UK AI adoption: the government data on the integration gap

UK AI adoption sounds settled. Ask what "using AI" means in practice and the picture changes fast. The UK Business Data Survey 2026, published by the Department for Science, Innovation and Technology (DSIT) in June 2026, found that 41% of businesses that handled digitised data said they used AI-based technologies. Not 41% of all UK businesses - 41% of the businesses that already had digital data to work with. That distinction matters, and so does what happened next in the same survey.

Of the businesses that said yes to AI, only 21% said their AI tools were integrated into existing business systems. The report's own example is Microsoft Copilot sitting inside Word, Excel and PowerPoint - AI that lives where the work already happens. The other four in five are using AI somewhere else: a browser tab, a standalone chat window, a tool nobody connected to the accounts package, the CRM, or the job sheet. That gap between adoption and integration is the entire reason so many businesses say AI hasn't given them any time back.

UK AI adoption vs AI integration: what's the difference?

Adoption means someone in the business has opened an AI tool and used it. Integration means that tool is wired into the systems where the business actually runs - its documents, its records, its workflow. DSIT's data shows plenty of the first and not much of the second: 41% adoption among digitised businesses, 21% integration among those adopters. Do the maths - 21% of the 41% - and fewer than one in ten businesses handling digitised data have AI doing anything more than sitting in a tab.

We've been making this argument from our own case work for a while. What's new is that it now has official statistics behind it too. The assessments we run keep finding the same pattern at the individual business level: an owner or a member of staff has a ChatGPT tab open, gets genuinely useful answers out of it, and none of that ever touches the spreadsheet, the booking system or the inbox where the actual admin sits. DSIT has just measured that pattern at national scale.

What UK businesses are actually using AI for

The survey also asked what businesses use AI for, on the same base of businesses handling digitised data. The two most common answers were researching information (28%, which the report describes explicitly as using AI "in place of a traditional search engine such as Google") and summarising or collecting in-house information, or drafting reports and correspondence (21%). Both are useful. Neither is integration. Looking something up and drafting a paragraph are one-off tasks you do inside the AI tool itself - they don't require the tool to connect to anything else, which is probably why they're the most common use cases. The uses that would actually save recurring hours - AI reading your job sheets, updating your records, chasing your invoices - need the wiring that only one in five adopters have done.

Why our own stats page says 54%, not 41%

Our own stats page on UK small business AI adoption leads with a 54% figure from a different survey. DSIT's figure here is 41%. Neither number is wrong. The 54% comes from the British Chambers of Commerce Insights Unit with the University of Essex, surveying UK SMEs in early 2026. The 41% comes from DSIT, surveying UK businesses that handle digitised data. Different surveys, different businesses, different definitions of "using AI", different bases to calculate a percentage from - DSIT's own report says as much, noting that "estimates from other sources vary, and differences in how AI use is defined, as well as variation across tasks and roles" affect the headline figure. Adoption surveys are notoriously sensitive to exactly what counts: does one person trying ChatGPT once count the same as a business running AI across its operations? Different surveys answer that differently, and you get different numbers as a result.

What DSIT's data adds is the first breakdown we've seen of what happens after adoption, not a more "correct" adoption number. Whether adoption sits at 41% or 54%, the integration figure - 21% of adopters - is the one that explains why so many businesses using AI still haven't seen it save them any real time.

The regulation gap is the same shape

DSIT also asked the businesses using AI about regulatory guidance, and the pattern repeats. 53% had at least heard of it and 46% were not aware of any guidance at all. Even among those who had heard of it, only 19% found it clear. Awareness without understanding, in other words - the same shape as adoption without integration. Businesses know AI exists and know rules exist. Very few have got past knowing to actually using either one properly.

Closing the gap

None of this means the four in five businesses without that wiring did anything wrong. Opening a tab and asking AI a question is the obvious first move, and it does produce real value - faster research, quicker drafts. The mistake is stopping there and assuming that's what "using AI" was supposed to deliver. It was never going to hand back hours from a tab that nothing else talks to.

This is a different problem from the one we've written about before, where a pilot stalls because nobody owned it, nobody measured it, or the timeline was wrong - or where a promising trial loses momentum and needs restarting. Those posts are about formal projects that lose momentum. This is about everyday tool use that was never wired into anything to begin with.

Closing that gap - moving AI out of the tab and into the systems where the admin actually lives - is what the HoursBack quiz is built to spot in your business, in under ten minutes. If you want the fuller picture, including where the hours are actually going and what to fix first, the HoursBack Assessment goes through it properly.