The AI jobs panic is mostly wrong, with one exception that is very real. At the macro level, unemployment in AI-exposed jobs is not spiking. But the first rung of the ladder — the entry-level job that used to pay you to do the repetitive work — is the part AI actually took.
That is the signal under the noise. The crisis is not "AI is taking all the jobs." It is "AI took the tasks that used to be how you got your first job." For an AI-fluent person without a traditional pedigree, that is the opening, not the threat.
The claim: the bottom rung is the one AI took
The entry-level ladder broke before the overall job market did, and a new on-ramp is forming in its place.
MIT Technology Review called the broad AI-jobs panic overblown in May 2026 — unemployment in the most AI-exposed occupations is actually lower than in less-exposed ones. But the same reporting flagged a sharp exception for young workers, and that exception is the whole story for anyone breaking in.
The evidence: who is actually getting squeezed
The pain is concentrated at the start of careers, not across them.
The Stanford finding behind that 16% number is the anchor: a measurable employment drop for 22-to-25-year-olds in exposed occupations, with no matching drop for older workers in the same fields or for younger workers in safer ones. The companion MIT piece on the looming crisis in entry-level work is blunt about the mechanism: the first job is where you learned by doing the small, repetitive tasks, and those are exactly the tasks AI now does.
Fortune framed it as an experience gap: employers still want experienced people, but the jobs that used to manufacture experience are thinning. On Hacker News, Sean Goedecke's essay arguing that software engineering may no longer be a lifetime career drew 764 comments — most of them from mid-career people watching the bottom of their own field hollow out.
And graduating into it feels exactly as grim as it sounds. PBS NewsHour spent a commencement-season segment on the class of 2026 entering "a world being transformed by AI," and the anxiety in it is not abstract. It is people who did everything right and found the first rung missing.
Why the first rung disappeared first
Entry-level work was always a bundle: do the boring tasks, and in exchange learn the judgment that makes you senior. AI unbundled it.
The boring tasks — first-draft research, basic tickets, data cleanup, boilerplate, summarizing, simple builds — were the training wages. They were also the most automatable work in any role. So the tasks left first, and the learning that used to come attached to them left with them. The senior people are fine. The on-ramp is what broke.
It is worth being honest about the counter-argument, because it is strong. The Financial Times asked whether remote work, not AI, is the real reason junior hiring is weak — when nobody is in the office, the cost of training a junior who needs constant guidance goes up. That thread drew 369 comments because it is partly right. The honest read is that AI did not act alone. It removed the tasks while remote work removed the cheap mentorship, and the two together cut the rung.
But the diagnosis does not change the move. Whether AI or remote work or both took the rung, the response is the same: stop waiting for an employer to hand you the experience, and manufacture proof of it yourself.
The on-ramp that replaces the ladder
The rung that broke was tied to one path: get hired junior, climb. The on-ramp that is replacing it grades you on outcomes you can show, not years you have served.
This is the part that favors the AI-native applier. Most of the work types on this board are judged on artifacts, not seniority: a shipped app, a running workflow, a clean eval, a client system that went live. You do not need permission or a first employer to produce those. You need fluency and one real project.
Three on-ramps that do not require a first employer to hand you the experience:
- Build something and ship it. Use AI coding tools to put a real product live, then write the judgment behind it. This is Build Products with AI work, and the bar is a working thing, not a clean codebase.
- Automate a real process end to end. Take a workflow you understand, rebuild it with AI in the loop, and run it for a month. An AI automation portfolio piece is the artifact that opens Automate Workflows with AI roles.
- Prove an AI skill on a real task. Rebuild one task you already do with an AI step, measure the before and after, and publish it. That is what employers mean when they ask for "AI skills" — applied judgment, not a certificate.
The common thread is the same one the broken ladder used to provide: judgment, earned by doing. The difference is you produce it in public, on your own schedule, instead of waiting for a junior seat that may not come back.
- Pick one real task or product you understand. Narrow beats ambitious.
- Build or automate it with AI tools. Ship the smallest version that actually runs.
- Run it on real work for two weeks. Log what broke and what you fixed.
- Publish a one-page teardown on your own domain: the before state, what AI does, where you reviewed it, and one metric that moved.
- Put that link above your résumé. The artifact is the experience now.
The takeaway
The data is real: the first rung is the one AI took, and pretending otherwise helps no one. But the conclusion is not "wait it out." The ladder was a permission structure, and AI fluency lets you route around it. Ship the proof the junior job used to give you. The on-ramp is open to whoever can already apply the tools.