Best AI for Business in 2026: What UK Firms Use

Which AI tools are UK businesses using in 2026?

ChatGPT has a substantial lead in LOCALiQ’s marketing research. Among respondents using AI in their marketing, 83% used ChatGPT, compared with 29% using Google Bard/Gemini and 27% using Bing AI/Copilot. The full survey covered more than 500 UK businesses. [2]

That is a strong result for ChatGPT. It does not mean 83% of all British businesses use it. Nor does it tell us which subscription they bought, how often they opened it or whether the work was any good.

The survey also includes video, image and automation products. CapCut, for example, reached 37%. It is left out of the assistant comparison above because putting a video editor beside a general-purpose chatbot answers a rather different buying question. [2]

Are ChatGPT, Gemini and Copilot actually LLMs?

An LLM, or large language model, is the underlying technology that works with language. ChatGPT, Gemini and Copilot are the names people encounter as products or assistants, with models and other tools behind them. Buying an assistant also means choosing its features, access to your files, administration and terms.

That distinction matters here. The ONS counts types of AI technology; LOCALiQ asks about named tools. Neither gives us a complete league table of individual models used across British businesses.

Where does Claude fit?

Claude is absent from the named figures used in this comparison. That leaves us without a comparable percentage; it does not establish that British businesses hardly use it. If it is on your shortlist, put it through the same work sample as the others. A missing survey entry is a poor reason to rule out a tool.

How many UK businesses use AI?

The ONS puts adoption at around 35% of UK businesses with ten or more employees in June 2026, up from about 12% in late 2023. BCC and FSB research reports higher levels in their own samples. [1] [3] [4]

These surveys agree that use has grown. They ask different businesses different questions, so their percentages are not interchangeable. The FSB’s latest figure shown here comes from November 2025 fieldwork, even though it appears in research coverage published in 2026.

Do not average the three results and call it the UK adoption rate. The arithmetic would be easy. Explaining what the answer measured would be rather harder.

Which kinds of AI are businesses using?

Language models lead the ONS technology categories at 18%, followed by visual content creation at 16%. These figures use businesses with ten or more employees as their base, including businesses that do not use AI. [1]

An adopting business uses an average of about 1.6 AI technology categories. That figure cannot tell us how many chatbots it has bought. Three different writing assistants might still sit in one category. It also cannot tell us whether one well-used tool is doing a better job than five neglected ones. [1]

Do small or large businesses use AI more?

Larger businesses are more likely to have adopted AI. They have budgets, IT departments and enough middle managers to commission a PowerPoint about commissioning a PowerPoint. The ONS figures show a clear size gap. [1]

Other studies point the same way, although their populations differ. DSIT reports 51% adoption among small businesses handling digitised data and 82% among large businesses in that population. NatWest’s SME and mid-market survey reports 67% for firms with more than 100 employees, compared with 36% for smaller firms. [5] [6]

Using AI extensively is a different question

Among businesses already using AI, the smallest firms are more likely to describe that use as extensive: 17%, against 9% of the largest firms. [1]

It is tempting to read that as “big companies buy the licences; small companies use the bloody things”. Tempting, but too neat. A five-person business and a company with thousands of staff have very different tasks when they try to spread a new habit.

The ONS also found more uncertainty among large businesses about how extensively AI was being used. These are early findings from new questions. They are a reason to investigate how work changes inside a business, not proof that small firms are universally better at AI. [1]

Which UK industries use AI most?

In the ONS research, 58% of information and communication businesses report AI use, compared with 13% in construction. The difference makes sense: a business that spends much of its day working with documents, code and data has an obvious place to start. [1]

That does not make a construction firm a lost cause. Tenders, supplier comparisons and site reports still need doing. The useful comparison is whether AI improves one of those jobs, rather than whether the firm looks sufficiently modern beside a software company.

These DSIT figures have a different base from the ONS estimates. Keep them within their own comparison.

The sources reviewed here do not provide a reliable national table of ChatGPT, Gemini, Claude and Copilot use by both industry and employee band. I would not turn ChatGPT’s lead in a marketing survey into a claim that it leads every sector. We do not have that evidence.

What are businesses actually using AI for?

In LOCALiQ’s marketing sample, writing dominates: 90% use AI to produce written content. Images and video follow at 39%; customer service sits at 20%. [2]

There is nothing inherently trivial about a product description, proposal or customer email. Doing those jobs well matters. The problem starts when success is measured by the volume of text coming out of the machine.

A business that publishes twice as many forgettable blog posts has doubled its output. Whether it has improved anything is another question.

Take a recurring proposal. The useful job includes gathering the client’s requirements, checking what you can deliver, writing the offer, confirming the price and getting it approved. AI might help at several points. Saving ten minutes on the opening paragraph while leaving the rest untouched is a modest improvement, and should be reported as one.

What is the best AI for your business?

My starting point would be the work, followed by the software you already use and the information involved. Here is how I would narrow the shortlist.

ChatGPT is a sensible starting point, not an automatic winner

Its lead in the marketing survey makes it an obvious tool to include. If staff already use it competently, that familiarity is useful too. I would still ask them to show me the work before buying more seats.

Check what “included” actually includes

Microsoft distinguishes Copilot Chat, available with qualifying Microsoft 365 subscriptions, from capabilities requiring an add-on licence. Google includes Gemini features in Workspace Business and Enterprise editions, with availability depending on the edition and settings. [7] [8]

So check the actual account, licence and task. “We already pay Microsoft” is the beginning of a purchasing conversation, not the end of it. The integration may be valuable; it is not proof of a saving until somebody uses it.

When should a small business pay for AI?

A free tool can be enough to test a straightforward task using material you are permitted to put into it. Pay when limits, administration, data requirements or missing features are getting in the way of useful, repeated work. You may need a business plan from the outset; waiting for a usage limit is not the only reason to buy one.

Price the finished job. Include briefing, checking, corrections, training and the subscription. A cheap assistant that creates an expensive mess is poor value.

How to choose an AI tool without wasting a month

You do not need a committee to begin. You need a recurring task, a person who understands it and a way to judge the result.

Run a short trial on real work

1. Choose one job and keep the brief consistent

Try a weekly customer update, a proposal outline or a summary of approved meeting notes. Give each tool the same source material and instructions. Include an awkward example; the easy ones rarely tell you enough.

2. Agree what a good result looks like

Write down the facts that must be right, the details that must be included and who checks them. Ask a colleague to assess the output without being told which tool produced it.

3. Count the whole job

Measure from preparation to approved output. Keep a note of errors, corrections and anything that could have caused a problem if it had slipped through.

4. Teach the method that worked

Write down the brief, the permitted inputs and the review steps. Let a second person try it. If they cannot reproduce the result, you have a useful experiment to improve before you roll it out.

The ONS identifies difficulty finding use cases, cost and lack of expertise as barriers. Among businesses with ten or more employees citing an expertise gap, around 62% report training or retraining staff to integrate AI skills. [1]

That is where I would put the effort: helping people do recognisable work properly. Another subscription is easy to arrange. Changing what happens on a Tuesday morning takes a little more attention.

Choose the tool that makes the work better. Keep the evidence. And if your AI strategy still amounts to “we have ChatGPT”, it is probably time to write the second sentence.