6 minEconomy
AI Is Cutting the Hidden Subsidy That Turned Entry-Level Jobs Into Careers
A former Fed economist argues that entry-level white-collar jobs were never about immediate output — they were apprenticeships funded by the productive work of juniors. As AI absorbs that work, firms are hiring fewer juniors, and the pipeline that turns them into senior professionals is thinning.
Entry-level white-collar jobs were never really about the work they produced. They were apprenticeships in disguise, funded by the modest output of junior employees while firms invested in turning them into senior partners, analysts, and researchers. Now, as artificial intelligence absorbs more of that routine output, companies are hiring fewer juniors, and the pipeline that manufactures professional judgment is thinning.
The debate over AI and entry-level work has been dominated by a single narrative: that automation is destroying the first rung of the career ladder. Stanford's Digital Economy Lab has 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. That finding has fueled widespread panic that college degrees are no longer worth it and that a generation of graduates faces a broken job market.
But the timing tells a more complicated story. Job postings for the occupations most exposed to AI peaked in March 2022 — the same month the Federal Reserve's policy-setting committee began raising interest rates — and then began to fall. ChatGPT would not exist for another eight months. Zanna Iscenko and Fabien Curto Millet, analyzing 238 million job postings, tied the decline to the tightening cycle and noted that AI-exposed occupations tend to cluster in information, finance, and professional services, sectors especially sensitive to interest rates.
The Economic Policy Institute adds a further complication: young workers without college degrees, whose occupations score negative on AI exposure, also saw their unemployment rate rise at a similar pace over the same period. If AI were the sole culprit, unexposed workers would not be suffering too. The Stanford researchers themselves caution that their findings are «descriptive patterns, not causal estimates» and that they «do not see widespread, economy-wide job displacement associated with AI.» They have also pushed back on the monetary policy explanation, noting that the most exposed jobs are not generally the most rate-sensitive and that the employment gap for young workers in exposed occupations has continued to widen even as rates have come down.
The truth is that it is too early to know exactly what will happen to the labor market as AI spreads, because overlapping shocks cannot be separated in real time. Yet businesses and universities must make expensive decisions right now about hiring, education, and regulation as if the answer were already known. One pattern has become clear in the data: firms are not firing their junior employees. They are hiring fewer of them, and the decline concentrates where AI automates work rather than where it complements it.
That raises a question the Stanford team's rebuttal does not fully answer. Why are firms not firing juniors? What were they getting all those years when they hired them? They clearly were not hiring for output productivity. A first-year associate's document was checked by a partner and frequently had to be redone. The patient history a resident took at 2 a.m. often had to be retaken by the attending physician. By any honest accounting, the work was unproductive. But firms bought it anyway because that is how a junior employee becomes a senior partner. The output was the byproduct; the formation of the employee was the point, and what the junior produced helped offset the cost.
AI can now do the work of the junior analyst, so firms invest less in these hires, and the pipeline that turns juniors into seniors thins. Matt Beane watched this mechanism in operating rooms years before ChatGPT, as surgical robots quietly cost residents the case time that made them surgeons. Employers still want experienced people, but many have simply stopped funding the process that produces experience.
The same problem faces universities. The Ph.D. is an apprenticeship funded by the productive value of early-career researchers' work, and the editors of Nature warned this spring that those researchers now face the danger that «tasks that are crucial to their training as scientists are done by a machine.» Classroom evidence points in the same direction: students learn when the tool is constrained so that effort cannot be skipped, and fail to learn when it hands over answers.
Forming judgment requires friction — dealing with tough problems, failing, and trying again. Getting the right answer is not the point. As math teachers have long insisted, «Show your work.» There is a version of this transition in which 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. It arrives only if employers keep booking formation as a cost they can finally cut, and universities keep certifying work the machine did. The postings data will recover when the hiring cycle turns, but the training that was quietly subsidized by junior output may not come back on its own.
8
