Does AI make people better at their work, or does the boost only last while the tool is switched on?
Do gains from AI assistance disappear when workers complete tasks alone?
This explores whether the performance boost workers get from AI tools stays with them once the tool is taken away, or whether it only exists while the AI is in the loop.
This explores whether AI makes workers better, or only makes their output better while the AI is present. The corpus says mostly the second. When people who did tasks with generative AI later did similar tasks on their own, their performance showed no improvement. The gain did not carry over to new contexts Does AI assistance help workers learn lasting skills?. One note calls this an exoskeleton. While you wear it, you produce skilled-looking work. Take it off and you're back to baseline. Real skill is different because it stays with you Does AI assistance build lasting skills or temporary abilities?.
The more surprising point is that the strongest productivity studies were never built to answer this question. Much of the headline evidence measured people doing tasks they already knew how to do. When workers used AI to learn something new, the gains disappeared and learning itself suffered When does AI actually boost worker productivity?. Read the well-known results with that in mind. A field study of 5,172 support agents found AI raised issues resolved per hour by 15%, with the biggest jumps for the least experienced agents Does AI assistance help less experienced workers most?. A Procter & Gamble experiment found individuals with AI matched two-person teams working without it Can generative AI replace the benefits of having a human teammate?. Both studies measure performance with the AI present. Neither can tell you what those novices could do alone a year later. It's possible that the workers who gained the most are also the ones who built the least independent skill.
There's a second-order problem. The skill AI may wear down is the same skill you need to check the AI's work. Anthropic engineers reported 50% productivity gains, yet most said they could fully hand off only 0–20% of their work. They worried that letting Claude do routine coding takes away the hands-on practice they need to catch its mistakes Does AI assistance erode the skills needed to oversee it?. A study that mapped 8,356 workplace AI risk scenarios reached the same conclusion: 'augmentation' (AI helping rather than replacing) is not safe by default, because relying on it too much slowly erodes both the skill and the ability to oversee the AI Does AI augmentation protect workers from skill erosion?. Help can also have a cost while it's happening. Correct AI suggestions can still break a person's concentration and pull them out of their own line of reasoning Does AI assistance always help reasoning or does it carry hidden costs?.
Why don't organizations notice? Their measurements are tuned to output. Usage data records work done with AI, but nothing records what a worker could do without it. Systems see expertise being used much better than expertise being built Can we measure whether AI erodes independent skill?. Work culture adds to this. Interviews show workers protect cues like their personal voice and authorship, while signs of effort and uncertainty disappear into the finished deliverable Which workplace cues survive AI mediation and which disappear?. A finished task gets treated as proof the person can do it.
So the honest answer is: for skills workers already have, AI raises output. For skills they're building, the evidence suggests the gains mostly don't survive working alone, and the standard metrics can't see this. The bigger open question isn't whether AI makes workers productive. It's whether a workforce that grows up with AI will ever build the skills needed to supervise it.
Sources 10 notes
Wu et al. found that workers using generative AI performed substantially better on content tasks, but when performing similar tasks independently afterward, their performance showed no improvement. The capability did not transfer across contexts.
Research shows AI assistance creates temporary capability extensions—workers produce skilled-looking output while AI is present but revert to baseline performance when access is removed. This differs fundamentally from true skill, which persists independently.
Studies showing AI productivity gains measured tasks within workers' existing domains. When workers used AI to learn new skills, productivity gains disappeared and learning suffered, suggesting prior findings do not generalize to skill acquisition.
A study of 5,172 support agents at a Fortune 500 firm found a 15% average productivity gain from AI assistance, with gains concentrated among less experienced workers who improved both speed and quality. The most experienced agents saw small speed gains but slight quality declines.
In a randomized field experiment with 776 P&G professionals, individuals using AI produced solutions as strong as two-person teams without AI. AI also reduced functional silos by prompting more balanced solutions across professional backgrounds.
Show all 10 sources
Anthropic's 132-person survey found 50% self-reported productivity gains and 67% more merged pull requests, yet most engineers can only fully delegate 0-20% of work. Employees fear that relying on Claude for routine tasks erodes the hands-on coding practice needed to catch its errors.
Research mapping 8,356 workplace AI risk scenarios found that augmentation mode does not inherently prevent harm. Overreliance on AI agents can gradually erode worker skills and their capacity to provide meaningful oversight, undermining augmentation's core safety justification.
Well-intentioned AI suggestions can damage reasoning performance by severing cognitive immersion, forcing users to rebuild focus before continuing. Evaluation must measure flow preservation across entire tasks, not just local suggestion accuracy.
Usage data registers assisted output but not independent capability. A stock-formation gap means current systems observe expertise in use better than expertise being built, leaving AI's skill effects fundamentally undetermined.
Analysis of 1,250 interviews found workers preserve identity-bearing cues like voice and provenance but allow effort, attention, and uncertainty to vanish into deliverables. This asymmetry occurs because output-centered work treats finished tasks as proof work happened, leaving labor-bearing cues unexamined.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- How AI Impacts Skill Formation
- Toward Measuring AI's Effects on Skill Formation: The Stock-Formation Gap
- Does generative AI narrow education-based productivity gaps? Evidence from a randomized experiment
- AI Skills Improve Job Prospects: Causal Evidence from a Hiring Experiment
- Generative AI at Work
- Adoption of Generative AI in the Workplace: Increasing and Shifting the Balance of Productivity and Communication Activity
- What 81,000 people told us about the economics of AI
- Research: Gen AI Makes People More Productive—and Less Motivated