UK private-sector businesses with at least five employees doubled their reported use of artificial intelligence between successive surveys, with 32% using at least one listed technology in 2026, up from 16%. The measure includes businesses piloting AI, rather than only those using it routinely.
Published on 29 October 2026, the report is the second instalment of a repeat survey series. It used 3,503 telephone interviews conducted between 7 January and 21 April 2026, with respondents screened for responsibility for, or oversight of, technology. Results were weighted by business size and sector. Education, human health and social work, and public administration and defence were among the excluded sectors.
A further 6% said they were not currently using AI but planned to adopt it, compared with 5% in 2025. The proportion saying AI was not relevant to their organisation fell from 51% to 38%.
London’s reported adoption rate was 33%, compared with 20% in 2025. However, the report found no significant regional differences in adoption in 2026. The figures are part of a national survey and do not establish a London-specific trend.
Reported use increased with employer size, from 27% among businesses with five to nine employees to 34% among those with 10 to 49. Adoption reached 47% for firms with 50 to 249 employees and 52% for those with 250 or more.
Information and communication recorded the highest adoption rate at 61%, up from 43% in 2025. Business services and administration reached 39%, against 23% previously. Lower rates were recorded in hotel and catering at 23%, transport and storage at 24%, and construction and retail and distribution, both at 25%.
Natural-language processing and text generation were the most commonly used technologies, reported by 30% of eligible businesses, compared with 13% in 2025. Computer vision or image processing and generation was reported by 9%, while agentic AI stood at 2%.
Usage and reported effects
Among businesses using or planning to use AI, administration was the most commonly named area at 81%, followed by marketing at 71% and IT at 66%. These figures cover both current and planned use. Among current users, 57% reported constant organisational use and 27% use at least once a week; these are not measures of individual staff use.
Improving efficiency and productivity was the main reason for adoption, selected by 71% of businesses using or planning to use AI. Improving employee experience was selected by 27%, while 25% cited competitive advantage.
Half of adopters reported an overall increase in employees’ productivity after adopting AI, while 42% reported no change and 4% a decrease. These are respondents’ assessments, not measured causal estimates, and the survey does not establish that AI alone produced the outcomes.
Operating-cost effects were mixed: 24% of adopters reported reductions, including 2% reporting a significant reduction. Most, at 64%, reported no effect, while 9% said costs had increased.
Only 14% of adopters rated their AI tools as highly important to operations on the report’s 0-to-10 scale. A further 42% rated their importance as medium and 44% as low. The report describes gains as mostly incremental.
Skills and business case barriers
Among current users, 67% cited limited AI skills, expertise or knowledge as a barrier, and 64% cited limited knowledge of how AI could help their business. Concerns about bias or lack of accuracy were cited by 56%.
For businesses planning to adopt AI, the corresponding figures were 78%, 67% and 60%. Among those neither using nor planning to use it, 65% cited no clear business case or proven return on investment.
Human oversight remained common: 81% of current users reported at least some human input or checking of AI outputs or decisions, while 5% reported none. The report describes oversight as particularly important for higher-risk uses.
Among respondents able to estimate AI spending, 47% reported no dedicated AI-related expenditure in 2025; among current users, the figure was 43%. The report links this partly to free tools and AI features embedded in existing software, rather than concluding that use was cost-free.
