Stanford study: AI isn't cutting jobs overall — it's closing the door to first ones
Researchers find AI is slowing hiring for career-starter roles built on textbook knowledge, while jobs leaning on hands-on experience keep growing for workers already established.
What happened: A Stanford study led by economist Erik Brynjolfsson found that AI's effect on hiring isn't spread evenly across the workforce — it's concentrated at the entry level. Using the government's O*NET database as a stand-in for how much a job depends on formal, textbook-style training, the researchers found a clear split: jobs built on that kind of codified knowledge are seeing entry-level hiring slow, while jobs built on hands-on experience keep growing hiring for people already established in their careers.
Why it matters: Overall job totals can look steady even while the market quietly stops making room for beginners. Brynjolfsson told the Washington Post he's increasingly worried about a labor market that holds its headline employment level while closing the on-ramp for people just starting out. That matters because entry-level jobs are traditionally where people build the hands-on judgment that later makes them promotable — if AI absorbs the codified parts of those jobs first, new workers may never get the reps that turn them into experienced ones.
How it works, plainly: The study splits job knowledge into two kinds. Codified knowledge is what you can learn from a manual, checklist, or textbook — the kind AI can absorb because it's already written down. Tacit knowledge comes from doing a job for years: reading a room, knowing which shortcuts are safe, judgment calls nobody wrote a manual for. AI is strong at the first and weak at the second, so jobs stacked with codified tasks — often the ones aimed at recent grads, like administrative support, data entry, and basic customer service — are losing entry-level openings fastest.
Where the safety net helps: The researchers found one buffer: in fields where most workers already hold a college degree, the gap between AI-exposed and less-exposed jobs narrows. In fields with fewer degree-holders, the split is sharper — AI-exposed jobs shrinking while less-exposed ones keep growing. Separate reporting points to rising demand for people who can manage or direct AI tools, suggesting schools and employers may need to rethink how new workers get trained, since the old learn-on-the-job path is what's under the most strain.