Do richer, better-educated workers use AI more or less? The evidence says it depends on the kind of work, not the pay grade.

This explores whether AI use climbs or drops as workers get more educated and better paid. The corpus doesn't give one slope. It gives two different pictures depending on the field, plus a warning that most of the data measures exposure rather than adoption. The old automation story says routine, lower-paid work goes first. The evidence on where workers have actually handed tasks to AI cuts against that. Delegation concentrates in information-intensive occupations and follows what the technology can do, not routine-task predictions. The wage pattern also reverses at advanced-degree levels (Where have workers actually delegated tasks to AI?).

The gender split shows why a single education-and-wage curve misleads. In male-dominated occupations, AI exposure concentrates among high-skilled, high-paid workers. In female-dominated occupations it spreads evenly across skill levels. That leaves lower-paid, lower-skilled women heavily exposed with fewer resources to adapt (Does AI exposure hit low-wage workers harder in some fields?). The same axis of more education and more pay points up in one part of the labor market and stays flat in another.

Skills matter for using AI, not just for being touched by it. Vacancy data from ten countries shows AI skill demand piling up in STEM jobs around Python, SQL, machine learning and data analysis. Non-technical occupations drift away from that core instead of converging on it (Is AI creating common skills across jobs or deepening divisions?). Productivity gains also show up when people apply skills they already have. When workers used AI to learn something new, the gains vanished and learning suffered (When does AI actually boost worker productivity?). Put together, these hint that existing expertise is what lets someone benefit from AI, but the corpus doesn't test that directly against wage or degree level.

Adoption is also decided by firms, not just individual workers. Firms with higher AI exposure replace online-marketplace freelancers with AI faster and more cheaply than less-exposed firms, which points to returns to scale in internal AI capability rather than even diffusion (Do firms substitute labor for AI at different rates?). How exposure is spread matters as much as how much there is. When it hits only a few tasks in a job, workers can shift to the tasks left over and employment falls only modestly. When it is broad, labor demand falls (Does concentrated AI exposure enable workers to adapt and reallocate?).

Two cautions apply. The long-run economic models suggest wages stop protecting anyone: as AGI takes over bottleneck work, human pay tracks the compute cost of replicating the work, not its value (What happens to human wages in an AGI economy?). And exposure is not the same as real-world readiness. Agents that clear benchmark contests still fail at long-horizon professional work (Why do agent benchmarks not predict real economic value?). The corpus has no clean adoption-by-wage curve, only signs that the answer depends on occupation, gender and the firm.