UK AI Training Statistics 2026: Adoption, Skills Gaps and Workplace Evidence

Which UK AI training statistics are useful in 2026?

Give a percentage a sufficiently large font and it can acquire the authority of a judge. Put it beside a photograph of a worried executive and you've practically written the sales pitch.

Somewhere beneath it, in type usually reserved for the ingredients of indigestion tablets, you may discover who was actually surveyed. Perhaps the research is two years old. Perhaps the respondents were large businesses and you're running a company of six. Perhaps the question had nothing to do with training at all.

I wouldn't sign a training budget on the strength of the headline. The research can be valuable, provided we read what it says. Here are the figures, checked against their original publications on 2 October 2026. The year in this article's title refers to this edition; the table makes clear when each survey took place.

Survey detail: the Ipsos employer study sampled 801 employers, excluding sole traders and the public sector; fieldwork ran 19 March–7 June 2024. The Reuters study sampled 1,004 journalists, 29 August–4 November 2024. The ONS business measures above use BICS and the stated employee-size threshold. Consult the original reports before generalising beyond those populations.

Why do AI statistics give different answers?

“Our business uses AI” could mean a single department has bought a tool. “I use AI at work” is an individual's answer. Neither tells you whether anybody has been taught to use it properly.

Training is a fairly roomy word itself. It can describe an introductory video, a practical workshop or months of study. All may teach something. Counting people who attended won't tell you whether they can now complete the job without help.

Before putting two percentages on the same chart, check the following:

• Who answered? Employees, employers and businesses are different groups.
• Who was excluded? Very small firms, public bodies or particular sectors may fall outside the study.
• What was asked? Using a tool, feeling confident and completing a task are different things.
• When were they asked? A new publication can report much older fieldwork.
• How was the study run? Look at recruitment, sample size, weighting and question wording.

Keep those details close to the number. Your reader shouldn't need a magnifying glass and a spare afternoon to work out what you've claimed.

What does the evidence say about the AI training gap?

The surveys give us reasons to investigate how people are learning and using AI. They can't tell you whether your finance manager checks a calculation or your sales team knows what information it's allowed to upload. You'll have to ask them.

My view is that buying the software, teaching people to use it and deciding where it's appropriate all need somebody's attention. Call the whole lot an “AI rollout” and it's easy to celebrate the licences arriving while the other jobs languish on a list.

I'd look closely at how the team handles an actual piece of work. Enthusiasm can coexist quite happily with carelessness. Equally, somebody who's reluctant to touch AI may simply need a patient explanation and a chance to try it without an audience.

The free AI skills matrix gives you a way to record what people can do and where they need help. Use the national research to inform the conversation, then make the training decision with the people whose work you're trying to improve.

Which AI training measures should a business owner track?

Pick a recurring task and a small group. Keep a record of how the job goes before training, then review comparable work afterwards. Resist the temptation to commission a dashboard so elaborate it requires its own induction course.

Record other things that might explain a change. Perhaps the second brief was easier, somebody lent a hand or the person simply got better through practice. Claiming the entire improvement for the course would be convenient for a training provider. It would also be a bit cheeky.

Be equally careful when turning time into money. If somebody finishes an hour earlier and uses that hour for other work, you've gained capacity. A cash saving requires an actual cost to change. Finance will probably notice the distinction, particularly if your spreadsheet has become a little overexcited.

If the team needs a grounding in everyday use, look at AI Basics Deep Dive. Our advanced training guide covers what comes after that. You can find all the available modules on the programmes page.

What should journalists check before quoting these statistics?

Link to the original research and retain its population, fieldwork period and question. Attribute the finding to the organisation that produced it. We've collected and commented on these figures; we didn't conduct the surveys.

And yes, I sell AI training. You should be as willing to question my interpretation as you would anybody else's. Ask how respondents were recruited, what they were asked and which groups are missing before accepting a dramatic conclusion from a provider's research.

I'd be interested in a story that followed people back to work after a course. What can they do now? What still goes wrong? Who helps when they get stuck? The answers would tell us considerably more than another photograph of delighted delegates holding certificates.

For commentary on teaching AI to business teams, contact The Oxford AI School. For questions about how a particular survey was conducted, go to the researchers who did it.

Sources and editorial method

These are selected primary sources, checked on 2 October 2026. The table separates training, adoption, self-reported knowledge and policy targets. The advice about choosing and measuring training is my interpretation.

1. Ipsos: AI Skills for Life and Work: employer survey findings. Published on GOV.UK, 28 January 2026.
2. ONS: Artificial intelligence in UK businesses, 2023 to 2026. Published 20 July 2026.
3. Thurman, Thäsler-Kordonouri and Fletcher: AI adoption by UK journalists and their newsrooms. Reuters Institute, 27 November 2025.
4. UK Government: AI Skills Boost explainer. Published 28 January 2026.

Check for a newer release before reusing a figure in a dated report. If you spot an error here, send us the figure and its source so we can review it.