Protect the unfinished thinker

The first version of a person's thinking can be awkward. It reaches for the wrong distinction. It mistakes a familiar example for a useful one. It contains a question the writer has not yet learned to ask. A capable assistant can remove much of that awkwardness before a manager ever sees it.

Sometimes that is a gift. Sometimes it removes the very material the manager needed to help the person grow.

Imagine a hypothetical small publishing team preparing an exhibition guide. A junior editor uses AI to produce a polished introduction from the curators' notes. The paragraphs are clear, the transitions smooth, and the central argument sounds convincing. In the review meeting, the senior editor asks why one work appears before another. The junior editor cannot explain the sequence beyond saying it reads well.

There is no need to accuse the junior colleague of laziness. The team asked for a good introduction quickly. That is what arrived. The missing thing is an opportunity to develop and defend an interpretation. If nobody asks for it, the finished-looking page can conceal how much learning remains unfinished.

In his speculative letter Darwin Among the Machines, Samuel Butler imagines humans becoming dependent attendants to the machinery they have developed. His provocation is about dependence and power, not an empirical account of learning. It gives a manager a useful question: what happens to our own capacities inside the arrangement we are improving?

The answer cannot be to keep every inconvenience. Some routine work teaches little. Some barriers simply exhaust people who could be doing more worthwhile work. Better tools can help a newcomer attempt something that was previously beyond reach. The managerial task is to distinguish a barrier worth removing from a difficulty through which understanding can grow.

In the publishing example, retyping the curators' notes is unlikely to be the valuable struggle. Choosing what the visitor should notice, tracing a claim back to an artwork, and explaining why the sequence matters are closer to the craft. Protecting those decisions gives the junior editor something worth becoming good at.

That protection costs senior attention. A manager who insists that people think for themselves while providing no time for discussion has preserved an ordeal. The point is to arrange a challenge that can be attempted, examined, and attempted again. Feedback needs to reach the person's reasoning, not merely replace it with the senior colleague's preferred wording.

AI may have a constructive role after an initial attempt. The editor could compare alternative structures, ask what each emphasizes, and test a choice against the source notes. The useful outcome would be greater command of the argument. A beautifully finished page remains part of the work; it stops being the only evidence of progress.

The practice

Choose one task a less experienced colleague is learning. Ask them which decision inside it they find hardest to explain. If you do not manage anyone, use a part of your own work where you regularly accept a fluent result you cannot defend. Keep the first experiment modest enough to complete within a week.

Write a short practice agreement. Name the decision to be learned, a first attempt to make without generated wording, and the feedback someone will provide. Preserve one stage of reasoning rather than banning assistance from the whole task. The output might be a rough structure, a comparison of two choices, or a paragraph explaining an interpretation.

For the junior editor, the first attempt could be a five-sentence account of what the visitor should understand and why the artworks appear in that order. The senior editor spends fifteen minutes asking about the choices. They then use whatever permitted assistance helps turn the interpretation into a finished introduction.

Include a second, slightly different task. Could the editor arrange a new selection of works and explain the rationale with less help? That is an opportunity to notice whether understanding travels beyond the example. It is a proposed learning check, not a validated measure of expertise.

Keep the agreement humane. Do not use a novice's rough attempt as a surprise performance test or demand unassisted work where errors would harm others. Bound the exercise, provide support, and review whether the difficulty is earning its place. If the person only becomes more anxious or more dependent on guessing your preferences, change the practice.

A manager can admire the speed of a finished result and still make room for an unfinished mind. The person learning the craft needs enough room to become its author.