Firms are setting their 2027 budgets right now, and one of the first numbers on the table is the size of next year’s entry-level class. The easiest hire to cut is the one you haven’t made yet. Nobody gets a severance package, it doesn’t make the news, and the seat just never gets filled. More and more, the reason given is AI.
Last week, Statistics Canada, the Canadian government’s national statistics agency, published a study that tested that reason against 20 years of hiring data. Its answer was no. Not yet.
What we’ve been hearing
For about a year and a half, the story has been alarming. The biggest piece of evidence is Stanford’s Canaries in the Coal Mine paper, which uses ADP payroll data on millions of US workers. The August update found that employment for 22- to 25-year-olds in the jobs most exposed to AI is about 19% below where it would be if it had kept pace with young workers in less exposed jobs. Experienced workers in the same jobs show no comparable gap, and the gap comes from fewer young people being hired, not more being let go.
Accounting has its own version. In the UK, KPMG cut its 2023 graduate intake from 1,399 to 942, and Deloitte cut its intake by 18%, according to Accountancy Age. A Work Foundation survey this summer found 43% of UK employers said AI or automation had reduced their entry-level roles. Put together, the headlines point one way: AI is eating the first rung of the ladder.
What Canada found
The Statistics Canada study, by Tahsin Mehdi and Marc Frenette, looked at labor force data from January 2006 through August 2026. The researchers sorted occupations by how exposed they are to AI and by whether AI is more likely to replace the work or help the person doing it. Then they measured how often people were hired into those jobs and how often people left them.
Since ChatGPT launched, hiring and separation rates have stayed within their historical ranges or followed trends that were already in place. The agency found no broad evidence that AI has cut hiring or increased job losses, including for young workers.
Young Canadians are having a harder time getting hired, but it’s happening everywhere. In the jobs with the lowest exposure to AI, the hiring rate for workers aged 15 to 29 fell from about 4% to just over 3%. If AI were the cause, hiring wouldn’t fall in jobs it can’t do.
The US data points the same way. Yale’s Budget Lab, which tracks employment data month by month, found no clear evidence of AI disruption in its September update. A University of Pittsburgh-led team found that job prospects in AI-exposed occupations started getting worse in early 2022, months before ChatGPT, as interest rates rose and companies that over-hired during the pandemic pulled back.
Canada and Stanford can both be right. Stanford zooms in on 22- to 25-year-olds in the most exposed jobs. Canada looks at a wider age band across the whole economy, and a narrow effect can get lost in a broad average. Stanford’s own authors describe their numbers as descriptive indicators, not proof that AI caused the gap.
Accounting has been here before
If most of this is a weak market, it’s temporary. Markets turn. In 2008 and 2009, many firms, especially the largest, cut thousands of staff positions and pushed back start dates. Many of the people who were laid off or never hired left the profession. By 2015, the Journal of Accountancy was describing a talent shortage that AICPA leaders called more intense than anything they’d seen, and it named those recession-era cuts as one of the causes. UK recruiters later traced a shortage of mid-level accountants back to firms cutting graduate programs in 2008. The cut saved money in 2009. The bill came due six to ten years later.
Think of a baseball team’s farm system. The minor leagues are where you develop players cheaply, over years, so they’re ready when the big-league team needs them. Shut it down to save money and you still need players. You just have to sign free agents, and they cost more. Wharton’s Matthew Bidwell measured that cost using personnel data from a large investment bank: outside hires were paid about 18% more than people promoted into the same jobs, got lower performance ratings for their first two years and were more likely to leave.
A weak market also makes this a buyer’s market for talent. In the UK, graduate employers now report an average of 140 applications per vacancy, the highest since they started tracking in 1991. If you have room to hire, you have more choice than you’ve had in years.
Key takeaways
- The evidence points to a weak hiring market, not AI replacing entry-level work. Stanford’s gap is narrow and doesn’t prove cause.
- Before your team cuts the campus class, ask whether you’re assuming the slowdown is temporary or permanent. If it’s permanent, someone should be able to point to the work AI is doing today. If nobody can, it’s a budget cut, and it should be called one.
- In a weak market, make choices you can undo. Defer start dates instead of pulling offers, keep your campus relationships and internship program, and set a date to revisit the number.
- Use the market. If you can hire, be selective and go after people who would normally go somewhere else.
Want the CPE credit? Take the full lesson on EverydayCPE and earn 0.2 CPE credits: AI Isn’t Why Juniors Aren’t Hired
Leave a Reply