AI Productivity Gains Bring Hidden Costs

New research finds AI-generated “workslop” is widespread: about 40% of workers received it in the past month, companies spend roughly two hours per incident to clean it up, and it costs about $186 per employee monthly (≈$9 million annually for a 10,000-person firm). In software, AI coding assistants produce a mixed productivity effect — some developers complete tasks about 26% faster while others take 19% longer — as AI introduces correction cycles, technical debt and security flaws. Security analysis from Veracode finds roughly 45% of AI-generated code contains vulnerabilities. The surge in AI-driven expectations is also changing Silicon Valley workplace culture, with increased weekend and after-hours activity resembling 996 norms that disproportionately burden caregivers and women founders. Economists warn advanced AGI could allow overall prosperity to rise while labor’s share collapses and income concentrates with owners of compute infrastructure unless ownership and policy models change, and the booming market for AI productivity tools risks reinforcing low-value outputs without careful oversight.