Two days before this weekend, a commentary in Fortune put one of the sharpest framings yet on the entry-level AI debate — and it came from someone who spent a decade watching the US labour market from inside the Federal Reserve.
Patrick T. Harker, former president and CEO of the Federal Reserve Bank of Philadelphia and now a Wharton professor, published a commentary in Fortune on September 25 with an argument that reframes the “AI is destroying graduate jobs” story: entry-level white-collar roles were never really about the work. They were about making senior people.
The argument in one line
“Firms are not firing their junior employees; they’re hiring fewer of them,” Harker writes. The output of a first-year analyst was never the point. A first-year associate’s document got checked and frequently redone; a junior doctor’s patient history often got retaken at 2am. “By any honest accounting of the work, it was unproductive. But firms and organizations bought it anyway because that’s how you make a junior employee into a senior partner.”
In other words: the routine work was the byproduct. The formation of the future senior professional was the product — and its cost was quietly offset by the useful things juniors produced along the way. AI, able to do much of that routine work, has cut the subsidy. Firms now see junior hires as a cost line they can shrink, and the pipeline that turns juniors into seniors is thinning.
Why the diagnosis is contested
Harker’s piece is a direct engagement with Stanford Digital Economy Lab’s “Canaries in the Coal Mine” research, which reported that workers aged 22 to 25 in the most AI-exposed occupations are running roughly 19 percent behind peers in less-exposed fields, a gap that has widened over the past year.
But the timing question is genuinely open, and Harker — who argued in March 2022 that rate-hike cycles drove the same posting declines before ChatGPT even existed — points to countervailing evidence: young workers in occupations scoring negative on AI exposure also saw unemployment rise at a similar pace, per Economic Policy Institute analysis. And, in fairness to the Stanford team, their own paper states these are “descriptive patterns, not causal estimates” and that they do not see widespread, economy-wide job displacement from AI. The Stanford researchers have also pushed back on the monetary-policy reading in their own follow-up.
Harker’s synthesis: it is too early to be sure what is causing what, because overlapping shocks can’t be separated in real time. What is clear in the data is that the decline concentrates in hiring, not firing.
What this looks like in New Zealand
New Zealand’s graduate market is telling a similar story in its own accent. Reserve Bank of New Zealand’s analytical work has examined how the first rung of the graduate ladder is thinning as AI takes on junior tasks — we covered it earlier this year in our piece on the RBNZ graduate ladder. The pattern Harker describes — fewer first rungs, not mass firings of existing juniors — matches what NZ employers report: they are not cutting graduates already on staff, they are simply posting fewer graduate positions.
SEEK’s New Zealand data puts numbers around the other side of the market: job ads rose for a 21st straight month in August, with 4.0 percent of listings now referencing AI skills — up 93 percent year on year. Hiring is rotating, not disappearing.
The fix costs money — and someone has to budget for it
The most practical part of the piece is where Harker lands. Stop booking junior hiring as a cost line “that automation just erased,” he writes. “It was never an operating expense; it was a capital investment mislabeled.” Firms that still want experienced senior staff in ten years have to deliberately fund what used to be cross-subsidised: structured rotations, real mentoring, deliberate training.
He points to the same mechanism in science: Nature’s editors warned this spring that early-career researchers face the risk that “tasks that are crucial to their training as scientists are done by a machine.”
And he offers a memorable line for the classroom version of the problem: “No friction, no traction.” Students learn when the tool is constrained so effort can’t be skipped; they don’t learn when it hands over answers.
The open question for workers
There is a version of this transition where AI does the routine work and an entire generation never gets the repetitions that turn talent into judgment. Nothing in the technology makes that outcome inevitable, Harker concludes — it arrives only if employers keep treating formation as a cost they can finally cut.
For young workers, the practical read is neither panic nor comfort: the scarcity is in formation, not in talent. Roles and employers that still invest heavily in deliberately training juniors are where the ladder is being rebuilt — and as Amazon’s recent moves to rehire AI-era leavers suggest, firms are discovering what it costs when the pipeline runs dry.
FAQ
Is AI really the cause of the entry-level hiring slowdown? It is contested. Stanford’s Digital Economy Lab finds young workers in the most AI-exposed occupations trailing peers by about 19 percent, while other research ties the timing to interest-rate tightening. The Stanford researchers themselves call their findings descriptive, not causal.
Are companies firing junior employees because of AI? Data cited in the Fortune commentary suggests the opposite: firms are not firing juniors — they are hiring fewer of them, with the decline concentrated where AI automates routine work.
What should job seekers entering the workforce do? The argument points toward seeking employers that still fund structured training, mentoring and rotations — the places deliberately rebuilding the junior-to-senior pipeline.
— CJ Murden, editor of Singularity.Kiwi. Former digital technologies teacher, author of AI-focused books. Writing with a New Zealand focus.
