A reflection on how AI-enabled efficiency can weaken the process through which tomorrow's experienced people are formed — and what leadership must do to protect it.
Much of what we know at work was never taught formally. We learned by sitting beside someone more experienced. We watched how a senior colleague prepared for a difficult meeting. We noticed which numbers they questioned, where they paused and how they responded when the conversation moved away from the plan.
We learned not only from their answer, but from the path they took to reach it. This is informal learning — it happens through observation, participation, correction and repeated exposure to the judgement of others. It is rarely documented because experienced people may no longer be fully aware of what they know.
Artificial intelligence may make work more efficient while unintentionally removing some of these learning moments. A junior employee once prepared the first draft of a report and received detailed corrections from a manager. The process was slow, but the corrections revealed how the manager thought. Today, AI may prepare the draft before either person becomes involved.
A young analyst once organised raw information and gradually learned which patterns mattered. Now an intelligent system may move directly from data to recommendation. A new executive once attended meetings partly to observe how difficult decisions were made. Increasingly, they may receive a concise AI-generated summary after the meeting has ended.
The final output may improve. The path through which professional judgement is transferred may become less visible.
This is not merely a training issue. It is a behavioural and cultural issue. The virtual communities examined in my research showed that participation depends partly on how the environment is designed and managed. A community may contain knowledgeable people, but knowledge will not necessarily move between them unless the environment creates opportunities for consultation, interaction and meaningful contribution.
The same principle applies inside AI-enabled organisations. Knowledge does not transfer simply because employees have access to the same intelligent system. AI may give everyone an answer while leaving fewer opportunities to understand how experienced people interpret the situation.
This matters because some of the most valuable organisational knowledge is tacit. It includes knowing when a customer's objection is not really about price. It includes recognising when a technically correct decision will fail because the people responsible for implementing it do not believe in it. It includes sensing when silence in a meeting represents agreement — and when it represents fear, confusion or respectful disagreement.
Such knowledge is difficult to reduce to a procedure. It develops through social experience. If AI increasingly handles intermediate work, organisations may become more productive today while weakening the process through which tomorrow's experienced people are formed.
Senior professionals may also change their behaviour. When AI can complete routine analysis, they may involve junior employees later in the process. The experienced person reviews the answer, makes a few adjustments and moves on. The task is completed efficiently, but the reasoning remains private.
Over time, an organisation may develop a thin middle layer of capability — many people able to use AI and a smaller number still able to recognise when it is wrong. This creates a succession risk. The senior people who understand the underlying business will eventually leave. If their judgement was never made visible, the organisation may discover that it preserved the output but lost the knowledge that made the output trustworthy.
The answer is not to preserve unnecessary manual work merely for tradition. It is to redesign learning deliberately — experienced employees explaining not only what they changed in an AI-generated answer, but why; teams reviewing difficult cases in which context mattered more than the model's first recommendation; junior colleagues invited into the framing of the problem rather than joining only after the system has produced a solution.
AI can also support informal learning when used thoughtfully. It can record decision rationales, compare alternative approaches and help experienced people make tacit assumptions more explicit. But the technology cannot decide that knowledge transfer matters. That remains a leadership responsibility.
When AI removes the steps between a question and an answer, organisations must protect the human learning that once happened along the way.