Is AI eating entry-level jobs? What the UK figures say

Reed has been making the case since the summer. In a BBC interview reported by Personnel Today in June 2026, he said AI was “burning through entry-level jobs” and suggested a surcharge on businesses that replace people with machines, adding that it was “going to take some designing”.

In an essay in July he set out the figures behind it. Graduate vacancies on Reed's own job site, he wrote, had fallen from about 180,000 to 50,000 in four years, and UK vacancies overall from over a million to around 700,000. He called the rise in employer National Insurance a historic mistake and floated two ways to tax AI instead: a surcharge on replacing staff, or a levy on AI services that works rather like VAT.

In October a Reed survey, reported by This is Money, found 52% of UK hiring managers believe AI is reducing job opportunities, and 36% say it is already lowering recruitment at their own organisation.

Vacancies are falling. The Office for National Statistics put UK vacancies at 691,000 for July to September 2026, the first time below 700,000 since early 2021.

Graduate hiring is not collapsing, though. The Institute of Student Employers' 2026 survey, reported by the Guardian, found graduate recruitment rose 4% this year after an 8% fall in 2025, with employers expecting a further 7% rise. Its chief executive, Stephen Isherwood, said employers were “not managing down graduate recruitment because of AI”, but because of economic uncertainty.

Cost is part of the story. The British Chambers of Commerce, citing research by the National Institute of Economic and Social Research, says rises in employer National Insurance and the minimum wage, along with employment rights reforms, have pushed up the cost of hiring at entry level by around 7% in real terms.

And employers do expect AI to change the work. In the ISE's spring survey of 144 employers, 87% said they expect AI adoption to reshape graduate and apprentice roles.

So both things are true. Hiring is weak for several reasons, of which AI is one and not yet the largest. And the jobs that remain are changing shape.

When we build AI into a business, it rarely takes a whole role. It takes tasks: sorting and answering routine enquiries, first drafts of letters and reports, moving information between systems, summarising long documents. Those are often the tasks a junior was given while they learned.

It is also still wrong often enough to need watching. A UK court has had to deal with lawyers filing case citations that AI invented. A major employer's AI hiring chatbot was left behind a test account with the password 123456. AI is good enough to change who does what. It is not good enough to be left alone.

This is the part we think matters most for a small business. Your experienced people learned their judgement doing junior work. If AI does all of that work, nobody learns it, and in five years you have a team that cannot check what the AI produces.

The answer is not to avoid AI. It is to redesign the junior role around it. Checking the AI's output before it reaches a customer, handling the cases it hands over, spotting when it is wrong and fixing the rules: that is real work, it builds exactly the judgement you want, and it is work AI cannot be trusted to do for itself.

Whether AI should be taxed to protect jobs is a question for government, and Reed himself says it would take some designing. We do not take a side. For a business owner the decision is closer to home: when you bring AI in, what happens to the people and the roles around it.

Automate the tasks, not the role. List the repetitive work AI can take, and decide what the person who did it will do with the time.

Keep a person in charge where a mistake would matter, and make that a junior's job to learn, with someone senior checking.

Keep hiring at entry level if you can. The cost of training someone is real; so is the cost of having nobody who understands how your business works when the AI gets it wrong.

That is how we build AI into a business: the repetitive part automated, a person in charge of the rest, and the roles designed around it. If you want to work out what that looks like for yours, book a 20-minute call.