What UK businesses are using AI for, and what that means for you

01

What does the 54% figure actually measure?

It is the share of respondents to a British Chambers of Commerce survey who said their organisation currently uses a specific AI technology. The survey had 668 responses in January and early February 2026, mostly from small firms. It is a sample of the BCC's network, not a census of UK businesses.

The BCC, with the University of Essex and partner Atos, added AI questions to its annual Business Outlook Survey. Of 668 answers, 54% said yes, 6% were not sure, 18% intended to use AI and 22% had no plans. The 2025, 2024 and 2023 surveys found 35%, 25% and 23%.

Of the respondents, 561 firms (84%) had 1 to 250 employees, 68 were sole traders and 39 had 250 or more. The authors report unweighted results because subgroups were too small to weight reliably, and say the findings describe the responding sample rather than population-representative estimates.

So the headline is better read as: a little over half of the mostly small firms that answered a BCC survey in early 2026 said they use AI. That signals how quickly use is spreading among engaged firms, not what every UK business does.

02

Why do UK surveys give such different AI adoption figures?

They ask different businesses, in different ways, at different times. A survey of a business network, a weighted statistical survey and a phone survey of technology leads will not match. Compare the base, the question and the date before comparing the numbers.

The ONS counts a business as a user if it reports using any technology on a list, and does not separate light use from embedded use. It excludes some industries, including finance and insurance. The DSIT survey would not capture staff using AI without the business knowing.

03

What does the AI adoption gap look like by size and sector?

Larger firms and professional, digital and creative sectors use AI far more than small, consumer-facing and manufacturing firms. In the BCC sample, 78% of IT and digital firms used AI against 31% in other services. The ONS and DSIT also find use rising with firm size.

• BCC sample: highest in IT and digital (78%) and creative and media (70%), lowest in other services (31%) and manufacturing (45%). Sole traders and micro firms were far more likely than larger firms to have no plans.
• ONS, September 2026: 29% of businesses with under 10 employees used AI, against 51% of those with 100 or more. Information and communication was highest at 60%.
• DSIT, 2025: 14% of micro, 23% of mid-sized and 36% of large businesses used AI. Construction, retail, transport and hospitality were the most likely to neither use nor plan to use it (86% to 90%).

These are patterns, not rules: a one-person digital firm can be a heavy user, and a manufacturer may reasonably see no use yet.

04

What are UK businesses using AI for, and how deep does it go?

Mostly general-purpose text tools for writing, summarising, research and administration. Bespoke systems are rare, and the ONS found only 10% of adopters with 10 or more employees reported extensive use. Occasional chatbot use and embedded AI are very different things.

DSIT found that 85% of businesses using AI used natural language processing and text generation, 35% computer vision, 21% machine learning and 7% agentic AI. Among those using or planning AI, the main uses were creative and content creation (77%), administration and support (70%), data and analytics (56%) and automation (47%).

The BCC says most firms use generic tools such as chatbots and writing assistants. About one in ten (65 firms) used bespoke systems, and were far more likely to report restructured job roles. The ONS calls adoption among businesses with 10 or more employees relatively shallow, at about 1.6 technologies per adopter.

Reported benefits are mostly self-reported. In the DSIT study, 75% of AI users reported improved workforce productivity, but 77% had not yet seen a change in revenue and 12% reported an increase. DSIT says these figures should be viewed as estimates.

05

What do the businesses that go deeper tend to do differently?

The evidence points to five habits: a defined task, trained staff, explicit permission and policies, human checking of outputs, and building AI into how work is done. These are associations in small samples, not proven causes.

The ISER authors stress that their results are associations, the sample is modest, and 65 bespoke adopters are too few for some findings to be more than indicative. Deeper is not automatically better: firms with bespoke systems and structured training were more likely to expect staffing reductions, which a small employer may not want.

06

Do I risk falling behind if I wait?

Possibly, but waiting can be reasonable. In the BCC sample 22% had no plans to use AI, and the DSIT study found 51% of businesses did not see AI as relevant. The risk is greatest where competitors use it for work you also do. Decide on evidence about your own tasks.

The most common reasons in the DSIT study were no identified need for AI (71%) and limited AI skills, expertise or knowledge (60%). Of those who cited them, 80% rated ethical concerns, 76% high costs and 72% unclear regulation as significant barriers.

The case for acting is mostly expectation. BCC respondents already using AI reported a net productivity expectation of +71 percentage points, against +46 for firms planning to adopt and -3 for firms with no plans. These are expectations, not measured results.

A practical test: list three weekly tasks that are text-heavy or repetitive. If you cannot name any, you may have no current need.

07

What is a realistic four-week plan to get started?

Start in week one by choosing one task and setting rules. Then pilot it with human review, measure net time saved and errors, and decide whether to keep, widen or stop. The aim is a decision backed by your own numbers, not a transformed business.

1. 01
   Week 1: choose one task

   Pick one frequent, low-risk, text-based task, such as routine enquiry replies or meeting summaries.

2. 02
   Week 1: set the rules

   Write one page covering approved tools, what must never be entered, and who checks output. Read the supplier's data terms.

3. 03
   Week 2: run a pilotFull step in the download ↓
4. 04
   Week 3: measureFull step in the download ↓
5. 05
   Week 4: decideFull step in the download ↓

Two checks belong in week one. The NCSC advises not to include sensitive information in queries to public AI tools, because providers store queries. The ICO's worked examples say UK GDPR still applies, and that an AI meeting-notes tool can keep existing lawful bases where the provider does not use the data for its own purposes and it is not used for new purposes or decisions about people.

08

When is outside help worth paying for?

When AI has to connect to your own systems or customer data, when many staff will use it, or when nobody has time to test it. For one person trying a chatbot on routine emails, the four-week plan is usually enough.

The BCC notes that smaller firms often struggle with the cost and complexity of AI, and DSIT found limited skills was the most common past barrier among businesses already using AI (60%).

Reeve Consult offers AI consultancy that starts with a specific workflow, the information it uses and who checks the output. It does not promise particular savings, and should say plainly when a task is unsuitable for AI.