Which AI economy is your GCC's leadership hiring built for?

In September 2026 Anthropic’s economics team published Scenarios for our Economic Future, an interactive model of how AI might change jobs, growth and unemployment in the US up to 2030. Anton Korinek, Chad Jones, Szymon Sacher, Tess Cotter and Peter McCrory built it, and the method is set out in their technical report, Economic Scenarios for Transformative AI. Anthropic sells the technology it’s modelling, and we read it with that in mind.

The starting idea is the most useful part. Every job is a bundle of tasks. Their example is a nurse, who draws blood, triages patients, charts vitals and orders supplies for the ward. AI can help with a task, take it over completely, leave it alone, or create a task that didn’t exist before. AI can’t bathe a patient, it might well chart her vitals, and somebody now has to check how well the AI triages.

The model then asks what happens to the whole economy under three levels of AI capability and adoption. In the modest one, AI does roughly what the internet did. In the extreme one, AI does nearly all knowledge work better than people, growth reaches 15% a year and the economy doubles every 4.5 years. Unemployment rises beyond typical recession levels, and knowledge workers’ wages fall by more than 10% by 2030. Anthropic says that would probably need AI that improves itself, adopted fast.

The middle scenario is the one we keep coming back to. Anthropic calls it substantial. AI is capable of half of all knowledge work by 2030, but most knowledge work tasks are still done without it, because companies don’t adopt everything they could. The economy grows at twice its normal rate, ending 2030 8.3% larger than the same economy without AI. Wages for knowledge workers don’t rise, and everyone else’s do. By 2030 knowledge workers are paid 0.3% less than they would be without AI, and all other workers 5.9% more.

That gap between what AI can do and what a company has adopted is where a global capability centre (GCC) lives for the next few years. Somebody inside the centre decides which half gets adopted, how fast, and what happens to the people doing it today. (Anthropic also surveyed more than 10,000 Americans in August, and their typical answers landed close to this scenario too.)

The model covers the US only, so reading it across to India is our own step, and we might be wrong about the size of the effect. But the people Anthropic says may have to change jobs are coders, heading towards work like electrician or nurse. Coders are the job family most GCCs in India were built on. Since October 2014 we’ve placed 2,350+ people into 64 global capability centres (as of March 2026), and about 86% of all our placements since then have been specialist roles. That’s the knowledge work this model treats as most exposed.

The line we underlined twice was the plainest one. Changing occupations entirely is hard, and it takes many people a long time to land the next job. Anthropic gives three reasons: people may not want to change, they may need new skills, and even with the skills it’s hard to get hired. In the substantial scenario, 2.5% of all workers leave knowledge work by 2030 and 0.7% are still looking for their next occupation.

The external reviewers pushed on this too. They pointed out that the model doesn’t follow individual workers, so it can only give a coarse picture of what losing a job costs a person. A GCC head works in that gap, managing specific people and deciding which of them can move from a task that’s being automated to one that’s being created.

So we’d put the redeployment question into every senior brief this year. When we hire a delivery head or a head of engineering, we’d want to know whether they have moved a team from one kind of work to another and kept most of the people. We wrote about the roles AI is already creating inside GCCs, and a leader with this skill is the one who gets your people into them.

Today about 60 cents of every dollar the US economy produces goes to workers and 40 cents goes to capital. Anthropic’s finding is that in the substantial and extreme scenarios the workers’ share falls, to 56.1% in the substantial scenario and 45.2% in the extreme one, even while the economy grows. In the extreme case, society is far richer and total labour income is barely changed by 2030.

The GCC business case has usually been written as a labour case: the same work, done well, for less. If more of each dollar starts flowing to capital, the centre’s value moves towards what it owns and runs. That means the models, the data, the platforms and the teams who keep them running.

That changes who you want in the chair. The GCC head we’ll need next can argue a capital plan to headquarters as well as a headcount plan, and can explain why the centre should own a platform instead of renting more people. We’d start assessing for it now, before someone else rewrites the business case. Sachith made a related point about tasks changing before titles do, and the GCC head’s own role is no exception.

We like that Anthropic published the arguments against its own model. The 18 reviewers included Nobel laureate Daron Acemoglu and David Autor, both of MIT, and David Romer of UC Berkeley. Some felt the extreme scenario is better read as a thought experiment. Others felt the modest one understates what’s already visible in the data. Several asked Anthropic to be clearer that the model leaves out the demand created by building data centres.

Several also questioned whether the jobs most exposed to AI will shrink at all, or grow. If they grow, the redeployment skill still matters, because the tasks inside them change either way.

The authors name their own gaps: no hyper-capable robots, no policy responses, no business cycles. They call the explorer “a tool for thinking” and say actual outcomes may differ materially. Our reading is that the model is useful even if nobody can say which scenario is right. It tells you to check which one your current plan assumes, and whether you’d still be comfortable with it one level up.

Take the brief for the next leadership role you’re opening. Write out the 10 or 12 tasks that person will spend their week on. Then sort them the way Anthropic sorts the nurse’s day: tasks AI leaves alone, tasks it helps with, tasks it takes over, and tasks that only exist because of it.

If most of the brief sits in the first two groups, you’re hiring for the modest scenario. That might be right for the role, and it’s still worth saying out loud. If the last group is empty, we’d rewrite the brief before anyone reads a CV. We’ve covered the roles a GCC should stop hiring separately, and this exercise is a quicker way to find them in your own organisation chart.

If you’d like a second pair of eyes on it, send us the brief. We’ve run 650 talent-mapping and workforce-planning engagements (as of March 2026), and we’ll sort the tasks with you before the search opens. Share a brief with us and we’ll come back within one working day.