Entry Level, Three Years of Experience Required

Stanford's Digital Economy Lab found that employment for 22-to-25-year-olds in AI-exposed jobs fell 16 percent relative to others, while a survey of about 1,000 hiring managers showed 31 percent raised experience requirements for entry-level roles. However, a June study of over 20,000 U.S. firms found that the heaviest AI investors grew entry-level headcount by 12 percent in two years, with the threshold at about $30 per employee per month.

Artificial Intelligence /us/basics/artificial-intelligence Entry Level, Three Years of Experience Required I said AI was eating entry-level work. However, it depends on the employer. Posted August 22, 2026 Reviewed by Lybi Ma /us/docs/editorial-process You have seen the posting. You may have approved one. It reads like a clerical error, and online it gets passed around as a joke. It is neither. That line is the visible edge of a decision thousands of organizations made over the past two years, almost none of them on purpose. I have argued here that AI is climbing two ladders at once — from routine tasks toward judgment, and from junior roles toward senior ones — and that the casualty is the entry-level job that has trained two generations of experts. Every senior anything, I wrote, was once a junior something. Then I spent the summer with the data. Some of it backed me. Some of it took me apart. The second half is more useful. What Held Up Researchers at Stanford's Digital Economy Lab, using payroll records of millions of American workers, found that employment for 22-to-25-year-olds in the jobs most exposed to AI — software development, customer service — fell 16 percent relative to everyone else. Older workers doing the same jobs were fine. The young absorbed it. The declines showed up almost entirely where AI replaces a person's work rather than assisting it. Where AI mostly helps, employment grew. Hold onto that. Employers confirm it themselves. In a survey of about a thousand hiring managers this summer, 31 percent said they had raised experience requirements for entry-level roles. Nearly a third watched AI absorb the work juniors used to do, then required that juniors arrive already experienced. That is not a labor market adjustment. That is a catch-22 with a job posting attached. What Didn't Hold Up Start with a number I have used myself. I have been citing rising unemployment among recent college graduates as evidence of AI displacement https://www.psychologytoday.com/us/basics/displacement . It mostly isn't. Graduates lost their unemployment advantage back in 2018, years before ChatGPT existed, and economists who have taken the increase apart find nearly all of it comes from more graduates entering and searching, not losing jobs. Worse for my argument: When Stanford's economists sorted occupations by how much AI touches them, unemployment since 2022 rose slightly more in the least-exposed jobs than in the most-exposed ones. Then came the finding that rearranged the question. In June, researchers matched company AI spending against workforce records at more than 20,000 American firms. The heaviest investors did not shed junior workers. Their entry-level headcount grew 12 percent in two years — faster than those firms grew overall. The dabblers showed no change. The threshold separating the two groups was about $30 per employee a month. The cost of one license per person. That is not a budget. That is a decision. The Fork Nobody Puts on the Agenda Young workers are losing ground in the most automatable jobs , gaining ground at the most committed companies . Both are true, but measuring different things. Occupation is not destiny here; what an employer chooses to do matters more than what the technology can do. AI absorbs the routine content of a junior role, and someone decides what that role now is. See it as a bundle of tasks: the tasks are gone, and cutting https://www.psychologytoday.com/us/basics/self-harm the position is rational. See it as a stage in how a person becomes good at something, and the tasks were never the point; the work is rebuilding a job that is the same person on the other end. Almost everyone taking the first path believes they are making an efficiency decision. They are making a succession decision, 15 years early, without ever putting it on an agenda. My field has spent 50 years on this fork. People grow into competence through variety, real responsibility, and correction from someone further along — all three of which can be made into a job deliberately. And raising a hiring requirement instead of building a pipeline moves your training costs onto your competitors. That works until everyone does it at once, which is roughly where we are. The 31 percent are not solving a hiring problem. They are misdiagnosing a training problem as a hiring problem. It shows up in the same survey: Only 22 percent of employers require AI training for everyone, while 57 percent have raised expectations for how much people produce. Raise the bar, supply no ladder. No mystery why the juniors are struggling. Artificial Intelligence https://www.psychologytoday.com/us/basics/artificial-intelligence Essential Reads Rebuilding the Rung Three moves, if you would rather be in the group that grew. Make the junior job about judgment, not volume. The old version produced raw material for a senior person to edit. The new one checks, corrects, and owns what the machine produces. That builds the person faster, but only if a senior human reads the work, which is the cost most organizations bought AI to avoid. Pay for the training you already require. If AI fluency is a condition of employment, it is a core competency, and core competencies get budgeted. An optional lunch-and-learn is not a strategy. Count how many people are being taught. The number that matters for 2041 is not headcount. It is how many are learning from someone further along. Almost no one counts it, and once you have the number, you cannot unsee it. And if you are 22, you are shopping less for a salary than for an employer that still knows how to build a person. Ask out loud in the interview. Who reads my work, and how often? What will I be able to do in two years that I can't do today? An employer with a real answer is in one dataset. I was right that the rung is under threat. I was wrong to think it was disappearing on its own. It is a choice — one most organizations make by default, in a budget meeting, without ever saying they have decided to stop growing people. References Brynjolfsson, E., Chandar, B., & Chen, R. 2025, November 13 . Canaries in the coal mine? Six facts about the recent employment effects of artificial intelligence. Stanford Digital Economy Lab. Mahoney, N., McEntarfer, E., & Wahal, S. 2026, July . What is really happening to jobs? Separating AI hype from reality. Stanford Institute for Economic Policy Research. Economic Policy Institute. 2026 . Class of 2026: Young college graduates face a weaker labor market, but a more mixed picture than the headlines suggest. Kharazian, A., Simon, J., & Stevens, L. 2026, June 30 . A new look at AI's impact on jobs: Firm-level AI spending and workforce adjustment. Ramp Economics Lab & Revelio Labs. ZipRecruiter. 2026, July 29 . More jobs, higher bar: The 2026 AI employer report.