Daily.Brief

Work

The entry-level job market when AI does the junior work

#005June 1, 20268 min readBy Joseph

Canada's unemployment rate for people aged 15 to 24 sat above 14 percent for much of 2025, the worst reading outside a recession since the 1990s. For new graduates the picture was not much better. Some of that is the economy. Some of it is that the tasks a junior hire used to do are now done by a model in ten seconds.

What a junior job used to be for

A first job in consulting, law, accounting, marketing or software was mostly a training arrangement disguised as work. The firm paid you a real salary to do the parts of the job that were tedious but teachable: draft the first version of the memo, build the model, pull the comparables, write the boilerplate. You were slow and the work needed checking, but two years later you could do the senior person's job.

Those tasks are precisely the ones that current AI models do well. A first draft, a first model, a first pass at the research. Not the judgement about whether the draft is right, but the draft itself. So the economic reason to hire a junior, which was that a senior's time was too expensive to spend on drafts, is weaker than it was.

What the data shows so far

The clearest signal is in hiring, not firing. Large firms are not laying off analysts because of AI. They are hiring fewer of them. Job postings for entry-level software roles in North America fell sharply from their 2022 peak and did not recover with the rest of the market. Postings for senior roles held up.

It is hard to separate this from the interest rate cycle, which hit hiring at the same time. Researchers at Stanford who looked at payroll data in 2025 found that employment for workers in their early twenties in the most AI-exposed occupations fell noticeably, while employment for older workers in the same jobs did not. That is the pattern you would expect if the junior rung were being automated first.

Canadian youth unemployment rate, ages 15 to 24
Canadian youth unemployment rate, ages 15 to 245%10.1%15.1%20.1%201920202021202220232024202514.2%
Annual average, Statistics Canada Labour Force Survey, rounded. The 2020 spike was the pandemic; the 2025 level came without a recession.

The pipeline problem nobody has solved

If firms stop hiring juniors, in five years they have no mid-level people, and in ten years no seniors. Every firm knows this and every firm is hoping someone else trains the next generation. It is a classic collective action problem. The rational move for one company is to skip the junior class this year. The rational move for the industry is to keep hiring. Those two things point in opposite directions.

Some firms are trying to redesign the first job instead of eliminating it. Fewer hires, but each one paired with a senior from day one and given real responsibility faster, with the model doing the drafting. Whether that works at scale is an open question, and it is the question I would ask in any interview.

What to do about it if you are seventeen

Stop optimising for the skills that a model does well. Perfect essays, clean spreadsheets, tidy code that follows the pattern. Those are still table stakes, but they are no longer the thing that gets you hired.

The thing that gets you hired is judgement, which is the ability to look at a draft and know it is wrong, and the ability to decide what to work on in the first place. You build that by doing real things with real consequences: running a club with a budget, building something people use, competing at something where you can lose. I would take a student who has done one of those over one with a perfect transcript, and increasingly so would the firms.

Subscribe

The next one, in your inbox.

Free, about 7:00 AM Eastern. English or Chinese, daily or weekly.

Free. No spam. One click to leave, and you can change language or frequency from any email.