Does concentrated AI exposure enable workers to adapt and reallocate?

Using novel task-level AI exposure measures across firms from 2010 to 2023, this study identifies two variables that jointly determine AI's impact on within-firm labor demand: an occupation's mean task exposure to AI, and the concentration of that exposure across tasks.

Higher mean exposure reduces labor demand — unsurprising. But more concentrated exposure (AI affecting a small number of tasks rather than spread across many) plays an offsetting role. When AI displaces specific tasks, workers reallocate effort to non-displaced tasks within the same occupation. The offset is empirically significant: "relatively modest net employment effects due to countervailing forces — reduced demand in AI-exposed occupations is offset by productivity-driven employment increases across all occupations at AI-adopting firms."

The concentration mechanism matters because it determines whether adaptation is possible. If AI displaces 3 out of 20 tasks in an occupation (concentrated), workers shift effort to the remaining 17. If AI partially affects 15 out of 20 tasks (diffuse), there is nowhere to reallocate. This means the distribution of AI impact within an occupation matters as much as the level — a finding that complicates blanket "X% of jobs at risk" estimates.

Since Does incremental AI replacement erode human influence over society?, the reallocation mechanism may be temporary. Workers who reallocate to non-displaced tasks maintain employment but shift toward tasks AI cannot yet perform — which may be the interpersonal and organizational skills the WORKBank study identifies as gaining importance. The question is whether this reallocation constitutes genuine human adaptation or merely the gradual concentration of human labor into the tasks that haven't been automated yet.

Since What makes delegation work beyond just splitting tasks?, the concentration finding adds an empirical dimension: tasks that are delegatable (high verifiability, low subjectivity) will be displaced first, concentrating remaining human work in subjective, hard-to-verify domains. The eleven axes predict which tasks get displaced; the concentration mechanism predicts what happens next.

Source: Work Application Use Cases