Traditional training often separates learning from work. People first memorise rules and observe examples, then handle real problems independently after gaining enough experience. Generative AI changes this timing. It can offer specific guidance while a task is unfolding, bringing some of the experience that once took months of observation into the immediate workplace.
A 2025 study in The Quarterly Journal of Economics examined 5,172 customer-support agents. With access to an AI assistant, workers resolved 15 per cent more issues per hour on average, with larger gains among newer and lower-skilled staff. Agents with two months of tenure and AI support performed as well as those with more than six months of experience but no AI access. The finding comes from one firm and one occupation, so it should not be treated as a universal result.
The deeper change is not that novices no longer need experience. It is that learning and execution begin to overlap. AI puts a possible action into the live situation; the worker develops judgement by using it, modifying it and observing what follows. Organisations still need experts because the system's usable experience originates in human work, and new problems still require people to contribute new approaches. AI shortens the time it takes for experience to reach the work. It does not remove the value of experience itself.
https://academic.oup.com/qje/article/140/2/889/7990658
Discover more from Geoffrey Chen
Subscribe to get the latest posts sent to your email.