Jobs OpenAI once said would be safe are being crushed four times faster, report says

Fast-forward: the disruption arrived early

What was meant to be a decade-long adjustment is happening in months. It has been reported that U.S. fintech firm Block cut roughly 40% of its staff — about 4,000 people — as it pivots “to become an AI company,” a move that sent its shares sharply higher and signalled how quickly firms are betting on automation over labor. That single case is emblematic: technologies that in 2023 looked like “screen-bound” language models are now being paired with vision, tool use and agentic capabilities that remake jobs far faster than researchers expected.

The numbers behind the acceleration

In March 2023 OpenAI published a paper estimating that roughly 19% of U.S. workers would see more than half their tasks affected by generative models over a ten‑year horizon. It has been reported that a 2026 update from IT services giant Cognizant now finds that the same shifts are unfolding four to five times faster: annual AI exposure growth rose from about 2% to 9%, and roles with more than 50% task exposure have jumped to 30% (versus a prior 2032 projection of 15%). Cognizant estimates this represents roughly $4.5 trillion in U.S. labor cost potentially shifted to AI — about 15% of GDP. Meanwhile, Stanford’s Digital Economy Lab reportedly found entry‑level hiring in high‑exposure sectors down 18–40% while demand for senior talent rises.

Who loses and who wins?

The pattern is not simply job loss but structural collapse: entry rungs close, middle layers shrink, and a small cohort of “AI operators” capture value. Finance managers top the list with 84% of tasks judged transferable to AI; computing and math roles, legal, operations and management are likewise heavily exposed. It has been reported that internal accounts at Anthropic show most code is now auto‑generated by their models. At the same time, purely manual crafts — bricklayers, slaughterhouse workers, dishwashers — remain among the least exposed, for now. Even water‑and‑field trades are changing: multimodal AI that “sees” a leak, diagnoses causes, orders parts and drafts reports restructures rather than eliminates the plumber’s work.

Policy and geopolitical context

This accelerated wave arrives amid broader tech geopolitics: export controls, chip sanctions and national AI strategies will shape who deploys advanced models and at what speed. If the Cognizant numbers are borne out, the social and economic stakes — from reskilling to unemployment insurance to competition policy — become urgent. Who benefits and who pays for the transition? Governments and companies must decide quickly whether to steer this disruption or simply ride it.