Stanford HAI · Stanford Institute for Human-Centered AI · 2026-05-16
Read the Economy chapter on hai.stanford.edu
The Stanford AI Index 2026 Economy chapter is the canonical reference for AI's measurable workplace effects this year — the document everyone else's "AI and jobs" essay is implicitly citing or implicitly arguing against. A few findings deserve to be lifted out and named concretely so readers know what to look up.
Software developers ages 22–25 fell ~20%. Employment in this cohort declined roughly 20% from late 2022 through mid-2025, while the over-25 cohort in the same occupation grew. This is the most-cited single data point in the chapter, drawn from the St. Louis Fed / Brynjolfsson-team paper and replicated against BLS microdata. It is the cleanest signal of AI-driven entry-level displacement on record, and it is concentrated — not economy-wide.
BLS-flagged AI-exposed occupations lost 0.2% of employment. The 18 occupations the Bureau of Labor Statistics has identified as most AI-exposed (≈10 million jobs — paralegals, customer service reps, technical writers, bookkeepers, translators, market research analysts, others) lost 0.2% of employment between May 2024 and May 2025. Small in aggregate; the first year the AI-exposed group has underperformed the broader economy in an annual series.
Agentic AI postings: +10,854% year over year. Job postings mentioning agentic-AI skills grew 10,854% from 2024 to 2025, reaching roughly 90,000 US postings (0.23% of all postings). AI-related skill mentions are now in 2.5% of all US job listings, a 297% increase over the prior decade. Standalone "prompt engineer" postings are down 30–60% — the skill has been absorbed into AI Engineer, LLM Engineer, and Applied ML roles.