Ask the expert: "what do you notice that a beginner wouldn't?" — this extracts what no manual ever captures.

Convert your daily meeting to 15 minutes standing, three questions per person, then dismiss.

For each task estimate: optimistic, most-likely, pessimistic — then compute expected.

Assign the model a specific background and expertise so its answer matches a real expert level.

Have the model play expert + student who asks "dumb" questions — extracts tacit knowledge.

Before relying on AI in specialist field, show to human expert — 30% of time uncovers core error.