Does using AI at work make your skills sharper over time, or quietly let them fade?

This explores whether some jobs lose skills to AI tools while others gain them. The direct answer is that the collection has no clean profession-by-profession scorecard, and the research gives reasons to doubt one could be built yet. Instead, the same pattern shows up across very different kinds of work. People produce better output while AI is present, then fall back to their old level once it's taken away. Wu et al. found that content workers did much better with generative AI but showed no improvement when they later did similar tasks alone Does AI assistance help workers learn lasting skills?. The corpus calls this an exoskeleton: the boost is real while you wear it, and it isn't the same thing as skill Does AI assistance build lasting skills or temporary abilities?.

The clearest case of a specific profession worrying about erosion is software engineering, and the worry is unusual. Anthropic's internal survey of engineers found self-reported productivity gains of about 50%, but most could fully delegate only 0–20% of their work. Their concern wasn't losing coding as such. It was losing the hands-on practice they need to *catch the AI's mistakes* Does AI assistance erode the skills needed to oversee it?. That's a trap worth knowing about: in jobs where people supervise AI output, the skill most at risk may be the one that makes supervision possible. A related finding makes this harder to notice from the inside. When AI-assisted work feels smooth and fluent, people tend to count it as evidence of their own ability Do AI-assisted outputs fool users about their own skills?.

There's a strong counterargument to the erosion story. Nielsen argues that studies which take AI away and then test people measure a situation that almost never happens at work. The better question is which higher-level skills grow when AI is always available Does removing AI tools actually measure real skill loss?. A second problem is measurement. Usage data from real deployments shows expertise being *used*, not expertise being *built*, so the long-term effect on skill is still undetermined Can we measure whether AI erodes independent skill?. So 'erosion vs. development' partly depends on which test you run.

The labor-market data adds a different kind of divide. Job vacancy data from ten countries shows AI skill demand clustering in technical roles (Python, SQL, machine learning), while non-technical jobs drift away from that core instead of converging on it Is AI creating common skills across jobs or deepening divisions?. Yet in hiring experiments, listing AI skills raised interview invitations across occupations, even though recruiters rarely checked actual competence Do AI skills help candidates get more job interviews?. The boost was strongest for office assistants and weaker for graphic designers, and it partly offset hiring penalties for older candidates and people without bachelor's degrees Can AI skills help older or less-educated job candidates?.

Put together, the surprising takeaway is that 'AI skill' as the job market rewards it and 'skill' as the learning research measures it may be drifting apart. In some roles, being seen as AI-capable improves your prospects, while the underlying independent ability may not be growing at all.