The AI jobs story may be 11 million career changes, not 11 million lost jobs

For all the predictions about AI wiping out jobs, one of the more useful numbers in the debate is not a redundancy figure at all.

It is 11 million career changes.

In its new Workforce in motion report, McKinsey Global Institute estimates that around 11 million US workers could need to move out of declining occupations and into different ones by 2035 as AI, automation and wider economic changes reshape demand for labour.

That matters because McKinsey’s base case does not predict a simple shortage of jobs. It estimates that structural growth could create more work overall than automation removes. The problem is whether the people whose current occupations shrink can realistically get into the jobs that are growing.

The jobs may exist. The routes into them may not

McKinsey estimates that automation could reduce demand for work equivalent to around 36 million jobs over the next decade. Growth in the same occupations could absorb the impact for roughly 25 million of them.

That leaves around 11 million workers who may need to switch occupation entirely.

And those moves are not expected to happen at anything like the normal pace. McKinsey estimates that around 770,000 workers a year may need to move between broad occupational groups, about 3.6 times the historical average outside the pandemic period.

That is a very different workforce challenge from teaching everyone how to use an AI assistant.

A customer service worker moving into another customer service role may need new tools and new skills. A worker moving from a declining occupation into healthcare, technical work or another growing field may need a new qualification, different experience, a pay bridge, time to learn and an employer willing to hire for potential rather than an exact previous job title.

The labour market can have vacancies and displaced workers at the same time.

Reskilling is only useful if it leads somewhere

This is where the conversation about AI skills becomes more concrete.

EmEx has already looked at why AI upskilling is becoming an enterprise skills strategy, but this research pushes the question one step further. Skills programmes are not enough if organisations cannot show people where those skills can take them.

For HR, that puts more weight on some decidedly unglamorous infrastructure: job architecture, internal mobility, skills adjacencies, redeployment routes and accessible retraining.

It also raises harder questions. Which roles in your organisation are genuinely likely to shrink? Which roles are likely to grow? What skills overlap between them? Can somebody move sideways without taking a large pay cut? Will managers release good people to another team? Do internal vacancies ask for experience that existing employees have never had the chance to build?

Those questions turn “future skills” from a learning agenda into a workforce planning agenda.

What HR should do with this now

There is plenty of uncertainty in McKinsey’s modelling. The 11 million figure is an estimate, not an observed outcome, and the report gives a range of roughly six million to 16 million depending on how quickly automation is adopted and how much it reduces labour demand.

So HR should not treat 11 million as a headcount forecast. It is better used as a stress test.

Take the roles in your own organisation that are most exposed to automation or falling demand and ask four practical questions:

• Where could those people realistically move next? Name actual roles, not broad skill areas.
• What is the gap? Be specific about the skills, qualifications or experience they would need to make that move.
• Can they build that experience before their current job disappears? That may mean secondments, project work, shadowing or protected learning time rather than another online course.
• Will your hiring system let them through? If internal vacancies still demand an exact previous job title or years of direct experience, the career route exists on paper only.

The useful metric is therefore not simply how many employees have completed AI training. It is how many people can move from roles that are shrinking into roles the business still needs.

If an employer cannot describe those routes today, it does not yet have a reskilling strategy. It has a training offer.