Where AI helps a small business, and where it does not.

5 November 2026

The short answer

AI is good at first drafts, summaries and sorting: turning notes into an email, pulling the key points out of a long document, tagging enquiries by type. It is poor at anything where a wrong answer costs you, such as prices, legal wording, medical or financial advice, or facts it has not been given. Use it where a person checks the result, and keep customer data out of tools you have not vetted.

How many businesses use it

The Office for National Statistics asks UK businesses about this every few months. In September 2026, 30% said they used at least one kind of AI, up from 9% three years earlier. Larger firms lead: around half of those with 100 or more staff, against 29% of those with fewer than ten. The most common uses were writing text and making images.

So if you are not using it yet, most small businesses are in the same place. There is time to do it carefully.

Where it earns its keep

First drafts. Replies to common enquiries, product descriptions, job adverts, the outline of a proposal. You still edit them, but you start from something rather than a blank page.

Reading for you. Summarising a long contract, a tender document or a pile of customer feedback into the points that matter, with references back to where each point came from so you can check.

Sorting and routing. Deciding whether an incoming email is a complaint, a quote request or an invoice, and putting it in the right place. Mistakes here are cheap because a person sees the result.

Searching your own information. Answering staff questions from your own handbook, price list or past jobs, when it is set up to answer only from those documents and to say when it does not know.

Where it causes trouble

Making things up. These tools write fluent, confident text whether or not it is true. Figures, quotes, laws and references need checking every time.

Talking to customers unsupervised. A tool that answers customers directly can promise a refund you do not offer or a price you do not charge. Your business is still responsible for what it says.

Personal data. Pasting customer details into a free public tool may mean sending them to another company, possibly abroad, to be stored or used for training. Under UK data protection law you need to know where personal data goes and why.

Replacing judgement. Decisions about people, such as hiring, credit or who gets a service, need a person who can explain them. Data protection law gives people rights when decisions about them are made automatically.

Trying it safely

• Start with one boring, frequent task where you can check every result.
• Use a business version of the tool with a clear promise not to train on your data, and read it.
• Keep names, addresses and anything sensitive out until you have checked where the data goes.
• Tell staff which tools are fine to use and for what. Many people are already using them quietly.
• Measure it: time saved per week, and how often the result needed fixing.

Where we come in

The useful work is usually joining AI to the information and systems you already have, so it answers from your documents and hands off to a person when it should. That is what our AI work covers, built so the data stays where you can account for it.

Where this comes from

• How many businesses use AI: Business insights and impact on the UK economy, 8 October 2026, Office for National Statistics.
• AI and personal data: Artificial intelligence guidance, Information Commissioner's Office.