The number that went up is the one to distrust

Jack Kennedy, senior economist at Indeed, published the company's mid-year UK hiring report on 3 August and put a plain description on the data: the labour market is increasingly splitting into two speeds. Software development postings rose 14 percent. Manufacturing fell 58 percent from its June 2022 level and 18 percent over the past year. Accounting and marketing both posted double-digit declines. Read quickly, that 14 percent looks like the technology sector recovering.

It is not a recovery. Kennedy's own framing is that demand is concentrating around experienced workers and roles directly connected to AI rather than flowing evenly through the profession. A 14 percent rise made up largely of senior and AI-adjacent openings is a statement about which half of your org chart the market is bidding for, and the answer is the expensive half.

The pool underneath the split is getting smaller

The Office for National Statistics counted 712,000 vacancies across April to June 2026, below pre-pandemic levels, down 7,000 on the previous quarter and falling in 10 of the 18 industry sectors it tracks. There were 2.5 unemployed people for every vacancy, a ratio that has not moved since the autumn of 2025. These are survey figures gathered by a national statistics office, entirely independent of a job board's own postings index, and they describe the same market from the opposite direction.

That independence is what makes the conclusion firm. If only Indeed showed the pattern it could be a quirk of who advertises where. With the ONS showing a contracting total and Indeed showing where the remaining demand has moved, the picture is a redistribution inside a shrinking market. Competition for the profile everyone now wants is therefore rising faster than the headline vacancy number suggests.

Entry level is a supply line, not a cost line

The reasoning that removes graduate roles is usually sound in isolation. AI tools cover a growing share of the work a first-year hire used to do, that hire needs supervision from someone senior, and the saving is immediate and easy to show. Youth unemployment in Britain is now at its highest in over a decade, which tells you how many organisations have run that calculation and reached the same answer.

The part that does not appear in the calculation is where the senior AI-capable engineer of 2031 comes from. There is no external supply of people with five years of experience who did not first have one year of it. Cutting the intake does not remove the cost of that person. It converts a training cost you control into a market price you do not, paid in the one segment where every competitor is bidding at once.

Decide which of the two you are actually choosing

This is a decision rather than a trend to observe, and it is worth making explicitly instead of by accumulation. An organisation that trims graduate intake for three consecutive budget rounds has committed to permanent senior hiring on the open market without ever putting that commitment to a vote. The base rate to check is your own: what share of your current senior engineers did you hire as juniors?

If that share is high, your pipeline is the asset producing your scarcest input, and cutting it is a decision about 2031 disguised as a decision about this quarter. If it is low, you already buy seniors on the market, and this data says the price is about to move against you. Either answer is workable. Not knowing which one applies to you is the expensive position, and it is the most common one.