The Hydrogen Center for Artificial Intelligence

NewsletterNo. 9

The missing first rung

Young college graduates are having a harder time finding work. Is AI the cause? A long look at what the best evidence says, where it disagrees, and why it matters for the next decade.

By Yonathan Shanko, Founding Editor

The short version

Something is going wrong at the start of the career ladder. Young college graduates are having a harder time finding work than they did a few years ago, even though the overall job market is not in a recession.

AI is the obvious suspect, and it gets most of the blame in headlines. The best research so far tells a more interesting story. AI appears to be part of it, especially in the jobs it can do on its own. But at least one major study points to a different cause entirely. This piece walks through the evidence.

What is happening

The Federal Reserve Bank of New York reports that the unemployment rate for recent college graduates ages 22 to 27 was 5.6% in March 2026, up from 3.6% in March 2019.

The New York Times put that in context in June. Recent graduates' unemployment has risen sharply over three years, and more than 40% of employed recent graduates work in jobs that do not require a degree, the Times reported. The rate is still below the peaks of the Great Recession and the early 1980s, but above the levels seen after the dot-com crash.

At the same time, companies are increasingly naming AI when they cut jobs. The outplacement firm Challenger, Gray & Christmas reports that employers cited AI in 120,136 announced job cuts through September, about 21% of the total. But total announced cuts were down 39% from the same period in 2025.

So the overall labor market is not collapsing. The pressure is concentrated at the entry level.

The case that AI is a cause

The strongest evidence comes from Stanford. Economists Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen used payroll data from ADP, the largest payroll processor in the US, covering millions of workers. Their paper is called "Canaries in the Coal Mine?"

They found that workers ages 22 to 25 in the occupations most exposed to AI, such as software development and customer service, saw a 13% relative decline in employment since generative AI became widely used. More experienced workers in the same jobs held steady or kept growing.

Three details make the finding hard to dismiss:

  • It holds up under checks. The pattern survives when the researchers exclude technology companies and when they exclude jobs that can be done remotely.
  • It shows up in hiring, not pay. Companies are adjusting by hiring fewer young workers, not by cutting their wages.
  • It depends on how AI is used. Declines are concentrated in jobs where AI can do the work itself. In jobs where AI helps workers do their work, employment for young people held steady or grew.

The researchers' reading is that young workers' main value often comes from knowledge that can be written down and taught, which is exactly what language models are good at. Experienced workers rely more on judgment and know-how that AI cannot yet reproduce.

The case that something else is going on

In June, three economists at the New York Fed, Natalia Emanuel, Emma Harrington, and Amanda Pallais, published a different explanation: remote work.

Their argument is that remote work makes it harder for managers to train and mentor new employees. If a company cannot easily bring an inexperienced hire up to speed, it is less willing to hire one. They estimate that remote work can explain 64% of the recent rise in unemployment among young college graduates, and they found little effect from AI.

Harvard economist Larry Katz made a similar point to the Times, saying the overall economy is not in a recession but it feels like one for young graduates, and worrying that remote work reduces chances to learn on the job.

Can both be true?

Probably, and the two explanations may reinforce each other.

The Stanford study finds AI-driven declines even in jobs that cannot be done remotely, so remote work cannot be the whole story. The New York Fed study finds that remote work explains a large share of the overall rise in graduate unemployment, so AI cannot be the whole story either. They also measure slightly different things: one tracks employment in specific AI-exposed occupations, the other tracks unemployment across all young graduates.

What both point to is the same weak spot. Entry-level jobs have always been partly about training. Companies hire inexperienced people, teach them, and get the payoff later. Anything that makes that training harder, or makes it easier to skip, puts the first rung of the ladder at risk.

Why it matters

The first job matters more than it seems. Research cited by the Times shows that graduating into a weak job market can lower earnings for many years. A study of the early 1980s recession by Lisa Kahn of the University of Rochester found the wage gap still visible about 15 years later.

UC Berkeley economist Jesse Rothstein told the Times that if AI disproportionately hits entry-level work, those long-term effects could be larger than in a normal downturn.

There is also a pipeline problem. Today's senior engineers, analysts, and managers learned their jobs as junior employees. If companies stop hiring juniors because AI or remote work makes them less useful in the short run, the supply of experienced workers shrinks later.

A note on the layoff headlines

Be careful with "AI caused X layoffs" headlines. Layoff trackers record the reason a company gives, not a verified cause. Blaming AI can sound better to investors than admitting a company over-hired or lost business. A company may also cut jobs to free money for AI spending, even if no AI is doing those jobs yet.

As Challenger's Andy Challenger put it in the firm's July report: "AI is shifting the labor market, it is not dismantling it." The payroll studies are the better guide to what is actually happening.

What to watch

  • Whether the recent-graduate unemployment rate keeps rising as AI tools spread further, or levels off.
  • Whether companies that return to the office start hiring more entry-level workers, which would support the remote-work explanation.
  • Whether employers redesign entry-level roles around working with AI rather than replacing them. The Stanford data suggests that is where young workers are holding their ground.
  • Whether new research separates the effects of AI and remote work more cleanly.

Sources

If you study labor markets or hire entry-level workers and see this differently, send us a pitch. Everything we publish is free. If you want to help keep it that way, become a supporter.

The missing first rung · The Hydrogen Center