AI hasn't destroyed jobs yet — it's quietly raising the bar to keep one
A year after CEOs warned AI would wipe out half of entry-level jobs, unemployment data shows little mass displacement — but the nature of work is shifting fast.
What happened: In 2025, Anthropic's Dario Amodei predicted AI could wipe out half of entry-level white-collar jobs, and OpenAI's Sam Altman warned entire job categories would disappear. A year on, that mass carnage hasn't materialized. A Stanford Institute for Economic Policy Research analysis found unemployment among the most AI-exposed workers rose 0.77 percentage points since 2022 — actually less than the 0.85-point rise among the least-exposed workers. Recent graduates do have higher unemployment (5.6% vs. 4.2% nationally), but researchers say overhiring hangovers and remote-work shifts are likely bigger factors than AI alone.
Why it matters: The absence of a jobs crash doesn't mean nothing is happening. Employers are quietly raising expectations instead of cutting headcount outright. ZipRecruiter found 74% of employers now see AI skills as a strong advantage or requirement, and half expect new hires to already be competent AI users on day one. Labor economist Nicole Bachaud calls this 'a rising bar rather than a shrinking pool' — meaning the real risk for workers isn't a vanishing job market, it's getting screened out by skills employers assume you should already have.
How it's playing out, plainly: NYU's Robert Seamans splits AI's effect into three buckets: jobs made obsolete, jobs created, and jobs changed — and says the third is by far the largest. At coding platform Bolt.new, a three-person analytics team built an AI agent that now does work equivalent to a 30-to-40-person team, saving 12-13 hours weekly, says CEO Eric Simons. Stanford's Nicholas Bloom describes this churn as 'turbulence': some roles disappear, new ones appear to build and maintain AI systems, and most existing jobs simply get restructured around what the tools can now do.
The rollout: MIT's Paul Osterman, author of 'Disposable Workers,' warns the bigger shift may be toward contract and freelance labor as companies figure out which skills they actually need long-term — leaving more workers without a stable career ladder. He estimates 35% of the US workforce is already easily replaceable, a share AI could grow. Workers are responding by pushing AI-use clauses into union contracts, says TechEquity's Tim Newman. Full effects will take years to surface, and could slow further if political backlash over AI's costs grows after the midterms, Bloom says.