Could AI create jobs while leaving workers behind? McKinsey models the US transition
A 29 September McKinsey Global Institute report models occupational moves through 2035. Its scenarios are planning assumptions, not observed redundancies or a forecast for an individual worker.
The short version
- What happened
- The scenarios concern US occupational transitions through 2035, not measured layoffs.
- Who it affects
- Workers need accessible pathways and income support, not just predictions about expanding occupations.
- What remains unproven
- Model assumptions are uncertain; job postings do not measure completed worker transitions.
Keep in mind
Model assumptions are uncertain; job postings do not measure completed worker transitions.
Results cannot be transferred directly to other national labour markets.
Source analysis · 1 linked source. Primary analysed source date: 29 September 2026.
Read the full report · 4 min The evidence, methods, implications and limitations
What the new report actually estimates
McKinsey Global Institute published its US workforce study on 29 September 2026. It estimates that about 11 million workers in declining occupations could need to change occupations by 2035, with a scenario range of six million to 16 million. The central estimate represents roughly 7% of current US employees. It is a model output, not a count of people already displaced.
The analysis combines occupational employment data, automation assumptions and demand scenarios. It uses Lightcast job postings to compare skills and advertised pay, alongside credential and training requirements. The report considers four transition dimensions: skills, credentials, training time and wages. A language model helped synthesise qualification requirements. There is no single survey sample that makes these projections representative observations of future employment.[1]
Our analysis: a growing market can still be hard to enter
The practical question is whether a worker can reach the new opportunity. Imagine an administrator offered a route into a growing healthcare occupation. The destination might have vacancies and better pay, yet require years of study, supervised practice and a licence. A household needing this month's income cannot treat that vacancy as an immediately available replacement. This hypothetical example illustrates why a net employment total tells only part of the story.
A useful local response would start with a transition budget, not a course catalogue. How many paid learning hours can the employer provide? Can the worker retain benefits while retraining? Is the destination within commuting distance, and are employers actually hiring entry-level applicants? A pathway that looks close on a skills map may remain inaccessible because of childcare, transport or the cost of an unpaid placement. These questions are our interpretation of what a credible transition programme should test.
What workers and employers can measure
For a worker, the most informative exercise is to separate changes within a job from a move to another occupation. Keep a short record of tasks that disappear, tasks that expand and new checking responsibilities. Ask an employer which changes are already funded and which are exploratory. That provides a basis for discussing training without treating a national scenario as a personal probability of redundancy. It also avoids assuming that familiarity with a chatbot is equivalent to a recognised professional qualification.
Employers evaluating a transition programme should follow the whole group offered support, including people who cannot enrol or who leave early. Count sustained placements, earnings after transition and the time taken to find work. Record the denominator for each measure. Reporting only successful graduates would conceal whether the programme works for staff with the greatest barriers. In a pilot, comparison with a similar group receiving existing support would make the findings more useful than a collection of selected success stories.
Limits, global relevance and the next evidence
The report is US-specific and produced by a consultancy with commercial interests in organisational change. Its assumptions can be examined, but its publication is not independent proof that those outcomes will occur. Our assessment would change with observed transitions, earnings and unemployment durations that either support or contradict the model, particularly if results were reported separately by occupation, age and region.
Readers elsewhere should use the questions rather than import the numbers. Qualification routes, employment protections and access to education differ between countries. A UK employer would need UK labour-market and licensing evidence before applying a US pathway. For policymakers, the most useful next step would be to publish the cost and distribution of transition support alongside any headline jobs estimate. The public benefit should be judged by whether people can secure sustainable work, not simply by the size of a projected market.[1]