When AI helps with a business problem, the education gap shrinks, and it may be supplying the viewpoint a colleague would.

This explores whether generative AI evens out the performance differences that come from different backgrounds: education, job function, or seniority. The short answer from the corpus is yes, often by a lot, at least on the task in front of you. In a randomized experiment with 1,174 adults working on a business problem-solving task, AI cut the advantage of higher-educated participants by about three-quarters, from 0.548 to 0.139 standard deviations Can AI narrow the education performance gap?. A field experiment with 776 Procter & Gamble professionals found a similar leveling across job functions. Individuals using AI matched the output of two-person teams working without it. Their solutions also became more balanced across the commercial and technical sides, as if the AI supplied the perspective a colleague from another department would normally bring Can generative AI replace the benefits of having a human teammate?.

It helps to ask why the gaps close, and here the corpus offers an angle the workplace studies don't spell out. Models trained on many imperfect experts tend to land on the consensus of those experts, which averages out individual errors and quirks. At low sampling temperature that consensus can beat any single expert Can models trained on many imperfect experts outperform everyone?. Different models also converge on strikingly similar answers to open-ended questions, an effect one study calls an 'Artificial Hivemind' Do different AI models actually produce diverse outputs?. Together these suggest that part of the narrowing may come from everyone being pulled toward the same competent consensus answer. That raises the floor, but it may also flatten the distinctive approaches that different backgrounds used to bring. The corpus doesn't measure this directly for the workplace studies, so treat it as a question worth carrying into them.

The bigger complication is the difference between closing a performance gap and closing a skill gap. Interviews with South Korean software engineers describe AI absorbing the entry-level tasks juniors once learned from. That work now flows into senior-plus-AI workflows, which removes the hands-on struggle through which expertise used to develop Does generative AI prevent juniors from getting entry-level work?. So the gap can shrink on today's deliverable while the pipeline that turns juniors into seniors erodes. The education experiment gives a partial counterweight: lower-education participants kept some of their gain after the AI was taken away, so the leveling wasn't purely borrowed capability Can AI narrow the education performance gap?. One plausible reading is that the outcome depends on whether the AI does the work for you or alongside you.

The wider evidence points the same way. A cross-domain review of information, work, education, and healthcare concludes that generative AI can either widen or narrow inequality. Which one happens depends on who has access, how the tool is built into the work, and what incentives surround it, not on the technology itself Does generative AI inevitably worsen or reduce inequality?. Usage data adds a further twist. Heavy AI users shifted toward solo documentation and away from communication and coordination Does generative AI shift knowledge workers away from communication?. If AI narrows gaps partly by standing in for the colleague you would otherwise consult, the leveling may come with less of the cross-background contact that used to spread expertise between people. The takeaway is that the gap usually does narrow, and the open question is what the narrowing costs.