Entry-level job postings in New York's computer and mathematical occupations have fallen 49% since 2022, according to a report from the nonprofit Center for an Urban Future, first reported by The City. Most predictions of mass AI job destruction have not come true. This is the exception, and it has landed on the people who were supposed to go on to build the systems: computer science graduates looking for a first job.
John Bowles, Urban Future's executive director, told The City that entry-level hiring has "collapsed" across the board, with the bottom rung of the tech career ladder under particular pressure. For years, he said, the tech sector grew fast and generated a large number of junior positions, which made it a good opening for people starting out. Those positions are now rarer, and competition for the ones that remain has intensified.
The decline is not confined to tech. The report also finds noticeably reduced entry-level hiring in office and administrative support, financial operations, arts, design, entertainment, sports and media, education and libraries, and other industries.
The report is careful about cause. AI is the most obvious factor, but it is not the only one: tech companies have been cutting jobs at scale to correct for over-hiring during the COVID-19 pandemic, and that correction happened to coincide with the fast spread of generative AI. The findings also arrive against a sharply deteriorating US labour market — as of July, close to a quarter of the American workforce was considered functionally unemployed.
Read that list of affected sectors again, because it is the most useful thing in the report and it argues against the headline. Libraries and administrative support are not being hollowed out by chatbots at the rate computer occupations are, and yet junior hiring fell in all of them. When a category as broad as "entry-level" contracts everywhere at once, the common cause is usually the cost of money and the mood of the people approving headcount, not a specific technology.
The 49% figure deserves the same scrutiny. It covers one city, one occupational family, and it counts postings, not jobs. A posting is what a company does when it intends to hire, which makes it the cheapest thing to withdraw and the first thing to move when the outlook turns. Postings can fall by half while employment falls by far less, and they can come back faster than payrolls can. That does not make the number meaningless — it makes it a measure of intent, and what it shows is that firms currently do not intend to hire juniors.
The framing in the report is the honest one: companies are using AI as an argument for cutting new hiring. That is a different claim from AI having replaced the work, and it is the more plausible of the two. An executive who over-hired in 2021 and needs to explain a shrinking headcount has been handed the most flattering possible reason — not a misjudgement of the cycle, but a technological leap. The report's own conclusion concedes as much: switching off every AI chatbot tomorrow would not by itself restore the entry-level openings that have disappeared.
What nobody in this discussion has taken up is the arithmetic on the other side. Senior engineers are made out of junior engineers, over roughly five to ten years, by having them do work that someone more experienced reviews. An industry that stops creating first jobs is not just declining to hire this year's graduates; it is declining to produce the people it will need to promote in 2032. If AI genuinely does the junior work, the pipeline problem is theoretical. If it merely provides cover for a hiring freeze, the bill arrives later and cannot be paid retroactively.
That is the real exposure in the 49%. The cycle will turn, the postings will come back, and the cohort that was supposed to be learning during this stretch will be several years into other careers.