How far have UK entry-level jobs fallen?

Adzuna data shows graduate vacancies down 45.6% to 8,383 in the year to July, and non-graduate entry-level roles down 8% to 192,864. Employers cite higher employment costs alongside AI increasingly handling tasks previously given to junior staff.

Which junior tasks should a small business automate?

The repetitive ones that teach nothing: copying data between systems, formatting routine documents, and standard payment reminders. Keep exceptions, reconciliations that do not match, and supervised customer contact, because those are where judgement is actually learned.

Why does this matter more for small businesses?

A small firm's succession plan is usually one or two people who joined junior and grew into the business. Removing the entry route leaves external senior hiring as the only option, into a business with undocumented processes and at a salary it was trying to avoid.

Can you automate junior work and still train people?

Yes, by inverting the usual allocation. Give the learner the exceptions and odd cases with a senior person available, rather than the routine volume. That compresses into a handful of interesting cases the judgement that used to take thousands of routine ones.

If AI Does All The Entry-Level Work, Who Becomes Your Senior In Five Years?

UK graduate vacancies fell 45.6% to 8,383 in the year to July, and non-graduate entry-level roles fell 8% to 192,864, according to Adzuna. Employers give two reasons: higher employment costs, and AI now handling tasks that used to be done by junior staff. Average advertised salaries weakened to £43,675 in July.

The immediate logic is sound. If software does the work for a fraction of the cost, hiring someone to do it is hard to justify. The problem is what entry-level work was also doing, which nobody was paying for explicitly and everybody was relying on.

What junior work actually produces

Two outputs, only one of which appears on a job description.

The visible output is the work itself: the data entry, the first-draft document, the routine call, the chased invoice. This is genuinely automatable and increasingly automated, and there is no strong argument for a person doing it.

The invisible output is judgement. People learn what "wrong" looks like by handling several hundred ordinary cases and a handful of strange ones. A person who has processed two thousand invoices knows instantly when one is odd. Somebody who has never processed any cannot supervise a system that processes them, cannot spot when it fails, and cannot make the decision when it escalates.

Automate all of the first and you lose all of the second, on a delay long enough that the decision and the consequence never appear in the same financial year.

Why this lands harder on small businesses

A large firm can absorb it. It has enough people that some learn by exposure elsewhere, and it can hire experienced staff externally when it needs to.

A ten-person business cannot. Its succession plan is usually one or two people who came in junior and grew. If the owner wants to step back from operations in five years, the person who takes over is someone who is currently learning the business by doing its ordinary work. Remove that route and the only alternative is hiring a senior person externally, into a business with no documented processes, at a salary the business was trying to avoid in the first place.

The distinction worth drawing

Not all entry-level work teaches anything. The useful test is whether doing the task repeatedly builds a model of how the business works.

Task type	Teaches judgement	Automate?
Copying data between systems	No	Yes, entirely
Formatting and sending routine documents	No	Yes
Handling exceptions and odd cases	Yes	No, route these to a person
First-line customer contact under supervision	Yes	Keep, with support
Reconciling figures that do not match	Yes	Keep, assist with tooling
Chasing a routine payment	Marginally	Automate the routine, keep the difficult ones

The pattern is that the boring, repetitive version of a task teaches very little and should be automated, while the exceptions teach almost everything and should deliberately be given to the person who is learning. That is close to the opposite of how most businesses assign work, because exceptions are hard and get escalated to whoever is most experienced.

What to do instead of hiring nobody

Keep one entry-level route open, even if the role looks different from the one you would have advertised three years ago. It is more likely to be an assistant who supervises automated processes than one who performs them.
Give juniors the exceptions on purpose. With a senior person available, not absent. This compresses the learning that used to take two thousand routine cases into a much smaller number of interesting ones.
Document as you automate. Every process you hand to software is a process that leaves the heads of the people who knew it. If it is not written down at the point of automation, it is gone.
Do not treat apprenticeship funding as the whole answer, but do check what is available before assuming a junior hire is unaffordable.

The honest counterpoint

There is a real argument on the other side, and it is not silly: a small business struggling with employment costs cannot fund a training pipeline for a benefit that arrives in five years, and may not exist in five years. That is a legitimate position and this is not an argument that every SME should hire a graduate.

It is an argument for making the choice deliberately. Automating junior work to survive a hard year is a decision. Automating it without noticing that you have also closed your only route to a senior hire is an accident, and it is the more common of the two.