Do Not Automate Away the Work That Builds Judgment

AI is making it easier to remove junior work from product teams. A model can summarise interviews, draft requirements, analyse support themes and prepare a first version of a roadmap update. At the same time, companies are questioning how many junior people they need, while product leaders are being told to move faster, stay closer to customers and improve the quality of their decisions.

These developments do not fit together as neatly as they first appear. The work we are automating is often the work through which people learn to exercise judgment.

A junior product manager does not become a strong product leader by learning to write better documents. They improve by seeing why a plausible customer request should not become a feature, why an apparently positive metric is misleading, or why a commercially attractive opportunity would create an operational mess. Writing the document was never the real value. Investigating the situation, forming a view and being accountable for the recommendation was.

If AI performs the investigation and a senior person approves the output, the team may complete the task faster. But the junior person has only observed a decision, not made one. Repeat that across customer research, prioritisation, experimentation and stakeholder conversations, and the organisation slowly removes its own training ground.

I do not think the answer is to preserve low-value administration. Nobody needs to spend half a day manually combining notes just because previous generations had to. The answer is to separate the task we should automate from the judgment we still need people to develop.

For example, AI can produce a first synthesis of churn interviews. A less experienced product manager should still listen to some of the original conversations, inspect what the synthesis excluded and decide which explanation deserves further testing. AI can draft a proposal, but the person should be expected to defend its assumptions. AI can calculate and visualise results, but somebody learning the craft should explain whether the outcome is strong enough to change the roadmap.

This matters even more as product experiences become easier to copy. When features can be reproduced quickly, advantage comes from understanding the customer’s situation better than competitors do. That understanding does not come from reading increasingly polished summaries. It comes from repeated contact with inconvenient details: the customer who uses the product differently than expected, the salesperson who has learned an unofficial workaround, or the support issue that exposes a flaw in the original strategy.

Senior leaders are not exempt. The growing concern that technology leaders are too distant from ordinary users is partly a consequence of how information travels upward. Every layer cleans it up. Dashboards remove context, summaries remove hesitation and presentations remove contradictions. AI can accelerate that cleaning process until leaders receive a precise answer to a question nobody should have asked.

In the teams I lead, I want AI to reduce preparation time, not decision exposure. That means giving less experienced people bounded decisions with real consequences. They can own a small experiment, recommend which customer problem to pursue or decide whether evidence is sufficient to continue. A senior colleague can challenge the reasoning without taking the decision back at the first sign of uncertainty.

The review should also happen after the outcome is visible. What did we expect? What happened? Which assumption was wrong? Too many organisations review the quality of a presentation before approval but never revisit the quality of the reasoning afterwards. That teaches people to secure agreement, not to improve judgment.

Product leaders should therefore look beyond delivery capacity when deciding how to apply AI or shape a team. We also need enough learning capacity. If every task is routed to the person or system that can complete it fastest, the experienced people remain experienced and everyone else remains dependent on them.

That may look efficient for a few quarters. It is a poor way to build an organisation that can make good decisions without a small group of senior people in every room.