AI in UK business, 2026: every number, sourced.

Ask five UK institutions what share of businesses use AI and you will get five answers between 16% and 71%. None of them is wrong. They ask different businesses, in different months, and they mean different things by "use". A trade body whose members are engaged enough to answer a survey will report a higher rate than a statutory survey of 38,000 firms drawn at random, and a question that counts a staff member using a free chatbot will score higher than one that requires AI to be built into a business process.

The table below is the map. Every figure on this page is quoted with the base it was measured on and the month of fieldwork, and where two sources disagree we say so rather than average them. The ONS series is the one to quote if you want a number that will still stand up in a year, because it is the largest sample, the most conservative definition, and the only one with a consistent history back to 2023.

Sources: ONS (Jul 2026 and Jan 2026), DSIT (Feb 2026 and Jun 2026), BCC (Mar 2026), Lloyds (Aug 2026), BFI working paper 2026-47 using Bank of England data (Mar 2026). Full references at the foot of the page.

Size is the strongest predictor of adoption in every survey. The ONS puts AI use at 49% among businesses with 250 or more employees and 28% among those with fewer than ten, and the gap has not narrowed as overall use has climbed. DSIT’s survey of data-handling firms finds the same shape at a higher level: 82% of large businesses, 58% of medium, 51% of small, 41% of micro businesses and 40% of sole traders.

Sector matters almost as much. Information and communication leads on every measure, at 58% in the ONS series and 62% in DSIT’s; professional, scientific and technical services follow at 54%. Construction sits at 13% in the ONS data, the lowest of any sector, and the British Chambers’ panel finds manufacturing at 45% against 78% for IT and digital firms. Regionally, adoption in the BCC sample is highest in southern England and noticeably lower in the North East, Yorkshire and the devolved nations.

Two figures are worth holding on to. Large language models are now the most used AI technology among businesses with ten or more staff (18%), ahead of visual content creation (16%) and machine-learning data processing (12%). And of the businesses that use AI at all, only 10% say they use it extensively. Most adoption is still shallow.

Improving business operations is the most common use in every size band, reported by close to 60% of adopters in the ONS series. Personalising products or services is a distant second, at around 31% among the smallest firms. This matches what we see in our own work: the first thing a small business automates is admin, not the product.

The barriers are consistent across sources and they are not what vendors tend to say. Skills come first. DSIT finds 60% of businesses citing limited AI skills, expertise or knowledge as a factor preventing adoption, rising to 68% among firms that intend to adopt. Lloyds puts cost, data quality and skills almost level at 18%, 17% and 17%. The ONS finds that 41% of businesses with ten or more employees report no barriers at all, and that cost affects only 7% to 14% depending on size.

The most striking gap is between adoption and governance. Only 5% of UK businesses have a formal written AI policy, against 56% of large ones. Just over half are aware of AI regulatory guidance, but 10% say they understand it and 19% find it clear. Among small firms, 92% still have concerns about data security, copyright and liability, and a third say better guidance on implementation would encourage them to adopt.

Spend is smaller than the coverage suggests. Among UK firms that have invested in AI, the largest group, a third, spent under £25,000. Eighteen per cent spent between £25,000 and £100,000, 8% between £100,000 and £250,000 and 7% more than that. The last official estimate by business size is now old but still the only one: in 2020 the average small business adopter spent £9,500, the average medium business £380,000 and the average large business £1.6 million, on a national total of £16.7 billion.

For the people, the cleanest data is advertised vacancies. IT Jobs Watch measures a median contract rate of £575 a day for an AI consultant in the UK and the same £575 in London, up 6.98% on a year earlier, with a 90th percentile of £734 nationally and £738 in the capital. That is drawn from 95 UK contract vacancies in the six months to 13 September 2026, against ten in the same window two years before. Permanent AI engineers are advertised at a median of £87,500. Across every permanent IT job that mentions AI, the median is £74,915 and such jobs are now 8.61% of all UK IT vacancies.

Two notes on reading these. A day rate is a price for time, not a price for an outcome, which is why our own engagements are quoted as fixed prices against a written scope (see the consultant cost survey linked below for how the two compare). And IT Jobs Watch figures are rolling six-month windows that change daily, so always record the "six months to" date when quoting them.

Every UK survey finds a large majority of adopters reporting productivity gains and a much smaller share reporting revenue gains. Lloyds finds 87% of AI users reporting higher productivity and 48% higher profits. DSIT, with a stricter sample, finds 75% reporting improved productivity but only 12% reporting increased revenue, with 77% seeing no change in revenue at all. The FSB finds 59% of small-firm AI users seeing productivity gains and 24% increased revenue, with an average revenue uplift of 3% among firms that adopted in the last year.

The most sobering figure comes from the Bank of England’s Decision Maker Panel: 89% of UK firms report no impact of AI on their productivity over the past three years. They expect it to raise productivity by 1.9% and lower employment by 1.4% over the next three. PwC’s survey of UK chief executives is similar: 21% say AI increased revenue in the past year, 74% report little or no change, and only 45% have a clear roadmap for AI.

On jobs, the evidence points the same way from every angle. The ONS finds that 63% of businesses using AI to improve operations saw no change in headcount, 6% saw a decrease and 1% an increase; in the December 2025 wave, 4% of AI users said headcount had fallen because of it. The British Chambers find 95% of SME adopters reporting no impact on workforce size. Lloyds finds 54% of businesses saying AI has created new jobs. The one exception is bespoke AI: around one in five firms that have built their own systems report staffing reductions, against 3% of firms using only generic tools.

Individuals are ahead of their employers. The ONS notes that 55% of employees report using AI for work or education while 35% of businesses report using any AI technology, and puts it plainly: staff use AI tools independently. Ofcom’s 2026 tracker finds 54% of UK adults using AI tools such as ChatGPT, Copilot or Gemini, up from 31% a year earlier, with 79% of 16 to 24 year olds and 74% of 25 to 34 year olds.

The scale of that use is visible in traffic. ChatGPT recorded 252 million UK web visits in August 2025, up 156% on the year, and 1.8 billion in the first eight months of 2025; Gemini, Claude and Perplexity each more than doubled. More than half of UK adults (53%) say they often see AI-generated summaries in their search results, and Ofcom estimates around 30% of searches now show one. For anyone whose customers find them through search, that is the number to plan around.

Attitudes are mixed and worth quoting accurately. In June 2026, 41% of adults in Great Britain had used AI for work or education in the past year, 27% at home and 23% for hobbies; 33% had not knowingly used it at all. Thirty-six per cent expected AI to benefit them personally, 27% did not. Eighty-one per cent were worried about telling fake news from real, and 38% thought AI posed more risks than benefits.

The Google figure is included because it is widely quoted. It is a sponsor-commissioned poll and sits well above the statutory and regulator series; name the sponsor if you use it.

The UK has no AI Act. It regulates AI through existing law and sector regulators, and the ICO’s data is the best measure of how far AI has reached into decisions about people: in 2024, 8% of UK organisations reported using AI decision-making tools when processing personal information and 7% reported using facial or biometric recognition. The public’s tolerance is narrow. In recruitment, 64% believe employers will rely too heavily on AI and 61% are concerned it will do worse than a human decision-maker.

Security lags adoption. Among businesses using, adopting or considering AI, only 24% have cyber security practices in place to manage the risks of AI, and 43% of all businesses experienced a breach or attack in the last year. Fifty-seven per cent of adults are concerned their data is not held securely. These are the figures behind the questions a careful buyer asks a supplier about where data goes and who can see it.

For UK firms selling into Europe, the EU AI Act’s transparency duties for chatbots and AI-generated content apply from 2 August 2026, while the high-risk obligations have been deferred to December 2027 and August 2028. At home, the government reports an AI Security Institute of more than 100 researchers that has tested 30 frontier models, a commitment to AI training for 10 million workers by 2030, and £28.2 billion of investment through AI Growth Zones.

London is where the UK AI industry is. DSIT counts 5,862 AI companies in the UK in 2024, up 58% on the year, and finds that London, the South East and the East of England account for around three quarters of registered offices. Beauhurst, using a broader definition, counts 6,030 active AI companies in London alone and records £5.73 billion raised by London AI companies across 682 funding rounds in 2025, the strongest year on record. The two counts cannot be combined; they define an AI company differently.

Adoption is high too. Lloyds puts London business adoption at 78%, second only to the West Midlands at 83%, against 46% in Scotland. Between 47% and 69% of the UK’s AI assurance suppliers are based in the capital. The price of AI expertise, though, is the same as everywhere else: the median London contract rate for an AI consultant is £575 a day, identical to the national median, with the 90th percentile four pounds higher at £738.

For the first half of 2026, UK AI startups raised $12.6 billion, more than four times the same period a year earlier and close to three quarters of all UK venture capital. Most of that money lands within a few miles of our office. Very little of it reaches the 10 to 50 person firms that make up most of London’s economy, which is the gap the rest of this site is about.

Three things follow from the data. First, you are not behind. Roughly a third of businesses your size use AI at all and one in ten of those use it extensively, so the honest position for most firms in 2026 is "we have tried a chatbot and automated nothing". Second, the returns reported by adopters are in productivity, not revenue, and mostly in admin: 38% of businesses say staff spend less time on it. A project that promises new revenue from AI is making a bigger claim than the evidence supports; a project that promises to remove a repeated task is not. Third, the barrier is skills and clarity, not cost. The ONS finds cost holding back at most one business in seven, and the firms that say they lack expertise are the ones that invest in training.

That is also why our own engagements are shaped the way they are. An audit before a build, a fixed price against a written scope rather than the £575 a day the market charges for time, a running-cost figure stated before go-live, and training for the people who will use the system. The two cost surveys linked below put our prices in the same tables as everyone else’s, so you can check.

This page was compiled on 13 September 2026 from primary UK sources: the Office for National Statistics, the Department for Science, Innovation and Technology, the Bank of England and FCA, the Information Commissioner’s Office, Ofcom, the British Chambers of Commerce with the University of Essex, the Federation of Small Businesses, Lloyds Banking Group, Barclays, PwC UK, IT Jobs Watch, Beauhurst and Dealroom. Every figure was read on the source page or in the source PDF, with the base and fieldwork date recorded beside it. Where a widely repeated figure could not be found in its supposed source, it was left out; the best-known example is a "48% of adults have used generative AI" line attributed to Ofcom, which does not appear in the report cited for it.

Figures from sponsor-commissioned research (Microsoft, Google) are included in one place each and labelled. Lloyds Business Barometer releases do not publish sample sizes on their press pages. IT Jobs Watch data is a rolling six-month window and changes daily. The DCMS spend-by-size estimate is 2020 data and predates generative AI. Nothing on this page is an Augustova estimate, and nothing is drawn from our own client or product data.

To cite this page: Augustova, "AI in UK business, 2026: every number, sourced", augustova.co.uk/guides/uk-ai-statistics, accessed on the date you read it. We update it as new waves of the ONS, DSIT and Ofcom series are published, and the date at the top changes when we do. If you find an error, email info@augustova.co.uk and we will correct it and note the correction here.