Two Government AI Studies. Two Opposing Conclusions. One Policy Agenda?
The UK government has published a new UK artificial intelligence sectoral analysis survey hot on the heels of the results from an earlier end-user focused AI study published 28 Jan 2026. Both come from DSIT. Both use reputable research partners. Both claim to inform the same AI Opportunities Action Plan. Both are on track to be incompatible sources of policy guidance.
The 2024 AI Sector Study, produced by Perspective Economics, Ipsos, Beauhurst, the AI Collaboration Centre and Glass.ai, tracks companies that build and sell AI. The most recent findings, from the 2024 wave published September 2025, identified 5,862 AI firms generating £23.9 billion in revenue and employing 86,139 people, a 68% revenue jump in a single year. These are the headline numbers behind billions in public spending on compute infrastructure, talent visas and AI Growth Zones. Around eighty percent of that revenue sits with 5% of firms. Three quarters are registered in London, the South East and East of England. The 2026 wave has just opened for participants.
The AI Adoption Research, produced by IFF Research and Technopolis from fieldwork conducted February to May 2025, surveys businesses trying to use AI. Published January 2026, it found that only 1 in 6 UK businesses (16%) currently use any AI technology. Eighty percent neither use nor have plans to. Seventy-one% cited not having identified a use as a barrier in their organisation. Among the minority who have adopted it, 77% report no change in revenue. Most cannot formally attribute productivity gains to AI. The study excluded businesses with fewer than 5 employees, over 5 million firms according to ONS Business Population Estimates.
| AI Sector Study (2024 wave) | AI Adoption Research (2025 fieldwork) | |
|---|---|---|
| Commissioned by | DSIT | DSIT |
| Conducted by | Perspective Economics, Ipsos, Glass.ai | IFF Research, Technopolis |
| Published | September 2025 | January 2026 |
| Who it measures | Companies that build/sell AI | Companies trying to use AI |
| Who it excludes | Businesses that only use AI | Businesses with fewer than 5 employees |
| Number of AI firms | 5,862 identified | Not measured |
| AI adoption rate | Not measured | 16% (1 in 6 businesses) |
| Revenue | £23.9bn (+68% in one year) | 77% of adopters report no revenue change |
| Employment | 86,139 in AI roles | 30% of staff use AI among adopters |
| Concentration | Around 80% of revenue from 5% of firms | Not measured |
| Identified use for AI | Not measured | 71% cited no identified use as a barrier to adoption |
| Geography | 75% in London, South East, East of England | London slightly higher adoption (20% vs 16%) |
| Growth narrative | 58% more AI companies year on year | 80% of businesses have no plans to use AI |
| Skills | 71% cite need for increased skills as barrier to growth | 60% cite limited AI skills as barrier to adoption |
| Biggest gap | Doesn't ask whether anyone is buying | Doesn't ask whether the sector is growing |
| Policy role | Justifies billions in AI investment | Suggests most of the economy can't use it yet |
We'd like your views:
- The sector study reports £23.9 billion in AI revenue concentrated in 4% of firms, mostly in London and the South East. The adoption study reports 80% of businesses have no AI plans. When both inform the same national strategy, which set of numbers should carry more weight in deciding where public money goes?
- The adoption research found 71% of businesses haven't identified a use for AI. The sector study found a 58% increase in AI companies in a single year. Is the UK building an AI supply industry faster than it's building demand and does that matter?
- The sector study uses AI-powered web crawling to identify which companies count as AI businesses, partly based on how firms describe themselves online. As more companies add AI language to their websites, does this method risk inflating the sector's apparent size?
- The adoption study excluded businesses with fewer than 5 employees. The sector study excluded businesses that only use AI rather than sell it. Between them, who is actually being measured and who falls through the gap?
