UK business AI adoption nearly tripled, but extensive use remains rare

Reported artificial-intelligence use among UK businesses with 10 or more employees rose from about 12% in late 2023 to about 35% in June 2026. Yet adoption remained shallow: adopting firms used an average of about 1.6 AI technologies, up only 0.2 from 2023, and just 10% said they used AI extensively.

This study analyses public Office for National Statistics (ONS) findings alongside separate Department for Science, Innovation and Technology (DSIT) survey results. It asks whether the rise in headline adoption has translated into broad organisational use. It has not combined the surveys because their samples, questions and definitions differ. No Edworking customer, team or usage data was accessed.

• Among UK businesses with 10 or more employees, ONS reported use of at least one specified AI technology rising from around 12% in late 2023 to around 35% in June 2026—an increase of about 23 percentage points and nearly three times the starting rate.
• The average adopting business used about 1.6 AI technologies in June 2026, only 0.2 more than in late 2023.
• Only 10% of adopting businesses with 10 or more employees described their AI use as extensive, and 15% said more than half of employees used AI in daily work.
• ONS measured a 45-percentage-point industry gap: 58% of information and communication businesses reported AI use, compared with 13% in construction.
• In a separate DSIT survey of 3,500 businesses, estimated current adoption was 16%. Among adopters, the mean share of staff using AI was 38% in micro firms with 5 to 9 employees, 26% in small firms, 18% in medium firms and 20% in large firms. These estimates are not directly comparable with the ONS headline series.

AI adoption is often presented as a binary milestone: a business either uses AI or it does not. That framing misses how far a tool reaches across operations, how many types of technology are in use and how much of the workforce has adopted it.

The latest ONS synthesis offers a useful test. Its headline series shows rapid diffusion among businesses with at least 10 employees, but newer questions on intensity suggest that experimentation remains more common than deep integration. DSIT’s separate research points in the same broad direction while using a different methodology: adoption rates rise with business size, but the share of staff using AI among adopters is not highest in the largest firms.

The distinction matters for productivity, management and training. Buying or trialling one AI tool is not the same as redesigning a workflow, governing outputs or equipping most employees to use it safely.

• How quickly did reported AI use spread among UK businesses with 10 or more employees between late 2023 and June 2026?
• Did the breadth and organisational depth of AI use rise at the same pace?
• How large were the reported differences by industry and business size?
• What do public survey limitations allow us to conclude—and what remains uncertain?

ONS reported that the share of businesses with 10 or more employees using at least one specified AI technology rose from around 12% in late 2023 to around 35% in June 2026. The 23-point increase is substantial, but the measure records any reported use, not whether AI is embedded in core processes.

Over the same period, the average number of AI technologies used by adopters increased from about 1.4 to about 1.6. ONS explicitly treats that count as an imperfect proxy: not every technology is relevant to every business, and one use case can be either marginal or important.

Newer intensity questions sharpen the picture. Among adopting businesses with 10 or more employees, only 10% reported extensive use. Just 15% said more than half of employees used AI in daily work. Those questions were newly introduced, so ONS calls the results early indications.

Information and communication businesses reported the highest clearly quantified rate in the ONS summary, at 58%. Construction was at 13%. The 45-point gap is descriptive, not a causal estimate: industry mix, task suitability, capital needs, firm size and awareness can all affect reported adoption.

The sector chart is based on June 2026 Business Insights and Conditions Survey (BICS) microdata. Agriculture, oil and gas extraction, energy supply, public administration and defence, public education and health, and finance and insurance are outside BICS scope. Missing sectors are not zero.

DSIT’s AI Adoption Research surveyed 3,500 UK businesses and estimated that 16% currently used AI, with adoption increasing from 14% in micro businesses to 23% in mid-sized businesses and 36% in large businesses.

Among businesses already using AI, DSIT estimated that an average 38% of staff used it in micro firms with 5 to 9 employees, compared with 26% in small, 18% in medium and 20% in large firms. This does not mean micro businesses lead the economy in total AI use. It describes staff reach inside the subset of adopters, and a small denominator can make firm-wide adoption easier.

The ONS and DSIT headline adoption rates should not be subtracted or averaged. They were collected with different instruments, sample frames, definitions and fieldwork periods.

ONS found that 62% of businesses with 10 or more employees citing a lack of AI expertise as a barrier reported training or retraining existing staff, compared with 26% of businesses reporting no barriers. The 36-point difference is an association, not proof that the barrier caused training.

Across businesses with 10 or more employees, only 11% said more than half of their workforce had received AI-related training. This sits alongside the limited extensive-use rate: adoption can spread faster than governance, capability building and workflow redesign.

The public evidence points to a diffusion-depth gap. AI has moved from a minority experiment toward a common business tool, but breadth of technology, extensive organisational use and workforce coverage have moved more slowly.

Three interpretations are plausible. First, businesses may be in an experimentation phase and need time to redesign processes. Second, broad access to general-purpose tools can increase the adoption count without changing core production. Third, governance, skills and measurement can lag real use—especially when employees adopt tools informally.

The data do not establish a productivity effect. They also do not show that more AI is always better. Extensive use can introduce new risks around accuracy, privacy, security and decision accountability. A responsible implementation target is not maximum tool count; it is a documented, useful and governed workflow.

For managers, the useful audit is deeper than “do we use AI?”:

• inventory each AI-supported workflow and its business owner;
• record which employees actually use it and for what decisions;
• distinguish pilots, optional assistants and production dependencies;
• measure output quality, time saved, error rates and escalation needs;
• train staff on approved data, verification and accountability rules;
• retire experiments that do not produce a defensible benefit.

For vendors, consultants and policymakers, adoption rates should be paired with intensity, workforce reach and capability indicators. A one-tool trial and a governed production workflow should not be treated as equivalent outcomes.

UK business AI adoption rose quickly between 2023 and 2026, but organisational depth did not keep pace. The most important official finding is not simply that 35% of businesses with 10 or more employees reported using AI. It is that only 10% of adopters called their use extensive, the average adopter used about 1.6 technology types, and major industry and workforce-reach gaps persisted.

The next measurement challenge is to track durable workflow integration, training, governance and outcomes without confusing access with transformation.

This study transcribed published aggregate estimates from two public sources and reproduced only arithmetic comparisons supported by those estimates.

Office for National Statistics

The primary source is Artificial intelligence in UK businesses: 2023 to 2026, released 20 July 2026. It synthesises BICS Wave 159 fieldwork conducted 5 to 28 June 2026 and earlier waves. Wave 159 received 38,637 business responses, a 26.7% response rate. BICS is voluntary and its AI measures are official statistics in development.

The headline time series covers businesses with 10 or more employees and defines AI use as reporting at least one listed technology. Some size-band results include businesses with 0 to 9 employees. The article’s figures and notes identify when the population changes.

Calculations in this report are simple:

• adoption change: 35 minus 12 = 23 percentage points;
• relative multiple: 35 divided by 12 = 2.92, described as “nearly tripled” because both inputs are rounded;
• industry spread: 58 minus 13 = 45 percentage points;
• training-response spread: 62 minus 26 = 36 percentage points.

Department for Science, Innovation and Technology

The secondary source is AI Adoption Research, released 13 February 2026. It reports a survey of 3,500 UK businesses. Its definitions and size bands differ from BICS, so the estimates are shown in a separate figure and are not pooled with ONS data.

Reproducibility and missingness

Every chart row is stored in the study’s typed JSON and is downloadable as CSV, SVG or PNG from the page. Values marked unavailable would be represented as missing, never zero. Percentages are reported at the precision published by the source; no confidence intervals were supplied in the article-level summaries used here.

The estimates are self-reported and depend on respondents recognising the technologies they use. AI definitions are evolving, and embedded vendor features may be missed or interpreted differently.

ONS inputs described as “around” are rounded; derived differences and ratios inherit that uncertainty. The 10% extensive-use and workforce-share questions are new and should be treated as early indications.

BICS excludes several industries and some aggregate results exclude microbusinesses. Its voluntary response rate was 26.7%. Survey weighting reduces but cannot eliminate non-response and measurement error.

DSIT and ONS results are not directly comparable. This study does not infer causation, productivity gains, job losses or return on investment. It contains no private analytics and no Edworking customer or usage data. It has not been peer reviewed or independently human fact-checked.

• Office for National Statistics, Artificial intelligence in UK businesses: 2023 to 2026, released 20 July 2026.
• Office for National Statistics, Business Insights and Conditions Survey Wave 159 dataset, released 2 July 2026.
• Department for Science, Innovation and Technology, AI Adoption Research, released 13 February 2026.
• Office for National Statistics, Business Insights and Conditions Survey quality and methods information.

This study was produced for Edworking with AI-assisted research, writing and calculations using public sources. Edworking sells collaboration and AI-enabled workplace software and has a commercial interest in how businesses adopt workplace technology. No Edworking customer, team or usage data was accessed. The analysis does not evaluate Edworking’s AI adoption or product performance. It is editorial research, not a peer-reviewed paper; independent human fact-checking remains outstanding.

Edworking Team (2026). “UK business AI adoption nearly tripled, but extensive use remains rare.” Edworking Research. https://edworking.com/research/uk-business-ai-adoption-depth-2026