How many UK businesses use AI? What the ONS figures mean for a small firm

If you run a small business, you have probably been told that "everyone is using AI now". The official figures say something more useful: plenty of firms are trying it, but most are using it in a narrow way. This post looks at what the Office for National Statistics (ONS) actually reports, what it means for a business with a handful of staff, and how to try AI this week without taking silly risks.

What the ONS figures say

The ONS asks businesses about AI in its Business Insights and Conditions Survey. In its September 2026 release, 30% of all businesses said they used at least one AI technology. That is broadly unchanged from June 2026 and up from 9% when the question was first asked in September 2023.

The size breakdown is worth a look:

• Businesses with 10 or more employees: 38% used AI, up 3 points from June 2026.
• Businesses with fewer than 10 employees: 29% used AI.
• Information and communication firms: around 60%, the highest of any industry.

So the gap between a micro business and a mid-sized one is real, but not enormous. If you are a ten-person firm that has not touched AI, you are in a large group, not an odd one out.

The ONS also says these are official statistics in development, so the figures should be used with caution. That matters because other surveys, often run by vendors or trade bodies, publish very different headline numbers depending on how they define "using AI". Treat any single percentage, including this one, as a rough guide rather than a target to hit.

What businesses are actually using it for

Among businesses with 10 or more employees, the ONS reports that text generation using large language models (the chatbot-style tools) was used by 21%, up 16 points since September 2023. Visual content creation was used by 19%, up 15 points. In other words, most use is about writing and images, not complex systems.

The same release asked adopters which roles were most affected. Some 38% said administrative and clerical roles, and 24% said creative or design roles. For a small business that is a useful hint about where the tools tend to help: the repetitive paperwork and first drafts, rather than the work that needs your judgement or a customer relationship.

On jobs, among businesses with 10 or more employees using AI, 6% reported reduced headcount and fewer than 1% reported increased headcount, both broadly stable over the past year. For most small firms, then, the practical story so far is changing how tasks get done rather than changing who does them.

A sensible way to start

Do not begin by buying a tool. Begin by writing down the tasks that eat your week. For most owners the list looks familiar: answering the same enquiries, drafting quotes and follow-up emails, writing product descriptions or social posts, tidying notes after a meeting.

Then pick one task and test it for a fortnight. A good first task has these features:

• It is repetitive and low stakes, so a mistake is cheap to catch.
• It does not need customer personal data, at least to begin with.
• A person on your team can read the output before anyone outside the business sees it.
• You can tell whether it saved time, for example by noting how long the task took before and after.

That last point is easy to skip. Measuring your own result is more reliable than trusting anyone else's.

Check the data side before you paste anything in

The moment you paste customer names, emails or order details into an AI tool, UK data protection law is in play. The Information Commissioner's Office (ICO) has guidance on AI and data protection, with sections on lawfulness, transparency, accuracy, fairness, individual rights, and security and data minimisation. The ICO notes that, because of changes made by the Data (Use and Access) Act, this guidance is under review and may change, so check the current page rather than relying on an older summary.

In plain terms, before you use a tool with personal data, ask:

• Do we need to put personal data in at all, or can we remove names and details first?
• Does our privacy notice still describe what we do with customers' information?
• Do we know where the tool's provider stores and uses what we enter?
• Who checks the output for mistakes before it reaches a customer?

The ICO also runs a section for small organisations, covering SMEs, start-ups, sole traders, small charities and clubs, with self-serve tools and short practical guidance. It is a good place to start if you do not have a data protection specialist. If you want a fuller walk-through, our earlier post on using AI tools with customer data covers the checks in more detail.

Do not forget the basics around it

AI tools are another account, another login and another place data can leak from. The National Cyber Security Centre's guide for small organisations says that many of its recommendations take as little as five minutes, need no technical expertise, and that cyber security should be spread across the team, because "every member of the team should realise that cyber security is everyone's business". Its related resources include guidance on password managers and passkeys. Before you add a new AI tool, make sure each person has their own login and that two-step verification is switched on.

Your plan for this week

• Monday: list the five most time-consuming repetitive tasks in your business.
• Tuesday: choose one that involves no customer personal data and write down how long it takes now.
• Wednesday: try a tool on it, with a named person responsible for checking the output.
• Thursday: read the ICO's small organisation guidance and note anything your privacy notice would need to say.
• Friday: compare the time taken, decide whether to keep it, and set up individual logins with two-step verification if you do.

When off-the-shelf stops being enough

Chatbot-style tools are fine for drafting. When the same job needs to connect to your booking system, your shop or your accounts, it usually makes sense to build something purpose-made. If that sounds like where you are heading, our AI solutions page explains how we approach it, and the process automation page covers joining up the repetitive steps that do not need AI at all.

The short version

Roughly three in ten UK businesses use AI, with the biggest uses being writing and images, and administrative roles the most affected. You are not behind, and you do not need to rush. Pick one low-risk task, measure the result, keep personal data out until you have checked the rules, and secure the accounts. That is a better plan than any headline percentage.