White‑collar hiring is softening where AI is strongest
New analysis shows tech layoffs have surged this year and academic work suggests AI has prevented the creation of roughly 500,000 coder jobs, signalling weakening vacancy and entry‑level hiring rather than just headline redundancies. The result is a more subtle labour shock: fewer posted roles and slower promotion ladders, especially for mid‑skill, language‑heavy work. (techradar.com) (nbcnews.com)
Tech layoffs are still happening in public, but the quieter shift is in jobs that never get posted in the first place. Layoffs.fyi showed 71,447 tech employees laid off across 80 tech companies by April 10, 2026, and new labor research says the slowdown now reaches beyond pink slips into hiring itself. (layoffs.fyi) (federalreserve.gov) One reason economists watch coding first is that coding assistants are one of the earliest mass-market uses of large language models, the text-predicting systems behind tools like ChatGPT and Claude. A Federal Reserve paper released on March 20, 2026 says coding-intensive employment kept growing after November 2022, but “much more slowly” than before ChatGPT arrived. (federalreserve.gov) That same Federal Reserve paper estimates the gap is now big enough that roughly 500,000 additional coder jobs would likely exist without large-scale large language model use. That is not 500,000 people fired on one day; it is closer to a road that was supposed to get built and then never was. (federalreserve.gov) (nbcnews.com) The first workers getting squeezed are often the newest ones. Stanford researchers using payroll data from the largest payroll software provider in the United States found that workers ages 22 to 25 in the most artificial-intelligence-exposed occupations saw a 13 percent relative decline in employment after generative artificial intelligence spread widely. (siepr.stanford.edu) The Stanford team found the drop was concentrated in jobs where artificial intelligence is more likely to automate work instead of helping a worker do more of it. In plain terms, the pressure is strongest where the job is built from repeatable screen work, not from managing people, owning client relationships, or doing physical tasks on site. (siepr.stanford.edu) Job-posting data shows the same split from another angle. Harvard Business School reported in March 2026 that openings for routine, automation-prone white-collar roles fell 13 percent after ChatGPT’s debut, while demand for more analytical, technical, and creative roles rose 20 percent. (hbr.org) That helps explain why the labor shock feels strange on the ground. A company can say headcount is stable while still cutting the bottom rung, posting fewer junior roles, and asking one mid-level employee with an artificial intelligence assistant to do work that used to be spread across two or three people. (hbr.org) (nbcnews.com) The pattern is not economy-wide collapse, at least not yet. A February 2026 National Bureau of Economic Research survey of nearly 6,000 executives found 69 percent of firms are actively using artificial intelligence, but nine in ten said it had not yet changed employment at their own firms over the prior three years. (nber.org) Those same executives still expected artificial intelligence to cut employment by 0.7 percent at their firms over the next three years while lifting productivity by 1.4 percent. That is the shape of the shift: not one giant layoff wave, but a slower squeeze in the exact white-collar jobs where language models are already good enough to handle drafts, summaries, code, support tickets, and back-office paperwork. (nber.org)
