A Plain Memo
AI Isn't Taking Jobs — Except From One Specific Cohort
One week the headline says 300 million jobs are about to vanish. The next week a Yale economist says she can't find any disruption signal at all. Both pieces are sourced. Both are confident. They cannot both be the whole story — so I went to the primary sources and figured out what's actually in the data.
## Chapters
- 00:00 — Two headlines, two stories, same month - 01:19 — What the whole-economy data actually shows - 03:43 — Where the disruption signal actually is (the canary cohort) - 05:49 — Why companies are louder than the data (the AI-washing gap) - 07:35 — How to read the next AI-jobs headline you see
## Sources cited in this episode
- Brynjolfsson, Chandar, Chen — "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence," Stanford Digital Economy Lab (Nov 2025) — the ADP-payroll finding that 22-to-25-year-olds in AI-exposed jobs are down 13–16% relative to peers. - Yale Budget Lab — monthly AI labor-market tracker (Martha Gimbel, lead economist) — "no major macroeconomic effects." - U.S. Bureau of Labor Statistics — 2024–2034 employment projections (net +5.2 million jobs over the decade) and the Monthly Labor Review piece on incorporating AI impacts into projections. - Anthropic — "Labor Market Impacts of AI" (March 2026 Economic Index report) — "no systematic increase in unemployment for highly exposed workers since late 2022." - New York Fed — "Recent College Graduates in the Labor Market" data (5.6% unemployment for grads aged 22–27 vs. 4.2% overall). - Goldman Sachs Global Investment Research — the "300 million jobs exposed" estimate, with Goldman's own caveat that true impact is "likely smaller than their estimates suggest." - Daron Acemoglu (MIT, Nobel laureate, 2024) — AI productivity gain of ~1.1% of GDP over a decade; called Goldman / McKinsey estimates "hyperbolic." - Challenger, Gray & Christmas — 2025 year-end layoff report (54,000 AI-attributed layoffs out of 1.2M+ total — roughly 1 in 22). - Klarna — Sebastian Siemiatkowski on Bloomberg (May 2025) walking back the "AI replaced 700 agents" story; AI service was "lower quality," hiring humans back. - Harvard Business Review (Jan 2026) — the "anticipated capability vs. demonstrated performance" pattern. - IBM — Arvind Krishna on the AskHR agent automating 94% of routine HR tasks, with total IBM headcount rising as savings were reinvested. - Upwork freelance data — translators and content writers lost 20–50% of gig demand; freelancers who used AI to augment their work earned 40% more per hour than peers who didn't.
Plain Memo is a weekly editorial-essay channel on what AI is actually doing — investigated from outside the AI industry, sourced from primary documents, written for working professionals who don't have time for the hype cycle.