In short: Adoption figures make AI look mainstream, but depth tells a different story. Of the UK businesses that use AI, only around 16% use it extensively, and just 10% among firms with 10 or more employees. The rest are dabbling: a chatbot here, a licence there, on the edges of the business rather than at its core. That shallow use is why so few firms see a real return. The businesses that pull ahead will be the ones that move AI from the fringes into the systems they run on.
The headlines about AI in British business almost always focus on one number: how many firms use it. But what if that’s the wrong number to obsess over? Using AI and using it seriously are actually two very different things.
According to the Office for National Statistics, only 16% of the UK businesses that have adopted AI use it extensively. Among businesses with 10 or more employees, the ONS puts the figure at 10%. Either way, the vast majority use it on a limited basis, if at all.
That’s the real state of AI in the UK: a mile wide but an inch deep. Adoption has spread quickly, but depth has barely moved. This finding can (and should) reshape how businesses think about AI, and it’s a central theme of our study of UK AI adoption statistics.
Adoption Is Wide, but Use Is Shallow
The ONS data is blunt about it. Among businesses that use AI, the majority describe that use as limited rather than extensive. Only a small minority have woven it into how they actually operate.
The same survey shows the average adopter uses barely more AI than it did in 2023. The typical number of AI technologies per business has crept from around 1.4 to just 1.6 in almost three years.
Wide but Not Deep: UK AI Adoption and Depth, 2023 to 2026
| Survey wave | Businesses using at least one AI technology | Average AI technologies per adopting business |
|---|---|---|
| September 2023 | 11.9% | 1.42 |
| September 2024 | 17.8% | 1.50 |
| September 2025 | 27.2% | 1.56 |
| June 2026 | 34.9% | 1.59 |
| Change since 2023 | Almost tripled | +0.17 technologies |
Businesses with 10 or more employees. Source: Office for National Statistics, Artificial intelligence in UK businesses: 2023 to 2026 (Business Insights and Conditions Survey). Contains public sector information licensed under the Open Government Licence v3.0.
In other words, most firms that have adopted AI did it once, in one place, then stopped. Put simply, they aren’t building on it.
This is what shallow adoption looks like in practice:
- A marketing assistant drafting the odd email
- Someone summarising a document
- A single licence used by a single team
Sure, it’s useful. But it’s nowhere near the core of the business, and nowhere near enough to move the needle.
How Extensively AI-Using Businesses Use It
| Business size | Used extensively | Used on a limited basis | Testing or pilots only | Not currently used | Not sure |
|---|
Percentage of businesses that report using at least one AI technology, or are not sure whether they do. Rows may not sum to 100 because of rounding. Not published means the ONS suppressed the cell because the sample was too small. Source: Office for National Statistics, Business Insights and Conditions Survey, Wave 159 (15 to 28 June 2026). Contains public sector information licensed under the Open Government Licence v3.0.
The way businesses use AI confirms this. Close to 60% of adopters use it simply to improve existing operations. Fewer than one in four use it to develop new products, and fewer than one in five to reach new markets. These are the applications with real room to grow.
What AI-Using Businesses Use It For
| Business size | Improve existing operations | Provide or personalise products or services | Develop a new product or service | Explore a new market |
|---|
Percentage of businesses that report using at least one AI technology, or are not sure whether they do. Businesses can select more than one purpose, so rows do not sum to 100. “Other”, “not sure” and “not applicable” are excluded. Source: Office for National Statistics, Business Insights and Conditions Survey, Wave 159 (15 to 28 June 2026). Contains public sector information licensed under the Open Government Licence v3.0.
The pattern is clear: AI mostly shaves a few minutes off tasks people already did. It isn’t changing what a business is capable of. Independent coverage of the same figures reached the same verdict: UK firms are adopting AI widely but using it shallowly.
Why Shallow AI Rarely Pays Off
Shallow use is an obvious missed opportunity. It also actively undermines the case for AI. When a business dabbles, it gets dabbling-sized results. Think a bit of time saved here and there, but nothing that shows up in the accounts.
Then, when leadership asks what the return has been, the honest answer is “not much”. The appetite to go further evaporates. And the tool gets blamed, when the real problem was the shallowness of the deployment.
There’s a governance cost too. When AI sits on the fringes as a scattering of personal tools, nobody really owns it. Staff use whatever they’ve found, usually without oversight, which raises real risks around data, accuracy, and compliance.
We’ve written about why most teams aren’t using Copilot properly, and the pattern is almost always the same. The technology is capable of far more than it’s being asked to do.
The businesses seeing genuine returns are the ones that have done the opposite of dabbling. They’ve picked a core process and embedded AI into it properly, then built from there. In short, depth beats breadth every time. A firm that goes deep on two or three uses will comfortably out-perform one that subscribes to ten tools and masters none.
From the Fringes to the Core
So how does a business move from shallow to serious? Not by buying more tools. The firms stuck at the surface usually have plenty of AI logins already. The shift that matters is moving AI out of individual apps and personal habits and into the systems the whole business runs on.
That’s the difference an enterprise resource planning (ERP) system makes. When AI is built into the platform that already handles your finance, sales, and stock, it stops being a novelty. It becomes part of the operational fabric, used every day by design, because it lives in the tools people already work in. Nobody has to remember to open a separate app.
This is exactly how Copilot in Business Central works. Rather than being a chatbot bolted onto the side, the AI lives inside the system. It handles routine work in the flow of everyday operations, drafting and matching invoices, and answering plain-English questions about your own data.
There are loads of ways it reshapes day-to-day operations for small and medium-sized businesses (SMEs). All of them share the same trait: the AI is doing real work at the core instead of tinkering at the edge.
The 16% figure is a warning and an opportunity in one. It’s a warning that most AI adoption so far has been too shallow to matter. And it’s an opportunity, because the businesses treating AI as core infrastructure rather than a collection of gadgets are still a small minority. There’s plenty of room to be one of them.
Go Deep, Not Wide
If your AI use has stalled at the surface, it’s time to stop the new subscriptions and start putting AI where the real work happens. Let us show you how Business Central builds AI into the systems that run your finance, stock, and operations. Speak to the team about making AI part of how you actually work.
Frequently Asked Questions
What Share of UK Businesses Use AI Extensively?
According to the ONS, of the businesses that use AI, only around 16% describe that use as extensive. Among businesses with 10 or more employees, the figure is 10%. The majority use it on a limited basis, which is why so few firms see a meaningful return from it.
What Does Shallow AI Adoption Mean?
It means AI is being used lightly and on the edges of a business. So, a single tool gets used by a single team, rather than embedded into core operations. The ONS finds the average adopter uses barely more AI than it did in 2023, rising only from about 1.4 to 1.6 technologies.
Why Does Shallow AI Use Matter?
Because dabbling produces dabbling-sized results. Light, fringe use rarely shows up in the numbers, so businesses conclude AI does not pay off and stop investing. It also creates governance risk when unmanaged personal tools handle company data.
How Do Businesses Get a Real Return from AI?
By going deep rather than wide. The firms seeing genuine returns embed AI into a core process and build from there, rather than subscribing to lots of tools and mastering none. Depth consistently beats breadth.
How Does an ERP System Help?
An ERP puts AI at the core of the business. With Business Central and Copilot, the AI is built into the finance, sales, and stock system your team already uses. It does real work every day instead of sitting unused on the fringes.
Where Does This Data Come From?
The figures are from the ONS Business Insights and Conditions Survey, Wave 159, June 2026, and the ONS article “Artificial intelligence in UK businesses: 2023 to 2026”. You can read our full analysis in our UK AI adoption statistics study.
