Estimated reading time: 7 minutes
Ask a room of business owners what is wrong with the UK workforce and you will get a version of the same answer. People do not have the skills. It has been the settled view for a decade.
The government’s own survey says otherwise.
The Employer Skills Survey is the largest study of its kind in the country. The 2024 edition, published in November 2025, interviewed 22,712 employers across the UK. It found skills gaps in the existing workforce at their lowest recorded level since the survey began. Vacancies fell as well, from 23% of employers reporting one in 2022 to 17% in 2024.
So if the workforce is not unusually short of skills, what has actually gone wrong?
Employers stopped paying to build them.
What employers are actually spending
Total UK training expenditure in 2024 was £53.0 billion, down from £59.0 billion two years earlier. Per employee that comes to £1,700, against £1,960 in 2022. Adjusted for inflation it is 29.5% below the 2011 figure, and the Learning and Work Institute puts the fall at 36% since 2005, when the equivalent spend was £2,634.
Fifty-nine per cent of employers provided any training at all in the twelve months to 2024. In 2017 it was 66%.
These are the lowest figures the survey has ever recorded. Not the lowest since the pandemic. The lowest since the series started.
Worth sitting with for a moment: the apprenticeship levy was introduced in 2017 specifically to raise employer investment in training. Spend per employee has fallen by almost a quarter since, even counting the levy itself.
The skills crisis that is not quite a skills crisis
None of this means everything is fine. Skill-shortage vacancies remain higher than they were between 2011 and 2017, and construction and the skilled trades still struggle badly. Regional variation is stark, and the East Midlands records the joint lowest training days per employee of any English region at 3.0, against 4.3 in the North West.
But the framing matters, because it changes what you do about it.
If the problem is that qualified people cannot be found, the answer is recruitment, and you are competing with everybody else for the same short supply. If the problem is that your existing people were never trained on the thing you now need them to do, the answer is training, and you already employ the people who need it.
The data points firmly at the second.
The gap that is unambiguously real, and it is AI
Here is where the picture changes shape.
AI adoption among UK SMEs reached 54% in 2026, up from 35% a year earlier, according to the British Chambers of Commerce. Alongside that, the BCC and Helium42 found that 97% of UK organisations report at least one significant AI skills gap, and two in three employees have had no formal AI training whatsoever.
One caveat on that 54%, because it is the kind of number that gets quoted without its footnote. Adoption figures vary enormously depending on what the researcher counted. The ONS Business Insights survey put UK business AI use at around 25% in January 2026. The government’s own DSIT research, using a tighter definition of deliberate deployment for a defined purpose, puts it nearer 16%. The BCC figure of 54% covers SMEs and counts any use at all.
They do not contradict each other. They measure different things, and all three are rising steeply.
What none of them dispute is the second number. Whatever share of businesses are using AI, almost none of them have trained anybody to use it properly. The tools went in. The capability did not follow.
That is the actual state of play in 2026: the fastest technology adoption most owners have seen in their working lives, landing on a workforce receiving less training than at any point on record.
What changed in the funding
Worth knowing if you have not been following it. The apprenticeship levy became the Growth and Skills Levy in April 2026, with a wider set of things the money can be spent on, including shorter modular units rather than only full apprenticeships.
Two changes matter most to smaller firms. Non-levy-paying SMEs now get apprenticeship training fully funded for anyone under 25, removing the 5% employer contribution. And levy funds now expire after twelve months rather than twenty-four, so larger employers holding a pot have less time to decide what to do with it.
Level 7 funding has been restricted to younger starters, which has taken master’s-level apprenticeships off the table for most existing staff.
What training costs, and what it returns
This is the question everybody asks and the one with the least trustworthy answers online.
Search for training ROI and you will find impressive multiples. Look at where they come from and a pattern appears: most are published by companies that sell training, cite a study nobody links to, or describe an organisation that no longer exists. We removed several from our own site for exactly that reason.
So here is a more useful way to think about it.
The benchmark you can actually check is the ESS figure: £1,700 per employee per year, which is what the average UK employer spends. That is your comparison point. If you are spending nothing, you are below the level of an economy that is itself investing less than at any time on record.
Then work out what the gap costs you rather than what the training costs you. Pick one thing your team cannot currently do well. Estimate the hours it wastes each week, or the work it sends elsewhere, or the errors it produces. Put a number on that annually. Compare it to the cost of fixing it. If the fix is cheaper, the case is made, and you did not need anybody’s ROI multiple to make it.
Most of the time the arithmetic is not close. A capability gap that costs four hours a week across five people is roughly a thousand hours a year.
The reason this approach is better is that it survives being questioned. If your finance director asks where the number came from, you can show them.
What to do about it
Start with one gap, not a strategy. The businesses that get value from training almost always begin with something specific and irritating rather than with a development plan, because a specific problem has a measurable before and after and a development plan does not.
Train the people you already have. Recruitment is slower, dearer, and competing against everyone else facing the same shortage.
Measure something before you start. It does not have to be sophisticated. Time on a task, error rate, how long a quote takes to get out of the door. Without a baseline you will be guessing about whether it worked, and guessing is how training budgets get cut in the next downturn.
And if AI is the gap, treat it as a skills problem rather than a software problem. Buying the tool is the easy half, which is precisely why 97% of organisations have already done it and still cannot use it.
We write about this because it is what we do. If you want to talk about what training would actually look like in your business, our business training page sets out how it works, or you can just get in touch and tell us what is not working.
Sources
- Employer Skills Survey 2024, Department for Education and Skills England, published November 2025
- Falling short: understanding further falls in employer training, Learning and Work Institute
- Growth and Skills Levy reforms, CIPD
- AI Adoption Index 2026, drawing on ONS, DSIT and British Chambers of Commerce data
