Does using AI at work make you better at your job — or just make you look better while it's running?

This explores whether AI help at work gets in the way of people building skills they keep, and whether the answer changes by profession. The short version: across the domains the corpus covers, the most consistent finding is that AI raises how well people perform while they're using it, and that this tells you very little about what they've learned. Wu et al. found that workers using generative AI did much better on content tasks, but when they later did similar tasks on their own, they showed no improvement at all Does AI assistance help workers learn lasting skills?. The corpus has a useful image for this: AI-boosted ability works like an exoskeleton. Output looks skilled while the AI is there, and performance drops back to baseline once access is removed Does AI assistance build lasting skills or temporary abilities?.

This changes how to read one of the best-known workplace results. In the BCG experiment with 758 consultants, GPT-4 helped the bottom half of performers most and narrowed the gap with top performers Does AI help lower-skilled workers catch up to top performers?. That's often taken as AI 'upskilling' weaker workers. Seen through the exoskeleton lens, the gap closed in output, and nobody checked whether it closed in ability. The same study warns that when using AI is optional rather than required, the equalizing effect may reverse. For software engineering, the field where people use AI most heavily, Anthropic's survey of its own engineers adds a twist: people reported a 50% productivity boost, yet most said they could fully hand off only 0–20% of their work. Their main worry was that letting Claude do the routine coding wears down the hands-on practice they need to catch Claude's mistakes Does AI assistance erode the skills needed to oversee it?. So the risk goes beyond losing skills. The specific skill at risk is the one that makes it safe to depend on the AI in the first place.

The corpus also points to a mechanism below the level of any one profession. A four-month EEG study found that brain connectivity fell as reliance on AI grew. LLM users showed the weakest neural engagement and had trouble even recalling work they had just produced Does AI assistance weaken our brain's ability to think independently?. Separately, AI interventions can hurt reasoning even when the suggestion is correct, because they break a person's focus and force them to rebuild it Does AI assistance always help reasoning or does it carry hidden costs?. Both point to a cost that comes from how people engage with the work, not from the AI being wrong, which suggests it would carry across fields rather than belong to one.

Here's what you might not have expected to learn: the job market is currently rewarding the exoskeleton. Recruiters gave candidates who listed AI skills 8–15 percentage points more interview invitations across occupations, and they barely distinguished a certificate from a simple self-claim Do AI skills help candidates get more job interviews?. Meanwhile, demand for AI skills is concentrating in a technical core (Python, SQL, machine learning) instead of spreading evenly across jobs Is AI creating common skills across jobs or deepening divisions?. So the signals employers reward and the skills people actually build may be pulling apart.

A candid limit: the corpus can't yet give a clean answer for each profession, and one note explains why. Usage data captures expertise being used, not expertise being formed. Real-world deployment data can't tell you whether AI is wearing down independent skill, because nobody is measuring what people can do once the tool is gone Can we measure whether AI erodes independent skill?. The evidence here covers consulting, software engineering and general writing or content tasks. It's consistent across them, but it isn't a systematic comparison across professions.