Between 16% and 54% of UK businesses use AI, and both numbers are right

If you have felt lately that everyone else has worked this out and you have not, some of that is the statistics.

There are four serious pieces of UK research on how many businesses use AI. They were all done properly. They disagree by a factor of three. Here they are, with what each one is actually counting, because the counting is the whole story.

The four numbers

54% — the British Chambers of Commerce, with Atos and the University of Essex's Centre for Micro-Social Change, published March 2026. Around 94% of the firms surveyed were SMEs. The trend is the striking part: 23% in 2023, 25% in 2024, 35% in 2025, 54% now.

47% — Simply Business's 2026 SME Insights Report, up from 22% the year before.

35% — the Office for National Statistics, for businesses with ten or more employees, as of June 2026. Up from roughly 12% in late 2023. For the smallest businesses — nought to nine employees, which is most of our readers — it is 28%. For firms of 250 or more, 49%.

16% — the Department for Science, Innovation and Technology's AI Adoption Research, based on 3,500 business interviews. Around one in six UK firms have deliberately deployed at least one recognised AI technology for a defined business purpose. Another 5% have concrete plans to.

Why they disagree

Not because anyone is wrong. Because "using AI" is four different questions.

The ONS is admirably honest about its own limits. Its methodology, in its own words, "treats all reported uses as equivalent and does not distinguish between light use, user-level use, and more embedded, production-level adoption." One person who tried an AI summary in Outlook last Tuesday counts the same as a firm that rebuilt its quoting process around it.

It also measures named technologies rather than a vibe — large language models at 18%, visual content creation at 16%, machine learning for data processing at 12%, image processing at 6%. Ask "do you use AI?" and you get one answer. Ask "do you use any of these four specific things?" and you get a smaller, harder one.

The DSIT figure is smaller still because it adds two words that do a lot of work: deliberately and purpose. Somebody decided to do this, for a reason, and could tell an interviewer what the reason was.

So the 54% and the 16% are not in conflict. They are measuring the distance between somebody in the building has a ChatGPT tab open and the business has changed how it works.

The gap is the interesting bit

Sit those two numbers next to each other and you get the actual state of UK business in 2026. Roughly half of small firms have AI somewhere on the premises. Roughly one in six have done anything deliberate with it.

The ONS found the same gap from the other direction, and this is the single most useful figure in any of this research: 55% of employees report using AI personally, while only 35% of businesses report using AI at all.

Read that twice. In a meaningful number of companies, AI adoption is happening to the business rather than by it. Staff have quietly worked out that it makes their job easier and got on with it, on their own accounts, without anyone deciding anything. That is not adoption. That is what adoption looks like from a distance if you squint.

What this means if you are five people and feeling behind

Three things follow, and none of them is "hurry up."

You are less behind than the headlines suggest. When a trade magazine says half of your competitors use AI, the honest translation is that half of them have someone with a subscription. The share who have genuinely changed how the work gets done is one in six, and among businesses your size it is smaller than that — adoption tracks with headcount at every measure, 28% for the smallest firms against 49% for the largest.

The gap that matters is not the one being reported. The distance between you and a competitor who owns a ChatGPT licence is roughly zero, because that is a distance you can close this afternoon for £20. The distance between you and one of the 16% is real, and it is not made of software. It is made of somebody having sat down and decided what should change.

Uncontrolled adoption is a live liability, not a head start. If your staff are in the 55% and your business is in the other 65%, work is already being done on AI in your name, on personal accounts, with consumer privacy defaults, on data you are legally responsible for. That is worth a look this week regardless of what you decide about any of the rest of it.

One thing we are not going to do

Tell you what the number will be next year.

Every one of these surveys measures the past, and the honest reading of four surveys that disagree by a factor of three is that the thing being measured is still too loosely defined to forecast. Anyone quoting you a 2027 adoption figure to a decimal place is selling something.

What we would say instead is duller and more useful. The 23% to 54% climb over three years is real and it is fast. But almost all of it happened in the loose definition — people trying things. The strict number has not moved anything like as far. Whatever is genuinely hard about this, it is not getting hold of the tools.

Sources

• Half of SMEs using AI — with limited headcount impact so far — British Chambers of Commerce, March 2026
• Artificial intelligence in UK businesses: 2023 to 2026 — ONS
• What does firm-level data tell us about AI adoption in the UK? — Bennett School of Public Policy, University of Cambridge, on the DSIT figures
• Simply Business SME Insights Report 2026

Figures checked on 8 September 2026.