What happens when AI-dependent workers must operate without their tools?

This reads the question as one about the morning after: a worker has been producing skilled-looking output with AI, the access disappears, and we want to know what remains. The corpus is unusually direct here — the answer is mostly "baseline." One line of research frames AI-enhanced ability as an exoskeleton: it makes you stronger while you're wearing it, but the strength doesn't persist once it comes off Does AI assistance build lasting skills or temporary abilities?. A controlled study makes the same point from the learning side — workers using generative AI performed substantially better on the task at hand, but when they did similar work unaided afterward, their performance showed no improvement at all Does AI assistance help workers learn lasting skills?.

What's striking is the *mechanism* the corpus offers for why this happens — and it's not laziness. Skill seems to form precisely in the friction the AI removes. Learners working without AI hit more errors and resolved them themselves, and that independent struggle is what produced retention; AI-assisted learners delegated the debugging away and bypassed the very cognitive work that builds competence — even the ones who leaned on AI *most* to debug scored *lowest* on later skill tests Does AI assistance remove a core learning channel through error work?. So operating without tools doesn't just reveal a missing skill; the tool's smoothness is part of why the skill never formed.

Here's the part the worker themselves may not see coming: they often don't know the skill is missing. A cluster of work describes a competence misattribution — when AI output is fluent and seamless, people fold it into their own self-image and come to believe they possess abilities they don't Do AI-assisted outputs fool users about their own skills?. Four mechanisms compound this — attribution ambiguity, the fluency illusion, cognitive outsourcing, and pipeline opacity — and they multiply rather than add, so the gap between felt and actual competence widens silently How do AI tools trick users into overestimating their own skills?. The shock of operating tool-less, then, is partly a shock of self-knowledge.

The corpus also lets you zoom out from the individual to the system, which is where it gets genuinely uncomfortable. If institutions stay aligned with human interests partly *because* they depend on humans who care about outcomes, then quietly hollowing out that human capability erodes the implicit checks holding systems on course — a gradual disempowerment that's hard to reverse once it sets in Does incremental AI replacement erode human influence over society?. There's a more hopeful counter-thread: at the labor-market level, when AI exposure is concentrated in a few tasks rather than spread evenly, workers can reallocate to the parts that aren't displaced, softening the aggregate blow Does concentrated AI exposure enable workers to adapt and reallocate?. The catch is that reallocation assumes the worker still *has* transferable skill to reallocate — which is exactly what the exoskeleton findings say may have quietly evaporated.

If there's a design lesson buried here, it's the argument for keeping humans genuinely in the loop rather than racing to full autonomy — collaborative human-agent systems outperform autonomous ones on judgment, ambiguity, and accountability, and they keep the human's hands on the work instead of letting the muscle atrophy Should AI systems stay collaborative rather than fully autonomous?. The thing you didn't know you wanted to know: "can this worker function without the tool?" and "does this worker realize they can't?" are two different questions, and the corpus suggests the second one is the more dangerous blind spot.