ONS data shows about 35% of UK businesses now use AI, yet use stays shallow. Four questions that move leadership teams from AI access to measurable value.
Link copied
Around 35% of UK businesses with 10 or more employees now use at least one AI technology. In autumn 2023 the figure was around 12%.
That is the headline from the Office for National Statistics, published on 20 July 2026. The less quoted line sits further down the same article: "AI adoption to date has been relatively shallow."
Access has spread fast. The question on the leadership table has moved with it: the tools are in place, so what has changed in the work?
Moving from AI access to AI value means moving from a stage where people have the tools to a stage where those tools measurably change the outcome of a business process. Licence counts measure the first stage. Time, error or cost in a chosen process measures the second.
The ONS figures come from the Business Insights and Conditions Survey (wave 159, June 2026, 38,637 responses). They show breadth growing while depth barely moves.
| Indicator (businesses with 10+ employees) | Figure |
|---|---|
Use at least one AI technology | Around 35% (around 12% in 2023) |
Average AI technologies per adopting business | Around 1.6 (around 1.4 before) |
Adopters reporting extensive AI use | 10% |
Adopters where more than half of staff use AI daily | 15% |
Businesses where more than half of the workforce has had AI training | 11% |
Large organisations tell a different story about themselves. In Publicis Sapient's 2026 Global Enterprise AI Report, which surveyed 1,550 AI decision-makers at organisations with at least 500 employees and $100 million in revenue, 60% of UK respondents said AI is highly or fully embedded in their workflows.
Across all six countries in the same study, 73% use AI regularly or across most processes. Only 10% call it core to how the business operates. And 42% say AI is capable, but their organisation isn't set up to capture its value.
So the tools work. What lags is the way the organisation around them works.
To see value, count redesigned workflows.
Kyndryl's People Readiness Report (June 2026, 1,100 senior leaders in eight countries) singles out a group it calls Pacesetters: the 9% of organisations that redesign roles, manage change and build workforce readiness. They are 1.5 times more likely to report AI-related revenue growth.
In practice, start with one process. Proposal writing, customer query handling or the monthly management report all work. Write down today's steps. Mark the step where AI comes in and the step it removes.
If AI was added and no step went away, the process hasn't changed. A tool has been added to it.
According to Kyndryl, only 33% of organisations have clear policies on which decisions AI can and can't make. Meanwhile 81% of the same leaders expect AI agents to make impactful decisions within a year, and 25% completely trust AI systems running without human oversight.
That gap slows people down. They use the tool, then hesitate over whether the output counts.
A three-line decision table per priority process is enough to start: the step where AI drafts, the step where a person checks, and who gives final approval. We cover the same split for agents in our first AI agent decision matrix.
Once AI costs become visible, the conversation changes. Bloomberg reported on 1 June 2026 that Walmart had introduced a per-employee usage allowance for Code Puppy, its in-house AI agent, after a period of unlimited use. A company spokesperson said Walmart wants employees to use AI to create value and to choose the right tool for each task.
The lesson for UK leadership teams: the link between access and work needs managing too.
For 2027 budgets, we suggest tying the AI line to two or three priority processes. A year from now, the spend can then be judged against those same processes.
Middle managers often decide whether value reaches the work. A manager who can say "this is how we do this task now" turns a tool into a habit.
The ONS data points the same way. Among businesses that cite a lack of AI expertise, 62% train or retrain existing staff, against 26% of businesses that report no barriers. Kyndryl finds that a third of organisations have fully implemented training for working alongside AI tools.
We look at the manager's role in more depth in what leadership means when AI does the shallow work. A quick test: ask your managers to list three tasks their team now does with AI. How specific the answers are tells you where the programme stands.
Access indicators are useful at the start. If they are the only ones in the board pack, progress looks bigger than it is.
| Access indicator | Value indicator |
|---|---|
Active users or licences | Workflows redesigned |
Staff trained | Change in time, error or cost in a chosen process |
Number of tools in use | Share of processes with a defined decision table |
Sessions or prompts | Lasting use cases per manager's team |
For the cost and return side, see seven metrics that measure AI ROI. For the training sequence, see our six-step AI literacy programme.
Zack Kass, who led go-to-market at OpenAI, talks about what organisations can expect from AI in this keynote:
Paris 2023: Zack Kass, AI Futurist | International Bar Association (YouTube)
For leadership teams that want to work through these questions together, Zack Kass, Cassie Kozyrkov and Andrew Grill speak on AI strategy and how it changes the way a business runs.
It means moving from a stage where staff have AI tools to a stage where those tools measurably change the outcome of a business process. Licence and user counts show access. Change in time, error or cost in a chosen process shows value.
According to the ONS, around 35% of UK businesses with 10 or more employees used at least one AI technology in June 2026, up from around 12% in autumn 2023. Use is more common in larger firms: 49% of businesses with 250 or more employees reported AI use.
The evidence points mainly to how organisations work. In Publicis Sapient's 2026 survey of 1,550 AI decision-makers, 73% use AI regularly but only 10% call it core to their business, and 42% say their organisation isn't set up to capture its value.
Access indicators such as active users are fine at the start. For value, track the number of redesigned workflows, the change in time, error or cost in chosen processes, and the share of processes with clear rules on what AI decides.
Managers should make clear which tasks their team does with AI, define where AI output can be used in decisions and follow up until the new way of working sticks. The ONS found that only 11% of UK businesses have trained more than half of their workforce on AI.
After Dinner Speakers
Contents
