A Nobel economist knew 60 years ago that people learn best on the job, and now the Atlanta Fed says companies automating lower payroll costs may be sawing off the branch they’re sitting on

You spent four years and tens of thousands of dollars obtaining a degree. You polished your LinkedIn, tweaked your resume and applied to what seemed like hundreds of jobs, and then nothing. Or worse, a role that has nothing to do with your study.

If this sounds familiar to you, you are not imagining things. Right now, the entry-level job market in the U.S. is really tough, and a growing body of research suggests AI has a lot to do with that.

The first job is not just a paycheck; it is a classroom
There’s a reason economists speak of “learning by doing.” In a landmark paper published in The Review of Economic Studies in 1962, Nobel Prize-winning economist Kenneth Arrow argued that workers don’t just produce output on the job; the more they actually do the work, the smarter, faster and more capable they get. He said experience is the real productivity engine. You can't hack it with a textbook.


This idea is hugely relevant right now, because companies are increasingly using AI to automate the very kinds of work that used to be found in entry-level jobs: drafting documents, pulling data, answering emails, and doing research. Stuff that was part of the job description for a 22-year-old.

When companies automate away junior jobs, they aren't just cutting costs today, researchers at the Federal Reserve Bank of Atlanta recently found, looking back at Arrow's framework, something alarming. They are quietly gutting their own pipeline of talent for tomorrow. Researchers said entry-level work is basically a crash course to turn new graduates into the competent senior employees companies will desperately need years down the line. If you take that on-ramp away, you get a generation of workers who never got the reps in.

The data is already showing warning signs
This is not merely theoretical. A landmark study from the Stanford Digital Economy Lab, “Canaries in the Coal Mine?” found that early-career workers ages 22 to 25 in the most AI-exposed occupations, including software development and marketing, experienced a 13% relative decline in employment compared to older workers in those same jobs. The researchers drew on payroll data from ADP for millions of workers at tens of thousands of US companies, making it one of the most concrete, large-scale investigations of AI’s initial effect on the labor market. The hits were in roles susceptible to automation, while jobs where AI augments rather than replaces workers actually saw employment grow.

At the same time, the jobless rate for recent college graduates is consistently higher than the national rate overall, a reversal of historical trends that most economists find highly unusual. A college degree, long considered the surest ticket into the professional middle class, is starting to appear like a less reliable investment.

Companies are winning in the short run and losing in the long run
Here’s the real talk for employers: the short-term math on automating entry-level work is attractive. Lower payroll costs, faster throughput, and fewer management headaches. But the researchers at the Federal Reserve Bank of Atlanta warn that this is a tradeoff companies are making without fully factoring in the long-term bill.

Rob an entire generation of young workers of that basic, on-the-job experience and the economy doesn’t just generate fewer junior employees, it eventually generates weaker senior leaders, less capable managers, and slower innovation across entire industries. Arrow’s fundamental insight was that the gains from learning-by-doing do not remain within a single firm. They ripple outwards, raising the economy's overall productivity. So the decision of one company to automate its entry-level workforce is not just its own problem.

One practical solution proposed by researchers is to tax automation profits and subsidize companies that actively create and grow entry-level jobs, giving young workers a chance to learn the trade.

What it means for you
If you’re in your 20s and having trouble finding your footing in your industry, the forces working against you are real, not a reflection of your resume. The labor market has changed fundamentally, at least for now. Some analysts point to more general reasons: geopolitical uncertainty, corrections after overhiring in the pandemic era, and a saturated pool of applicants. But it’s hard to ignore AI’s role in wiping out the very jobs that have historically served as career launch pads.

The skills gap that is happening today could be the talent crisis that companies will be complaining about ten years from now.