AI for small business: where the evidence supports practical adoption
A practical evidence guide for small businesses deciding where AI can save time, where the commercial case is still uncertain and how to buy tools without turning an experiment into a data or security problem.
Last reviewed 2026-07-25
How common is AI use in UK businesses?
AI use is growing, but adoption is not universal. The Office for National Statistics reported that about 25% of businesses were using some form of AI technology in late December 2025, rising to 44% among firms with at least 250 employees. That size gap matters for small businesses: the market is moving, but “everyone is already doing it” is still sales-pitch mathematics.
Where can AI help a small business first?
The best starting points are usually repetitive, text-heavy or information-heavy tasks where staff already know what a good result looks like: drafting routine communications, summarising internal material, extracting information, preparing marketing variants or assisting customer-service triage. Begin with one measurable pain point, keep sensitive data out until controls are clear, and compare the new process with the old one.
Does AI adoption automatically increase revenue?
No. UK AI Adoption Research found strong self-reported productivity effects among adopters, but revenue effects were much less common. In that survey, 77% of AI-using businesses reported no revenue change yet and 12% reported an increase. That is a useful commercial reality check: saving time can be valuable, but it does not automatically turn into sales or profit.
What should a small business ask a supplier?
Ask what data the product receives, whether customer material is used for model training, where data is stored, which subcontractors or models sit underneath the service, how access is controlled, what happens when the system is wrong and how you can export or delete your information. Then ask for the price after the trial period. Procurement is where shiny demos meet invoices and liability.
What does the current evidence say?
Limitations
Different surveys use different definitions, samples and time windows, so adoption figures should not be blended into one trend line. Self-reported productivity and revenue effects are not the same as independently measured causal impact.
A counterpoint worth keeping
The most profitable AI decision can be not to automate. If a task is rare, poorly defined or cheap to perform manually, integration and checking costs can exceed the saving.
Sources and provenance
- Office for National Statistics · Business insights and impact on the UK economy: 8 January 2026 · 2026-01-08 · Primary source
- Department for Science, Innovation and Technology · AI Adoption Research · 2026-02-13 · Primary source
- UK Government · Code of Practice for the Cyber Security of AI · 2025-01-31 · Primary source