Eleven Million Americans May Need a Different Job by 2035

Roughly 11 million US workers, about 7% of current employees, may have to move into an entirely different occupation by 2035, according to a McKinsey Global Institute report published on 29 September 2026. Four hours before it appeared, the Bureau of Labor Statistics reported that the quits rate, the closest official proxy for workers actually moving, had held at 1.9% in August. The two numbers describe the same decade and point in opposite directions.

The report, "Workforce in motion: Skills and pathways to future jobs in the United States", models labour demand from 2025 to 2035 using the Bureau of Labor Statistics occupational classification mapped onto the roughly 1,800 occupations in the Lightcast taxonomy, and expresses changes in full-time equivalents against a 2025 baseline. Its authors are explicit that these are "model outputs under a base set of assumptions, not measurements of observed change". The headline arithmetic: automation could reduce labour demand by the equivalent of about 36 million jobs, while growth in the AI value chain and the broader economy could generate demand for more than 40 million. Around 25 million of the 36 million would stay in their occupation because growth in it offsets the automation. The residual 11 million would not.

Where the hours go

The exposure is concentrated, not general. McKinsey's estimate of expected average adoption by 2035 reaches roughly 80% of current work hours in office and administrative support, 70% in technology and analytics occupations and 68% in retail and sales. Healthcare professionals sit at 26% and public safety and security at 28%. The gap between the top and bottom of that range is the whole story of who has to move.

Absorption is not the same as displacement, and the report is careful about the difference. It estimates that automation technologies could take on about 54% of current work hours across the economy by 2035, that organisational and market mechanisms would offset roughly 60% of the resulting labour impact, and that the net effect is a reduction in labour demand equal to about 21% of current work hours. That chain of three numbers, each an assumption, is where most of the uncertainty in the 11 million figure lives.

Parallax Nexus has reported on the two data-heavy studies from August 2026 that found no economy-wide AI job loss but a widening hiring gap for early-career workers in exposed occupations, with call-centre employment between 27% and 39% below trend across the United States, Canada and Germany. McKinsey's model is consistent with that: it does not forecast mass unemployment. It forecasts a sorting problem.

The constraint is the turnstile, not the total

Moving 11 million people between occupational groups over a decade implies about 770,000 such moves a year, which the report puts at roughly 3.6 times the historical average. The destinations exist on paper. McKinsey counts 16 million new openings in growing occupations against the 11 million jobs facing reduced demand, a surplus of about five million posts, with growth concentrated in healthcare, professional and technical services and construction, and the largest declines in retail trade and in accommodation and food services. The report's own formulation is that "the next decade's challenge is mobility, not scarcity".

Which is exactly what the morning's official statistics say is missing. The Bureau of Labor Statistics put job openings at 7.079 million at the end of August, down from a revised 7.335 million in July, with hires at 5.192 million, quits at 3.066 million and layoffs and discharges steady at 1.6 million. CNN, reporting both releases on 29 September, described openings as a five-month low and the quits rate as near a six-year low, with layoffs shrinking for a second consecutive month. That is a labour market in which very few people are being let go and very few are choosing to leave. The judgment worth stating plainly is that a model requiring 3.6 times the historical rate of occupational switching is being published into a market running well below the historical rate, and nothing in the report explains what would reverse that.

Two things temper the alarm. McKinsey's own range is wide enough to matter: faster adoption or a larger demand effect would push transitions above 16 million, slower adoption would hold them near six million, a spread of nearly three to one. And the pay news is better than the mobility news, with the report finding that only 3% of transitioning workers, roughly 305,000 of the 11 million, would follow a pathway requiring a pay cut, an estimate that holds occupational wage levels constant and so cannot capture what happens to wages as several hundred thousand people a year arrive in the same growing occupations.

What the report does not attempt is a path. It says so: these estimates describe the long-run balance between declining and growing labour demand, not the route the labour market takes to reach it. The next observable test is narrow and soon. If AI is already pushing workers between occupations, it should show up first as a rising quits rate among the office and administrative support occupations McKinsey puts at 80% exposure. In the August data, it has not.