Summary

• AI adoption reached 32% of UK businesses in 2026, up from 16% in the previous survey.
• Half of adopters reported improved overall employee productivity, while 64% reported no immediate change in operating costs.
• Skills and knowledge barriers affected 67% of adopters and 78% of businesses planning to introduce AI.

Artificial intelligence adoption among British businesses has doubled within a year, reaching 32% according to new government research, although most organisations report relatively modest productivity improvements and limited immediate changes in operating costs.

The Department for Business, Innovation, Science and Trade published its second annual AI Adoption Survey on 29 October, comparing current usage with research conducted during 2025. The earlier survey found that 16% of businesses were using AI.

The latest study is based on interviews with 3,503 UK businesses conducted between 7 January and 21 April 2026. Researchers also completed 50 follow-up interviews and seven site visits to examine how the technology was being introduced into everyday business operations.

Adoption increased across the economy, but usage remained concentrated in activities that can be supported by relatively accessible software. More complex applications involving integrated business processes, specialist models, or autonomous decision-making were less widespread.

Knowledge-intensive industries lead adoption

Large businesses were more likely to be using AI than smaller organisations, with 52% reporting adoption. Sector differences were also substantial: usage reached 61% in information and communication, 45% in real estate, and 39% in business services and administration.

These industries contain substantial volumes of work involving documents, analysis, communication, and information processing, where employees can use AI features without necessarily changing the underlying business systems.

Among businesses using or planning to use AI, 77% identified approaches based on large language models. Administration was the most common application area, selected by 81%, followed by marketing at 71% and IT functions at 66%.

These figures relate to the group of businesses already using or intending to introduce AI, rather than the entire UK business population. The distinction is important when assessing how far particular technologies have spread into different activities.

Improving efficiency and productivity was the leading motivation, cited by 71% of businesses using or planning AI. Employee experience followed at 27%, while 25% identified competitive advantage.

Within organisations already using AI, 57% reported using it constantly and another 27% at least weekly. Many businesses have therefore moved beyond occasional experimentation, although frequent use of a tool does not necessarily mean it has become central to commercial operations.

Time savings have not consistently reduced costs

Among businesses adopting AI, 71% reported improvements in workforce productivity and 60% said they had developed more efficient or flexible processes. When asked specifically about employees’ overall productivity, 50% reported an increase, generally describing relatively modest gains.

These measures examine different aspects of performance. A business may recognise improvements in particular tasks without recording a comparable change in productivity across its entire workforce.

For example, software may reduce the time required to prepare routine documents, allowing employees to complete more work or spend additional time on activities requiring human judgment. Whether that results in financial savings depends on how the organisation reallocates the capacity.

Operating cost findings reflect that distinction. The survey found that 24% of adopters reported reductions, while 64% had seen no immediate change and 9% reported higher costs.

Implementing AI can also require expenditure on subscriptions, training, data preparation, systems integration, and review procedures. Those costs can offset some of the immediate savings generated by individual automated activities.

For businesses introducing the technology gradually, the first benefit may therefore appear as additional capacity rather than reduced expenditure. Determining whether that capacity produces lasting commercial value requires measurement across the process being changed.

Skills restrict more ambitious deployment

Capability-related barriers affected a substantial proportion of businesses. Limited skills, expertise, or understanding of AI applications were identified by 67% of adopters and 78% of organisations planning to introduce the technology.

These difficulties go beyond knowing how to use a conversational interface. Businesses may need staff capable of identifying suitable applications, preparing information for automated processing, evaluating outputs, and integrating new tools with existing systems.

Uncertainty over financial returns also affected investment decisions. Among non-adopters, 65% cited the absence of a clear business case or proven return on investment.

That concern can be particularly difficult for smaller businesses, where a technology project may require management time and technical expertise that are not available internally. The cost of purchasing software may be relatively modest compared with the effort required to incorporate it into established processes.

Concerns about accuracy and bias were reported by 56% of businesses already using AI and 60% of those planning adoption. Data protection, security, and uncertainty over regulatory obligations were additional considerations, particularly in sectors handling sensitive information.

Financial services, legal work, transport, and healthcare can involve requirements that limit the circumstances in which automated recommendations or actions are appropriate. Organisations may need stronger review and approval arrangements before allowing AI to affect consequential decisions.

Human oversight remains widely used

The survey found that 81% of businesses applied some human input or oversight to AI-generated outputs. The finding is consistent with the continued use of AI as a supporting technology rather than a complete substitute for professional judgment.

Trust varied according to the activity being performed. Drafting and summarising information were more widely accepted than systems operating autonomously or influencing high-stakes decisions.

That distinction affects how businesses design their processes. A generated document can be checked before distribution, while an agent authorised to complete transactions or change records may require controls that prevent an error from producing immediate operational consequences.

Businesses also need to consider the information available to the system and the reliability of its outputs. Where AI is connected to internal databases or customer records, errors can become harder to detect if automated results are incorporated directly into later stages of work.

The research suggests that many organisations recognise these limitations and continue relying on employee supervision. The appropriate level of oversight will vary with the application and the potential effects of an incorrect result.

Investment plans point towards gradual integration

Despite the implementation difficulties, 65% of surveyed businesses expected their AI-related budgets to increase. Planned investment was concentrated on activities that could extend existing use rather than requiring entirely new technical systems.

Among businesses using or planning AI, 61% expected to implement ready-made applications, 57% planned to automate tasks, and 54% intended to embed AI into existing systems.

More advanced development remained less common, although 24% of the same group planned to adopt machine learning. These intentions indicate a continuing market for established software products alongside more specialised applications.

Workforce adoption is also expected to expand. Currently, 22% of businesses using or planning AI reported that more than half their employees were using it. Looking one or two years ahead, 31% expected to reach that level.

These figures describe expectations rather than completed investment or future outcomes that can be guaranteed. Businesses may alter their plans as costs, technical capabilities, and operational requirements change.

The reported strategic importance of AI remains limited despite growing access. Only 14% of businesses considered it highly or critically important to their operations, suggesting that many organisations still use the technology for specific activities rather than as a foundation of their business model.

The latest survey establishes a substantial increase in adoption, supported by widespread access to language-based tools. Whether that growth produces deeper improvements in productivity will depend on the ability of businesses to address skills gaps, integrate software into established processes, and establish reliable evidence of commercial returns.