Keep the Learning Seat in the Meeting

The product review ends in 18 minutes. The brief was prepared by AI, the senior people resolve two disagreements, and the final direction lands in the project tracker before the next call begins.

By ordinary measures, it is an excellent meeting.

But the newest person on the team was not invited. There was nothing for them to present, and the recap would tell them what was decided. What the recap cannot fully show is how an experienced colleague recognized the risky assumption, why the obvious option lost, or which small hesitation changed the room.

That is the hidden curriculum of work. As AI absorbs more drafting, research, and coordination, teams need to protect some access to the judgment that surrounds those tasks. One practical way to do it is to keep a learning seat in selected meetings.

The early-career warning is getting harder to ignore

In August 2026, Stanford's Digital Economy Lab published a revised analysis of payroll data covering millions of US workers through June. The researchers found no evidence of widespread economy-wide job displacement. They did find a sharper pattern for people aged 22 to 25: employment in AI-exposed occupations was 19% below where it would have been if it had kept pace with less-exposed peers. Experienced workers showed no comparable gap.

This is an employment study, not a study of meetings. It does not prove that removing junior employees from calls caused the gap. The useful implication is more modest: organizations should pay attention to how people enter expertise when some of the traditional entry-level work is automated.

Another recent revision makes the learning mechanism easier to see. Researchers studying software engineers at a Fortune 500 company found that sitting near teammates increased coding feedback by 18.3% and improved code quality. The gains were concentrated among younger and less-tenured employees. There was also a real cost: experienced engineers wrote less code when they sat near teammates.

The authors describe a tradeoff between current output and training for the future. That evidence comes from software engineering and physical proximity, so it should not be stretched into a universal rule about office attendance. Still, it captures something managers recognize. Learning often arrives through small, timely exchanges, and providing those exchanges consumes senior attention.

AI can make the current-output side of that tradeoff look irresistible. A concise summary reaches everyone. A draft appears before a junior colleague could have produced it. Fewer people are needed in the room. The danger is not that every streamlined meeting becomes harmful. It is that a whole layer of observation, questioning, and feedback can disappear without appearing on any efficiency dashboard.

A recap preserves outcomes, not expert attention

A strong recap can record the decision, owner, deadline, and open question. That is valuable. It is also different from witnessing someone work through uncertainty.

Consider a pricing discussion. The final note might say, "Keep the current plan structure and test annual billing copy." A learner in the room may notice much more:

• The finance lead distrusts a clean-looking forecast because the sample is too small.
• The product lead changes position after hearing one support example.
• The team treats a reversible copy test differently from an irreversible packaging change.
• Nobody has enough evidence to settle the larger question, so the group deliberately makes a smaller decision.

Those are not decorative details. They are how people learn what counts as evidence, when to challenge a confident answer, and how the organization handles risk. A transcript may contain every sentence and still leave that structure implicit.

Design a learning seat, not an extra attendee

The answer is not to invite every junior employee to every meeting. That creates calendar debt and turns development into passive observation. A learning seat is a specific role in a meeting with unusually high learning density.

Start with one recurring forum: a customer debrief, incident review, hiring calibration, roadmap tradeoff, or negotiation retrospective. Choose a meeting where experienced people interpret imperfect information, not a status call where everyone reads updates aloud.

Then give the seat a simple five-part structure.

Name the learning target

Before the meeting, tell the learner what to watch for. "Notice how we distinguish a customer request from evidence of a broader need" is useful. "Come see how leadership thinks" is too vague.

The target should be a judgment skill, not a topic. Examples include weighing contradictory signals, separating reversible from irreversible choices, spotting missing stakeholders, or turning an ambiguous concern into a testable question.

Narrate one hinge

During the discussion, the meeting lead pauses once to make hidden reasoning visible:

We are choosing the slower option because the fast one depends on an assumption we cannot verify before launch. If the deadline moved, that tradeoff might reverse.

This takes less than a minute. It reveals the evidence, uncertainty, and condition that shaped the choice. The learner gets more from one precise hinge than from ten minutes of generic career advice afterward.

Create a question window

People early in their careers often need time to formulate a useful question, especially in a fast room with senior colleagues. Reserve two minutes near the end. Let the learner ask aloud or submit a question privately to the meeting lead.

A good prompt is, "Which assumption would you verify before making this decision again?" Another is, "What did the team notice that was not obvious in the pre-read?" These invite explanation without forcing the learner to perform certainty.

Assign a reversible next step

Observation becomes learning when the person has to use the judgment. Give the learner a contained follow-up: draft the decision note, test one assumption, compare two options, or prepare the first version of the next brief.

Keep the stakes appropriate and make review explicit. The point is not to hand over an irreversible decision before someone is ready. It is to create a short loop between seeing, trying, and receiving feedback.

Preserve the decision trail

Record more than the winning option. A useful decision trail includes the alternatives considered, the evidence that mattered, the uncertainty left unresolved, and the condition that would trigger a revisit.

This should remain concise. The goal is not permanent surveillance or a perfect transcript. It is a reusable example of how the team reasoned, available when the learner faces a similar situation months later.

Protect both sides of the tradeoff

Senior time is not free, and a learning seat will fail if it quietly becomes another obligation with no boundary. Rotate the seat. Limit it to meetings with real judgment content. Ask the learner to decline when the stated learning target is not relevant.

After four meetings, look for transfer rather than attendance. Can the learner anticipate the tradeoffs the team will raise? Can they draft a decision note that distinguishes fact from assumption? Do they ask sharper questions or catch a contradiction earlier? If nothing changes, redesign the practice instead of adding more meetings.

Also watch for extraction disguised as development. Taking notes, scheduling follow-ups, and cleaning data can support learning, but only when the person receives context, feedback, and a path toward more substantive responsibility. A seat at the table is not an apprenticeship if the learner is only there to service the table.

Efficiency should make room for apprenticeship

The best use of AI is not to preserve every old task. Some repetitive work deserves to disappear. But when a task vanishes, teams should ask whether it also carried observation, feedback, or a first chance to exercise judgment.

Meetings are one place to rebuild those pathways deliberately. With a narrow learning target, one narrated hinge, a question window, and a reversible follow-up, a short discussion can do double duty: make today's decision and prepare someone to make tomorrow's.

Caspi supports meetings with live recap, suggested questions, contextual chat, proactive flags from connected tools, post-call action items, and persistent meeting memory. Those capabilities can reduce the burden of capturing what happened. The more important opportunity is to use that recovered attention to notice how judgment is being formed, and to make sure the next generation of colleagues can learn it.