and asks about current use. The British Chambers of Commerce surveys only its own members, whom its own release describes as more AI forward than UK businesses generally, and asks about active use. A higher or lower figure does not mean a survey is wrong. It means it counted a different group.
How many companies have actually scaled AI?
Far fewer than the adoption figures suggest. McKinsey's 2025 global survey found 88% of organisations use AI in at least one function, but only 7% have fully scaled it. BCG's 2025 executive survey put companies achieving AI value at scale at 5%, with 60% achieving no material value at all. IBM's CEO study found only 25% of AI initiatives delivered the ROI their own CEOs expected, and just 16% had scaled enterprise wide.
Are small businesses using AI?
At roughly half the rate of large ones, in every survey that breaks out firm size. DSIT found 14% adoption among micro businesses against 36% among large ones. The US Census Bureau's Business Trends and Outlook Survey found under 20% adoption among firms with four or fewer employees against 37% for firms with 250 or more. ONS does not publish a smallest-firm figure, but its 44% rate for businesses with 250+ employees follows the same pattern.
Related: why AI projects fail, the causes behind the numbers. And the AI project failure rate, reconciled the same way: RAND's more than 80% next to MIT's 95% and BCG's 5% at scale.