When AI makes work more productive, does it help everyone equally, or depend on which kinds of jobs you're stuck in?

This explores how the sorting of workers into different kinds of jobs (by gender, skill level and field) shapes who benefits when AI raises productivity. The corpus measures exposure to AI much more than it measures gains, so the picture is pieced together from several angles. What comes out is that segregation decides who is exposed, on what terms, and with what room to adapt.

Start with where exposure lands. In male-dominated occupations, AI exposure concentrates among the high-skilled, high-paid workers. In female-dominated occupations it spreads evenly across every skill and wage level, so lower-paid, lower-skilled women are exposed with fewer resources to adapt Does AI exposure hit low-wage workers harder in some fields?. The same skew shows up in what work actually gets handed to AI. Delegation clusters in information-intensive jobs and follows what the technology can do, not the older 'routine tasks get automated' story Where have workers actually delegated tasks to AI?. Vacancy data from ten countries add that AI skill demand is splitting in two. STEM jobs converge on a shared core of Python, SQL and machine learning, while non-technical jobs drift away from it Is AI creating common skills across jobs or deepening divisions?. Whoever sits inside the technical core gets a common toolkit. Everyone else has to work out their own version.

Exposure only turns into gain under certain conditions. Productivity boosts from AI mostly appear when people apply skills they already have. When workers used AI to learn something new, the gains disappeared and learning suffered When does AI actually boost worker productivity?. Occupational segregation is in effect a map of who already holds which skills, so those in the technical core arrive with the most to build on. A second angle is more tentative, because it is about tasks inside firms rather than whole occupations. When AI exposure is concentrated on a few tasks, workers can shift to the tasks it doesn't touch, and net employment effects stay modest Does concentrated AI exposure enable workers to adapt and reallocate?. It would be a stretch to say that exposure spread evenly across an occupation leaves less to shift into, but that is the worry the two findings raise together, and the corpus doesn't test it.

Segregation isn't destiny, though. In a randomized experiment, generative AI shrank the performance advantage of higher-educated adults on a problem-solving task by about three-quarters, and lower-education participants kept part of the gain after the AI was taken away Can AI narrow the education performance gap?. That looks like a leveling effect, and it sits awkwardly beside the 'you need existing skill' result. One way to square them is that the leveling appears in a single task with the tool in hand, and may not survive the occupational structure around it. A broader review reaches a similar conclusion. Generative AI can worsen or reduce inequality across work, education and healthcare, and the direction depends on access, integration and incentives, not on the capability itself Does generative AI inevitably worsen or reduce inequality?.

The corpus doesn't have direct evidence on who ends up with higher wages or output once AI is deployed, and that is the gap. It does suggest that occupational segregation is a good early predictor of who is exposed, how concentrated that exposure is, and who arrives with skills AI can amplify. Whether that becomes a gain or a loss is set by choices about deployment.