A new McKinsey Global Institute report, published on 29 September, has put hard numbers on the question everyone in the AI-and-jobs debate keeps circling: how many people will actually have to change careers? Its base estimate — 11 million US workers, about 7 percent of the workforce, may need to switch occupations entirely by 2035.
But the same report’s other numbers point the other way, and that tension is the story.
The headline numbers, in both directions
McKinsey’s model estimates automation could reduce US labour demand by the equivalent of roughly 36 million jobs by 2035. Over the same window, it projects growth in AI-related fields and the wider economy generating demand for more than 40 million jobs. Net effect: more jobs, not fewer — roughly 41 million jobs created against 36 million reduced, in an economy that will also be short of workers as the population ages.
“The next decade’s challenge is mobility, not scarcity,” the report states. About 25 million of the workers whose roles get touched by automation are expected to stay in their occupations because industry growth offsets the pressure. The remaining 11 million — a range that stretches from 6 million to 16 million depending on adoption speed — would need to retrain into different work: what McKinsey calls “the largest and most sustained workforce transformation in US history.”
Where the burden lands
The distribution is anything but even. More than 75 percent of workers needing to switch occupations are concentrated in just three fields: office and administrative support, retail and sales, and transportation and logistics. Lower-wage workers are 7.6 times more likely than higher-wage workers to face a full occupation change, and workers without a degree are about 1.8 times as likely as bachelor’s graduates to need to move. Meeting the shift would take roughly 770,000 workers changing occupational groups every year — about 3.6 times the historical average.
There is a brighter side in the destination data too: McKinsey projects roughly 60 percent of growing employment in the top two wage quintiles, concentrated in healthcare, construction and management. Growth, in other words, is projected to happen mostly in better-paid work — if people can reach it.
The pathway problem is the real story
Whether those 11 million transitions happen is where the report gets uncomfortable. Every worker is modelled as having some route into growing occupations, but only about one in seven has what McKinsey calls a direct pathway — minimal retraining, no wage loss. About four in ten face a “winding” pathway requiring meaningful reskilling or some pay cut, and nearly half face an “unpaved” one with large skill gaps or blocked credentials. The single biggest obstacle is paper: roughly 85 percent of growing jobs require some form of credential or certification.
Demand for the moving skills is already visible in postings. McKinsey counts an 11-fold increase in demand for AI fluency between 2022 and 2026, a fivefold rise in adaptability, and a tripling in resilience, curiosity and willingness to learn.
The timing, and what it means for New Zealand
The report landed into a US labour market already in a two-year “low-hire, low-fire” stall. Bureau of Labor Statistics data released the same day showed job openings at a five-month low, voluntary quits — the standard gauge of worker confidence — near a six-year low, and layoffs falling for a second straight month. The Conference Board’s confidence index dropped to 81.9 in September, a 12-year low.
For New Zealand, the transferable idea is not McKinsey’s US-specific numbers but its framing: the risk in this transition is not mass joblessness, it is blocked mobility. The same structure applies here — an ageing workforce, growth clustered in healthcare and construction, and admin-heavy roles facing the same automation pressure. New Zealand’s own SEEK data shows job ads requiring AI skills up 93.3 percent year on year, so the skill-shift dynamic is already priced into local listings. The policy question McKinsey effectively raises for any country is whether credential barriers and training pathways move fast enough for the workers who need them.
McKinsey is careful to label its output a model, not a measurement — scenario modelling, not a forecast. But as a sizing of the problem, it does something the doomscrolling doesn’t: it separates the jobs that vanish from the workers who need a bridge, and points the debate at the bridge.
FAQ
Will AI create more jobs than it destroys? McKinsey’s base case says yes in the US through 2035 — about 41 million jobs of demand created against 36 million reduced by automation. Other researchers disagree, and the range depends heavily on how fast AI is adopted.
How many workers will need to change careers because of AI? McKinsey estimates 11 million US workers (6-16 million in the range) may need to switch occupations entirely by 2035, with a further 25 million staying in their occupation as industry growth absorbs the pressure.
Which occupations are most exposed? In the McKinsey modelling, more than 75 percent of workers needing to change occupations sit in office and administrative support, retail and sales, and transportation and logistics.
Is there an equivalent study for New Zealand? MBIE’s AI-exposure mapping of NZ roles and SEEK’s NZ AI-skills job-ad data cover part of the same ground; no McKinsey-scale occupational mobility model exists for NZ yet.
— CJ Murden, editor of Singularity.Kiwi. Former digital technologies teacher, author of AI-focused books. Writing with a New Zealand focus.
