AI Exposure vs Adaptive Capacity: the workforce chart every leader should study

Most AI workforce debates start with the wrong question: which jobs will AI take? The better question is where exposure to AI intersects with low adaptive capacity, because that is where the real human, operational, and reputational risk sits.

That framing comes from the January 2026 NBER paper by Manning and Aguirre, “How Adaptable Are American Workers to AI-Induced Job Displacement.” I built the interactive chart below so leaders can explore 356 U.S. occupations, covering 95.9% of the workforce, across two axes: AI exposure and adaptive capacity. Bubble size reflects employment. Colour reflects vulnerability. You can filter by gender, age, sector, and job family. For the source paper, start with the NBER working paper.

The headline finding is counterintuitive enough to stop most people in their tracks: exposure and adaptive capacity are positively correlated. The jobs most exposed to AI are often held by people who are also best equipped to adapt.

That matters. Of the 37.1 million U.S. workers in the top quartile of AI exposure, 26.5 million have above-median adaptive capacity. In plain English: many of the people most exposed to AI are not headed for oblivion. They are more likely to be disrupted, reconfigured, and made more productive than permanently displaced.

Where the real risk concentrates

The real danger sits in a much smaller but far more fragile pocket: 6.1 million workers, or 4.2% of the U.S. workforce, who are both highly exposed and low in adaptive capacity. These are the kinds of roles that exist in every large services organisation: secretaries, administrative assistants, information clerks, tax preparers, and routine back-office processing roles.

That is why “AI will replace jobs” is too blunt to be useful. A senior compliance analyst may be highly exposed and still adapt well, because the underlying skills transfer. A routine reconciliation or processing role may be similarly exposed, but with no obvious next step for the person doing it. Those are very different leadership problems.

What leaders should do with this chart

Use the interactive chart as a map, not as theatre. Find your own roles. Find your team’s roles. Then ask two questions, role by role.

• How much of this role’s task list can AI now do?

• If AI does more of it, where do these people go next inside my organisation or outside it?

That second question is the one most leadership teams are still avoiding. It is also where your moral obligation and your reputation risk sit. Handle the high-exposure, low-adaptive-capacity pocket badly and you will create an avoidable transition crisis for people who were never the ones setting the strategy.

If you spend ten honest minutes with this chart, the workforce conversation gets much sharper. It stops being “AI is coming for jobs” and becomes “which parts of my workforce are resilient, which parts are vulnerable, and what is my plan before I get surprised by it in 2027?”

For the adjacent questions of synthetic trust and Australian policy direction, read The video in your inbox may not be real anymore and What the 12 May 2026 Federal Budget says about where AI money is really going.