Anthropic Sharply Cuts Entry-Level Hiring as AI Agent Workload Surges

Anthropic published internal research showing that since early 2025, the share of entry-level roles in its technical division fell from 21% to 14%; among new hires, the junior share dropped from 27% to 9%, junior roles still being hired fell to 6%, while non-technical divisions stayed roughly stable. By August 2026, its AI agents logged six times the work hours of human employees, reaching 19:1 in the research division; since March 2026, agent work hours have exceeded human ones. Coding, bug-hunting and similar tasks are increasingly handed to Claude. On productivity, engineers' code commits rose to 5.1x, data scientists' query volume to 3.1x, and legal contract throughput to 3.1x; sales closed deals at 1.7x, recruiting completed hires at 1.4x, and accounting processed invoices at 1.4x. Entry-level roles shrank most where Claude takes on the most complex work. The study itself concedes Anthropic is an extreme case whose results don't extrapolate directly to the broader economy, and rapid business growth may distort the figures. Nolan Lovett, a researcher at the NATO Special Operations University, raises a "tragedy of the cognitive commons": firms gain efficiency by replacing entry-level roles with AI, while society bears the cost of developing senior talent, with effects potentially only visible between 2030 and 2045.