What does DSIT's AI adoption survey actually measure?
Someone in the room cites a government number and declares UK businesses barely use AI, or that adoption is surging, then looks at you for a verdict on your own company. Before you agree or argue, ask what was counted. The AI adoption survey DSIT ran asked one kind of question, of one kind of respondent, about one slice of the economy, not about your organization specifically.
DSIT's own published survey report states it completed 3,500 interviews with UK businesses, with fieldwork running between 12 February and 2 May 2025.
We think a single national snapshot like this is useful for reading the overall climate, not for grading one company. Treat it as weather, not a scorecard for your own building.
Sources: AI Adoption Research - GOV.UK
Who answered, and who's missing from DSIT's count?
In the AI adoption survey DSIT ran, interviewers spoke with the person responsible for technology at each business, not a cross-section of staff. That's a reasonable way to ask about formal AI investment decisions.
It also means the figures describe what leadership signed off on, not what every employee quietly does with a chatbot on a personal device. A tech lead can report minimal AI use while half the sales team already drafts emails with it on the side.
Sources: AI Adoption Research - GOV.UK
Why do AI adoption statistics disagree across countries?
Put DSIT's UK figures next to other countries' numbers and the gap is jarring. The US Census Bureau's Business Trends and Outlook Survey, analyzed by St. Louis Fed economists, found firm-level AI adoption of only 5% to 7%, far below the 35% to 40% range that US worker surveys reported over the same 2023-2025 period.
Eurostat's 2025 EU-ICT-Firm survey, as described by St. Louis Fed researchers, covered roughly 157,000 firms across 32 European countries and asked about eight specific AI technologies, a design that looks nothing like DSIT's single tech-lead interview or the AI adoption survey UK businesses answered.
A DSIT- and DCMS-supported review distinguishes diffusion, the population-level take-up of a technology, from an individual firm's adoption decision. Confuse the two and any set of AI adoption statistics starts looking contradictory when it's really measuring different things.
Whether you're holding DSIT's figure, another country's survey, or a vendor's report, check who answered and what question they were asked before you trust the percentage.
| Survey | Who answered | Period covered | What it measures |
|---|---|---|---|
| DSIT UK business survey | Person responsible for technology at each business | Feb-May 2025 | Firm-level tech use and self-reported business impact |
| US Census Bureau BTOS | Business officials, as analyzed by St. Louis Fed economists | 2023-2025 | Firm-level AI adoption |
| Eurostat EU-ICT-Firm | Firms across 32 European countries | 2025 wave | Use of eight specific AI technologies |
| Gallup US employee survey | Individual US employees, self-administered web panel | February 2026 | Self-reported frequency of AI use and attitudes |
Sources: AI Adoption Research - GOV.UK, AI in the Workplace: What Separates Adopters and Holdouts, Measuring AI Adoption among Firms: How You Ask Matters | St. Louis Fed, AI Adoption in Europe and the U.S. | St. Louis Fed, Factors influencing firms’ adoption of advanced technologies: A rapid evidence review - GOV.UK
Is 'AI improved our productivity' a result or a perception?
Perception, not a measured result, and DSIT says so itself. DSIT's own report flags that its productivity and impact figures are self-reported by businesses and should be read as estimates, not verified outcomes. A business owner saying AI saved time is a belief worth noting, not proof of a financial gain.
Gallup's February 2026 survey of 23,717 US employees used self-administered web surveys of a demographically weighted panel to capture how often people say they use AI and how they feel about it. Useful for attitudes. It still isn't a measurement of output.
We're skeptical of any adoption report that leads with 'felt more productive.' If a number can't be traced to a measured outcome, call it a perception until someone actually measures the outcome.
Sources: AI Adoption Research - GOV.UK, AI in the Workplace: What Separates Adopters and Holdouts
Can you compare your team's numbers to DSIT's national figures?
No, not directly. DSIT surveyed a different population, asked different questions, and captured one moment in a fast-moving space; your 200-person logistics team or your 40-person finance department is a different animal entirely.
What you can do is run a consistent internal check instead of borrowing someone else's. AI Adoption Score's predefined employee surveys assess the human side of AI adoption, covering skills, habits and culture, so a Digital Transformation Leader can track consistent questions across teams and over time rather than leaning on one external headline.
Our view runs against the usual instinct to benchmark against a national average. We'd skip the comparison entirely and chase consistency with your own baseline instead, since that's what actually tells you if an intervention moved anything.
What should you actually do with DSIT's numbers?
Use the AI adoption survey DSIT ran the way you'd use a weather report before a trip: useful context, not a verdict on your plans. It tells you where the UK economy sits on AI; it doesn't tell you where your marketing team or your night-shift warehouse crew sits.
Run your own survey, keep the questions stable, and track the trend quarter after quarter. That consistency tells you more than any national percentage ever will. Start there.
Frequently asked questions
How often should I check my company against a survey like DSIT's?
Treat it as occasional context, not a recurring benchmark. DSIT's fieldwork covers a single window, February to May 2025, so each wave is already a snapshot; we'd glance at a new wave when it's published rather than building a regular comparison ritual around it.
Does DSIT's survey count informal AI use that IT doesn't know about?
Not directly, since DSIT's interviewers spoke only with the person overseeing technology at each business. Employee-level surveys, like Gallup's self-reported study of US workers, ask individuals directly and can surface use that a tech lead never sees, so pairing a leadership-facing question with a workforce survey closes that gap.
Can I reuse DSIT's exact survey questions inside my own company?
You can borrow the spirit of asking about technology use and perceived impact, but DSIT's questions were built for a national, cross-sector sample, not a single organization. We'd adapt the structure rather than copy it, and keep the wording identical across your own survey rounds so results stay comparable over time.
Who should own tracking AI adoption inside a company?
In most organizations this falls to whoever owns the technology rollout and the people side of it together, often a Digital Transformation Leader working with L&D or People teams. The important part isn't the title, it's keeping one consistent owner and one consistent set of questions over time.
What if I search for an AI adoption survey South Africa Incasu ran?
We don't have verified details on a specific Incasu-run survey, so treat any result you find the way you'd treat DSIT's figures: check who answered, what question they were asked, and over what period before comparing it to your own numbers. The checklist matters more than the source's name.
Sources: AI Adoption Research - GOV.UK, AI in the Workplace: What Separates Adopters and Holdouts
