McKinsey Global Institute's September 29 report puts a number on AI workforce disruption: roughly 11 million Americans — 6.5% of the current labor force — will need to switch occupations entirely by 2035. The pace is 3-to-4 times faster than the historical average, at roughly 770,000 workers per year versus the long-run 215,000. AI creates more net jobs than it destroys, but lower-income workers face career disruption at eight times the rate of higher earners.
1
Transition pace: 770K/yr vs. 215K historical
McKinsey's midpoint estimate is 11 million, but the range spans 6M to 16M workers depending on AI adoption pace. At the fast end, AI absorbs administrative and logistics roles within a decade. Historical transitions — industrialization, PC adoption — moved at roughly 215,000 displaced workers per year. AI runs at more than triple that.
2
Growth: healthcare, construction; decline: office work, retail, logistics
Growing roles concentrate in home health aides, infrastructure maintenance, and advanced manufacturing — all requiring physical presence. Declining sectors are office and administrative support, retail and sales, and transportation and logistics. The geographic mismatch is acute: growth jobs are not in the cities where displaced workers live.
3
High-cost states face the steepest structural disadvantage
McKinsey finds high-cost states like California and New York see AI automation adoption above the national average, as high wage floors lower the break-even for automation investment. Low-income workers in these states bear the brunt — facing displacement eight times more likely than higher earners, in states whose policy responses rely on government programs rather than competitive retraining markets.
The problem isn't AI taking jobs. It's that the replacement jobs are somewhere else.
Sources
- ✓ McKinsey Global Institute — AI and the Future of Work 2026 — 2026-09-29
- ✓ Bloomberg — AI May Force 11 Million Workers Into New Jobs McKinsey Says — 2026-09-29
- ✓ Fortune — AI will create more jobs than it kills McKinsey says — 2026-10-03
#AI