Job Market
What industries are most at risk from AI automation and how can workers adapt?
Quick answer
Exposure follows tasks, not job titles: work that is routine, screen-based and text-heavy sits most exposed, which is why customer support, data entry, basic content production, bookkeeping and routine analysis appear consistently in exposure research. Adaptation means moving toward judgment, relationships and physical-world work inside your field, and becoming the person who uses AI rather than competes with it.
The useful way to think about risk is tasks, not job titles. AI automates specific activities, and the most exposed ones share a shape: routine, screen-based, text- or number-heavy, with clear inputs and outputs and little physical or interpersonal component. That is why the same categories recur across exposure studies: customer support and call handling, data entry and processing, basic content and copy production, bookkeeping and routine accounting work, translation of standard material, and entry-level research and analysis. Precise percentages vary widely between studies and change with every model release, so treat any single headline number with suspicion; the pattern is the reliable part.
Exposure is also not the same as elimination. In most fields the realistic near-term shape is task compression: one person with AI doing work that took three, with the routine layer automated and the judgment layer remaining. That still moves the job market, especially at the entry level where the routine layer was the job, but it rewards a different response than panic: shifting your own mix of tasks toward what remains scarce.
Adaptation follows directly from that. Inside your current field, move toward the tasks AI handles worst: client relationships and trust, negotiation, judgment calls with incomplete information, physical-world work, managing people, and accountability for outcomes. Simultaneously become the person in your team who uses AI well, because the near-term displacement pattern is less AI replaces workers and more workers who use AI replace workers who do not. That position is available in almost every role today and costs only practice.
Concretely: audit your own week and estimate how much of it is routine and screen-based, because that share is your personal exposure regardless of your industry's headline risk. Then rebalance deliberately, pick up the judgment-heavy work nobody wants to define, and learn the tools on real tasks. Workers who did the equivalent during past technology shifts consistently ended up running the new tools rather than being displaced by them.