AI shrank the edge of advanced degrees on a business task by about three-quarters, but does matching a task mean matching the job?

This explores whether AI that helps weaker performers catch up could open up elite professions, or whether doing equally well on a task is a different thing from getting equal access to the work. The strongest result in the collection is striking. In a randomized experiment with 1,174 adults, generative AI cut the performance advantage of people with higher education by about three-quarters on a business problem-solving task. Participants with less education also kept part of their gain after the AI was taken away, which suggests they learned something rather than just leaning on the tool Can AI narrow the education performance gap?. On a single task, then, a credential can become much less of an advantage.

But elite professions don't hire people for single tasks. A separate analysis of 960 real occupational workflows found that AI agents do well on contained, contest-style problems and struggle with the long, multi-step work that real jobs involve Why do agent benchmarks not predict real economic value?. The same caution plausibly applies to the human results. A gap that closes on a bounded experimental task may not hold up across the judgment, context, and long time horizons that set senior professionals apart. Lab equality is a promising signal, not proof that people will advance more equally at work.

The labor-market evidence points the other way. Job-vacancy data from ten countries show demand for AI skills concentrating inside technical occupations, around Python, SQL, and machine learning. Non-technical jobs are moving away from that core rather than toward it Is AI creating common skills across jobs or deepening divisions?. So AI may level performance within a task while splitting the job market into a technical core and everyone else. Getting in the door is also changing. Hiring is turning into an AI arms race: applicants mass-submit applications and slip hidden instructions past AI screening tools, while recruiters spend more and more time filtering out spam Are job applicants and employers locked in an escalating AI arms race?. When everyone's application looks polished, employers tend to fall back on signals AI can't fake, such as networks, pedigree, and referrals. Those are the signals that favor people who already have advantages.

The broader review in the collection says the outcome isn't fixed by the technology. Across work, education, and healthcare, generative AI has both widened and narrowed inequality, depending on who has access, how it is built into workflows, and what incentives surround it Does generative AI inevitably worsen or reduce inequality?. One more argument fits here: because these models are built from humanity's shared written output, restricting access to them can itself create inequality by turning shared knowledge into a private advantage Should restricting AI access create new kinds of inequality?. Whether AI opens up elite professions may therefore depend less on how capable the tools are and more on who gets to use them, and on whether employers start trusting what people can actually do over where they went to school.

One gap in the collection: none of these sources directly studies law, medicine, finance, or academia over time. The answer is built from task experiments, hiring data, and policy reviews, not from evidence of who actually reaches the top of those fields.