When AI takes over a few parts of a job, workers can move onto the parts it leaves alone.

This explores whether the *shape* of AI exposure inside a job matters, not just its total amount. The clearest answer in the corpus comes from one firm-level study covering 2010–2023, and it separates two quantities people usually treat as one. Higher *average* exposure across a job's tasks does reduce labor demand. But when that exposure is *concentrated*, meaning AI can do a few tasks very well and leaves the rest alone, workers can move their time onto the tasks AI doesn't touch. That shift offsets much of the employment loss, so the net effect is modest Does concentrated AI exposure enable workers to adapt and reallocate?. The useful point is that a job isn't simply 'exposed' or 'safe.' A job where AI takes over two tasks completely can be easier to adapt than one where AI partly covers every task, because the first job still has untouched work to move into.

Other notes in the corpus show where this happens in practice. When workers actually hand tasks to AI inside structured workflows, the delegation clusters in information-heavy occupations and follows what the technology can do, not who chats with LLMs the most. The pattern also breaks from older predictions about routine-task automation Where have workers actually delegated tasks to AI?. Reallocation also happens at a smaller scale, inside individual tasks: AI often doesn't shorten total task time. Instead it moves time from doing the work to writing prompts and checking the output Does AI really save time, or just change how we spend it?. So when workers 'move to non-displaced tasks,' some of what they move into is new work that AI itself creates.

An outcome the reallocation story doesn't predict shows up in hiring. Payroll data show no economy-wide job losses from AI, which fits the offsetting effect. But young workers in AI-exposed occupations face hiring rates 19% lower than peers in less-exposed fields, and experienced workers show no such gap Is generative AI displacing workers at economy-wide scale?. One possible reading (an inference the corpus doesn't test directly): reallocation works for people who already hold a job with other tasks to move into, while entry-level roles, which tend to be built from the very tasks AI handles well, simply don't get created. Firms also differ in how fast they make these trades. Highly exposed firms replace outside freelance labor with AI faster and more cheaply than other firms, which suggests adjustment depends on each firm's AI capability, not on one market-wide rate Do firms substitute labor for AI at different rates?.

Who gets the room to reallocate is also uneven. Exposure in male-dominated occupations is concentrated among high-skill, high-wage workers, but in female-dominated occupations it spreads evenly across all skill levels. That leaves lower-paid women exposed with fewer resources to adapt Does AI exposure hit low-wage workers harder in some fields?. This note measures how exposure spreads across *workers*, not across tasks, but it points the same way: broad, even exposure leaves less shelter than narrow, targeted exposure. At the skill level, AI demand pulls technical occupations toward a shared core of Python, SQL and machine learning, while non-technical occupations move away from that core rather than toward it Is AI creating common skills across jobs or deepening divisions?. That limits how far a reallocating worker can move toward AI-adjacent work.

A note on how thin this is: only one study in the collection tests the concentration-versus-average question directly, so treat it as a promising finding, not a settled one. The longer-term warning comes from the gradual-disempowerment argument. Even if each step of reallocation looks benign, each task AI absorbs reduces how much institutions depend on people who care about the outcomes, and that dependence is one of the quiet ways those institutions stay aligned with human interests Does incremental AI replacement erode human influence over society?.