Stanford Study: AI Is Hitting Entry-Level Jobs Hardest in 2025

The Numbers Behind the Gap

Researchers rated occupations by their “AI exposure” using two frameworks: an established labor market impact gauge and the Anthropic Economic Index, which tracks how workers actually use Claude across different job types.

Economy-wide, the employment difference between high- and low-AI-exposed jobs is minimal. But for 22-to-25-year-olds, the divergence is stark:

• Employment in the top 40% of AI-impacted jobs has fallen roughly 11% since 2022
• Employment in the bottom 60% of AI-impacted jobs has grown by 10% over the same period

That’s a 21-point swing — driven almost entirely by one age group.

It’s About Hiring, Not Firing

Here’s the detail that matters most: this isn’t showing up as mass layoffs. It’s showing up as fewer people getting hired in the first place.

Entry-level workers in AI-exposed fields aren’t being pushed out. They’re not getting in the door. The on-ramp is quietly closing.

Pay rates for those who do get hired haven’t dropped significantly. The damage is happening upstream — at the point of hiring.

Automation vs. Augmentation: The Distinction That Changes Everything

Not all AI exposure is equal. The Anthropic Economic Index separates AI usage into two categories:

• Automative — AI fully replaces tasks a human previously did
• Augmentative — AI helps a human do their job better

The employment data breaks cleanly along this line.

Jobs where AI is primarily automative — think receptionists, accountants, data entry roles — are showing the steepest declines in entry-level hiring. Jobs where AI is augmentative — chief executives, registered nurses — show flat or even rising employment.

The researchers put it plainly: automation-oriented AI is substituting for labor. Complementary AI is not.

Codified Knowledge Is the Vulnerability

The study goes one level deeper with a concept worth understanding: codified vs. tacit knowledge.

Codified knowledge is formal, teachable, and documented — the kind you get from textbooks and training programs. Tacit knowledge is earned through experience, mentorship, and years of doing the work.

Entry-level jobs lean heavily on codified knowledge. That’s exactly what AI is best at replicating.

The data backs this up. Occupations requiring higher levels of formal education — used as a proxy for codified knowledge — are seeing slower entry-level employment growth. Occupations built on tacit knowledge are seeing faster growth for mid-career and senior workers.

AI is, in effect, compressing the value of the early career learning curve.

Does a College Degree Still Help?

Somewhat — but not uniformly.

The researchers found that in occupations with a higher share of college graduates, the employment gap between AI-exposed and non-exposed roles is more muted. The degree appears to provide some buffer.

In occupations with fewer college graduates, the pattern is sharper: AI-exposed roles are declining, while non-exposed roles are growing.

This doesn’t mean a degree is a guaranteed shield. It means the risk is distributed unevenly, and workers without degrees in AI-exposed fields are absorbing the most pressure.

What Lead Researcher Erik Brynjolfsson Is Warning About

Stanford researcher Erik Brynjolfsson framed the risk clearly: a labor market that maintains its headline employment numbers while systematically eliminating the entry points for new workers.

“The entry-level effects we’re measuring are real, persistent and widening,” he noted, “and I’m more worried than I was about a labor market that keeps its overall employment level while quietly closing the on-ramp for people starting their careers.”

That’s the scenario worth watching. Not mass unemployment. A labor market that looks healthy in aggregate while a generation of early-career workers finds fewer and fewer ways in.

What This Means If You’re Tracking AI Tools

For anyone evaluating AI tools — whether you’re building workflows, advising teams, or hiring — this research surfaces a practical question: is the tool you’re deploying automative or augmentative?

That distinction isn’t just academic. It determines whether AI is expanding what your team can do or quietly replacing the roles you’d otherwise be filling.

The tools that augment experienced workers are showing up on the right side of this data. The ones replacing codified, entry-level tasks are showing up on the wrong side — at least from a workforce health perspective.

Choosing AI tools with that lens in mind isn’t just an ethical consideration. It’s increasingly a strategic one. Related themes also appear in AI investment surges, but employees feel unready and Robot Relations Departments and the Future of HR.