When the first draft is easy, how do we learn?
Making room for practice, judgment, and the work of becoming good.
The output is not the whole lesson
A finished piece of work can hide the learning that happened along the way. Choosing an approach, getting stuck, making a mistake, and revising a decision are part of building expertise. If a tool takes over one of those steps, it is worth asking what practice the person still needs.
A different kind of apprenticeship
Consider an illustrative junior analyst using AI to prepare a briefing. The draft may be useful. But accepting it is not the same as understanding how its conclusion was reached. A manager could ask the analyst to explain the evidence, identify an unsupported claim, and defend an alternative interpretation. The exercise turns a convenient output into a chance to practice judgment.
Design for learning, not just delivery
There is no single right amount of assistance for every task. On a low-stakes exercise, a person might try the problem independently before comparing their approach with an AI-generated response. In a time-sensitive workflow, the team might use assistance immediately and review the reasoning afterward. The point is to make the trade-off deliberate.
Ask what transfers
Being fluent with one interface is not necessarily the same as understanding a field. A useful test is whether someone can explain the problem, recognize a weak answer, and adapt when the conditions change. Those questions put understanding alongside speed, rather than treating the two as interchangeable.
A question for every team
Which activities teach people how to do the work, and which merely consume their time? Before automating an entry-level task, identify the skill it helps develop. Then decide how that skill will be practiced, reviewed, and recognized in the new workflow. A learning pathway should be designed, not left to chance.