The AI job displacement risk is no longer a theoretical worry for economists to debate at conferences. It is showing up in employment data, in hiring statistics, and in the changed behaviour of the world’s largest companies, with younger and entry-level white-collar workers bearing the sharpest end of it.
The figures that matter most come from Stanford University‘s analysis of wage and employment data. For 22 to 25-year-olds overall, employment has fallen 2.7% since ChatGPT became widespread, rising to 12.8% in the most AI-exposed sectors: finance, software, and creative industries. Zoom in on software developers specifically, and the picture sharpens considerably. According to TechTimes, citing the Stanford HAI 2026 AI Index, employment among software developers aged 22 to 25 fell nearly 20% from its 2024 peak. Developers aged 30 and older at the same companies, meanwhile, saw employment grow between 6% and 12% over the same period.
That gap is not noise. It is a structural shift in who gets hired when a machine can do the introductory work.
Entry-level roles are disappearing across knowledge industries
The same Stanford data shows entry-level job postings across knowledge-economy and professional fields fell approximately 35% from January 2023 to late 2025. That is not a blip caused by interest rate rises, though some economists make that case. The OECD’s own analysis of job postings found that highly AI-exposed sectors, including telemarketing and legal services, shed postings at a materially faster rate than less-exposed ones like construction, cleaning, and food preparation. The UK’s service-sector concentration left it disproportionately exposed on that measure, and the trend predated last year’s National Insurance rise.
The OECD’s 2025 report on AI transitions puts the systemic scale bluntly: almost 40% of global employment could be exposed to AI disruption. In advanced economies, that figure rises to approximately 60%, driven by the concentration of cognitive, knowledge-based work. IT services, media, and telecommunications rank highest across all measured dimensions of AI intensity: human capital, innovation, exposure, and usage.
By contrast, the OECD Employment Outlook 2023 found that food preparation assistants, agricultural labourers, and cleaners remain among the least exposed to AI progress. The irony writes itself: the jobs that pay least and carry least status are, for now, the ones AI cannot easily take.
The economists who changed their minds on AI job displacement risk
What lends the current alarm more credibility than previous rounds of automation anxiety is who is raising it. Daron Acemoglu and Simon Johnson, the MIT professors who shared the 2024 Nobel Prize in economics, spent years pushing back on what they considered AI displacement hype. They are now among those urging governments to act urgently to ensure that AI delivers rising living standards rather than large-scale unemployment. When the sceptics reverse course, it is worth paying attention.
Goldman Sachs estimated in April 2026 that AI is eliminating roughly 25,000 US jobs per month. That is not a forecast; it is the bank’s read on what is already happening.
The counter-argument, and it is a real one, is that AI agents may prove more expensive than the human workers they replace. Several large companies that deployed internal token usage tracking, pushing employees toward the most advanced models, found the bills unsustainable and began rationing access. ‘Flat is the new up’, in workforce terms, may be a transitional phase as much as a permanent restructuring. And the rapid spread of cheaper AI tools derived from Chinese models, provided freely to the market, complicates any clean analysis of cost and displacement.
But the data on young workers is hard to dismiss. If the first cohort to enter the workforce after ChatGPT’s launch is already seeing employment fall nearly 20% in software and double-digit percentage-point hits across AI-exposed industries, the question is not whether AI is disrupting entry-level work. The question is how far up the career ladder that disruption travels, and how quickly. The next round of Stanford employment data, covering 2026 hiring, will be the one to watch.
