Institutional knowledge is most visible when it disappears. A senior operator leaves, an unusual case arrives, or a customer asks why a decision was made. The organization discovers that the real procedure lived in memory, private messages, and corrections applied after the official process ended.
Capture decisions, not just documents
Policies often describe the expected path while expertise lives in knowing when the path does not fit. Ask experienced people to walk through real cases: an ordinary success, a difficult exception, and a mistake they would now avoid.
For each decision, capture the facts considered, rule or principle applied, available alternatives, authority used, and evidence retained. This reveals reusable knowledge rather than merely recording a story.
Separate rule from judgment
Some knowledge can be deterministic: thresholds, required records, prohibited actions, and escalation timing. Other knowledge remains judgment: assessing credibility, balancing risk, or recognizing a novel pattern.
A strong model preserves both. It encodes stable rules and creates a governed place for human judgment, including who may exercise it and how the rationale is recorded.
Make knowledge operational
A handbook may be searchable yet remain disconnected from work. An operating model places knowledge where decisions occur. Required context appears with the task; constraints become validation; exceptions route to the right authority; outcomes create evidence.
Imagine a veteran estimator explaining how site access, material volatility, schedule compression, and client history change contingency. Those factors can become structured parameters while the estimator retains authority over nuanced assessment.
Keep provenance and change
Record who approved each rule, when it became effective, and which examples support it. Let new evidence challenge the model through review rather than silent workaround. Institutional knowledge remains valuable because it can evolve without losing its history.
Model conversion exercise
Use this exercise with a real workflow and the people accountable for its outcome. Record disagreements as modeling questions instead of silently choosing an answer.
Ask an experienced operator to teach three cases: routine success, difficult exception, and memorable failure. At each decision, pause for the facts considered, alternatives rejected, principle applied, authority used, and evidence retained. Ask a second operator to predict the same decisions. Differences expose tacit judgment and local habits. Encode stable agreement as rules; preserve nuanced judgment as an authorized review point with rationale. Link every element to its source example and approver so future teams can understand not only what the model says, but why.