The Fear That AI Is Erasing Entry-Level Jobs

The fear is easy to state and hard to shake: AI is quietly wiping out the bottom rung of the career ladder, and the graduates trying to climb onto it are the first to fall. Unlike a lot of what we cover in this section, this one is not a viral rumour or a doomsday hunch. There is real data behind it. The honest reading, though, is more tangled than the headline suggests.

The most-cited evidence comes from a Stanford working paper by economists Erik Brynjolfsson, Bharat Chandar and Ruyu Chen. Looking at payroll records, they found that early-career workers aged 22 to 25 in the jobs most exposed to AI have seen employment fall by about 13 percent since 2022. In the hardest-hit fields, entry-level software engineering and customer service, the drop was closer to 20 percent between late 2022 and mid-2025.

What makes it harder to wave away is persistence. The original paper ran through August 2025, but the authors have since extended their dashboard to April 2026, giving almost four years of post-ChatGPT data. As Brynjolfsson told Fortune, the effect has not reverted. It has grown, by roughly half a percentage point a month, month after month.

Where the fear overreaches

So the fear is not baseless. But it is often stated too broadly. Two things are worth keeping straight.

First, the decline is concentrated, not universal. The study measures specific occupations that overlap heavily with what chatbots do well, jobs built on tasks a new graduate learns through formal education. It is not evidence that AI is shrinking employment across the whole economy, and total employment has not collapsed.

Second, the picture flips depending on which firms you look at. At companies actually building AI, hiring has gone the other way. Analysis from Ramp and Revelio Labs found those firms grew their workforce by around 10 percent, with entry-level headcount up about 12 percent. The same technology that appears to be closing doors in exposed occupations is opening them where the tools are made.

What the evidence actually supports

Put together, the responsible claim is narrower than the fear and more uncomfortable than the reassurance. AI does seem to be reshaping the entry level of certain white-collar fields, and young workers in those fields are bearing the brunt because their tasks are the easiest to automate. That is a genuine problem for anyone graduating into software or support roles right now.

What the data does not support is the sweeping version, that AI has made junior workers obsolete everywhere. Whether the trend spreads or stays contained to a handful of occupations is still an open question, and the researchers themselves are careful not to overclaim. We have looked before at the related worry that AI has made programmers obsolete, and the pattern holds here too. The threat is specific and measurable in places, and it gets less useful the more it is generalised.

If you are early in one of the exposed fields, the sensible response is not panic but positioning: the roles that pair human judgement with these tools are the ones the same data shows holding up best.