When a report says 11 million workers may need to change occupation, the number does the talking and the rest of the sentence gets lost. That is what is happening with McKinsey Global Institute's new report, "Workforce in Motion", published on 29 September. The headline figure is easy to repeat and easy to fear. The finding underneath it is calmer, and more useful to anyone running a business.

First, the numbers. McKinsey estimates that about 11 million US workers, roughly 7% of the workforce, may need to move into a different occupation by 2035. That is around 3.6 times the historical annual rate of occupational change, so it is a real step up in pace. But the same analysis has the US creating around 40 million jobs against about 36 million lost, and finds that 97% of the people who would need to move have a feasible path that holds or improves their pay.

That is movement, not collapse. It is still a lot of movement, and it is worth being plain about the limits of the data. This is a US study, so none of these figures should be read as UK numbers. They are projections to 2035, not measurements of what has happened. And they describe the labour market as a whole, which is a long way from describing your team on a Tuesday.

So what does a business of 20 to 100 people do with it? The most useful move is to shrink the question. A national forecast cannot tell you whether your operations manager's role is changing. Your own work can. Most roles are a bundle of tasks, and AI tends to touch some of those tasks long before it touches the role itself. The sensible question is not "will AI replace this job?" but "which tasks in this job are already moving, and has anyone shown this person how to move with them?"

In practice, that usually looks like a short, honest conversation rather than a programme. Ask each person which parts of their week are repetitive, which parts need their judgement, and which part they would hand over tomorrow if they could. You will find that the answers are specific and rarely alarming. Drafting a first version of a report, summarising a long email thread, tidying meeting notes: these are tasks where a tool such as Claude can take a first pass while the person keeps the decisions. The person's time then moves towards the parts of the job that need knowing the client, the context and the history.

The reason this matters now is the speed point. If change is faster than it used to be, the gap that hurts is not between people who are talented and people who are not. It is between people who have had a chance to practise on their own work and people who have been told "just use AI" and left to it. McKinsey's own framing leans this way: the people who move well are the ones with a visible path and the skills to follow it.

Two things tend to go wrong in smaller firms. The first is waiting for certainty. Nobody can tell you what your sector looks like in 2035, and the report does not pretend to. The second is treating AI as something for the one keen person in the office. That leaves everyone else guessing, and guessing is where the worry grows.

A better first step is small. Pick one real task in one role, agree what good looks like, and let the person practise on it for a few weeks with support. Then look at what changed. It is a modest way to find out how your own business moves, and it does not need a big budget or a new system.

If you take one thing from the McKinsey report, take the proportions. Most people who need to move have somewhere good to go. The job for a business owner is to make sure the people in their own team are not finding that out alone.

Go wisely.