AI makes skilled people faster at what they already know, but does it really help anyone become skilled?

This explores whether AI's quick productivity wins come at the expense of the effortful practice that builds real expertise. The corpus points to a specific answer: the gains and the skill-building happen in different places. Productivity gains show up mainly when people apply skills they already have. When workers used AI to learn something new, the gains disappeared and their learning suffered When does AI actually boost worker productivity?. So the headline speedups mostly describe experts working faster. They aren't evidence that AI helps anyone become an expert.

The sharpest image in the collection is the exoskeleton. Workers using AI produce skilled-looking output, but when the AI is taken away they drop back to their baseline Does AI assistance build lasting skills or temporary abilities?. The ability lived in the tool and never moved into the person. This links to a wider argument that AI separates the finished product (the essay, the code, the analysis) from the thinking that used to be required to make it Does AI separate intellectual form from the thinking behind it?. If the output no longer proves anyone struggled to make it, neither the worker nor the employer can easily tell whether learning happened.

The risk compounds in a way you might not expect. Anthropic's own engineers report about 50% productivity gains, yet most say they can fully hand off only 0–20% of their work. They worry that letting Claude do the routine coding wears away the hands-on practice they need to catch Claude's mistakes Does AI assistance erode the skills needed to oversee it?. The routine work people are glad to give away is the same practice that keeps their judgment sharp enough to supervise the AI. An economic model scales this up to whole organizations. Automating entry-level tasks can raise output while slowing long-run growth, because novices stop working alongside the experts who pass on unwritten know-how. This happens even when junior jobs don't disappear Can automation raise output while slowing growth?.

There's a counterpoint worth taking seriously. One industrial case study wrote experts' rules and design principles directly into an AI agent's setup. Non-experts then produced work rated at expert level, with no specialist reviewing it Can codified expertise let non-experts match specialist output?. A related finding says skills help agents mainly by anchoring the steps they follow, not by supplying missing facts Do skills teach procedures or inject missing facts?. This suggests expertise doesn't have to sit inside a person to be useful. The catch is that someone had to struggle to become expert enough to write those rules down. Codified expertise lives off past struggle and doesn't replace it.

The same pattern appears at the level of a whole field. Scientists who use AI publish about three times as many papers and get far more citations. Meanwhile, science as a whole covers fewer topics and researchers collaborate less, because AI pulls work toward problems that already have plenty of data Does AI help individual scientists while narrowing scientific focus?. Individuals gain while the shared ground they learn from shrinks. The corpus doesn't directly measure "struggle" as a learning mechanism, so the claim that effort itself builds skill is assumed rather than tested here. But the evidence agrees on the risk: today's gains draw on expertise built the old way, and few of the studies ask where the next generation's expertise will come from.