Capture expert judgement
Tacit knowledge: how to capture what your experts know but rarely write down
Tacit knowledge is the know-how someone uses without fully spelling it out. An expert may recognise an unusual client request immediately, then struggle to explain how they knew. You can make parts of that judgement easier to share by examining real decisions, recording their clues and testing whether another person can use them.
Capture the clues behind a decision, including exceptions and uncertainty, then let another person test the explanation on a different example.
Look for judgement hidden inside ordinary work
A documented procedure might say “review the client report”. The expert also notices an odd comparison period, recognises a client’s preferred level of explanation and knows which figure needs checking before a meeting. Those details can be hard to retrieve from the procedure alone.
Some tacit knowledge comes from practice, such as interpreting tone in a client conversation. You will not turn all of it into a reliable set of rules. The practical aim is to preserve the explanation, examples and limits that help another person make a better decision.
Choose a decision that repeats and matters. Avoid opening with “tell me everything you know”. That invitation can produce a fascinating afternoon and very little that a colleague can use on Monday.
Use a real example instead of a general description
Ask the expert to show the last time they made the decision. Look at the source material together and ask what they noticed first. Then follow the order of their checks, including the apparently obvious ones.
- What told you this needed a different response?
- Which detail changed your first impression?
- What did you check before deciding?
- What would have made you choose the other route?
- When would you need to ask another expert?
Record the evidence as well as the explanation. “This client always wants more detail” is an impression. A current reporting agreement or a reviewed client decision provides something another person can inspect. Keep the source separate from the date that requirement became valid.
Compare similar cases with different decisions
Find a second example that looks similar but led to a different action. Ask what mattered between the two. The contrast often reveals a condition the expert left out of the first explanation.
Said describes reporting as a difficult process to map because every client’s requirements can take the work down different routes. Capturing one polished report can miss that judgement. Compare the iterations to find the shared method and the client-specific decisions.
| Capture | Example of what to record |
|---|---|
| Clue | The comparison period does not match the client agreement. |
| Check | Open the current reporting requirements and source-period record. |
| Decision | Hold the explanation until the correct comparison is confirmed. |
| Exception | A changed agreement needs the account owner’s review. |
| Limit | This example does not establish every client’s reporting requirements. |
Ask someone else to use the explanation
Give a colleague the captured method and a different approved example. Ask them to explain their decision, its supporting evidence and what remains uncertain. The expert reviews the reasoning, including where the colleague correctly stopped.
If the colleague needs another question answered, update the record with that question and its source. Repeat with a contrast case. Success is an acceptable decision in the chosen job, rather than agreement that the notes look comprehensive.
Give the explanation an owner and a review trigger. A client requirement, process or system change can make an old rule misleading. Our current build returns reviewed human decisions to its knowledge records; that feedback updates the business context, without necessarily retraining a model.
Keep the interview connected to a decision
Worked example
Capturing a reporting review decision
Invented teaching example. No client outcome is claimed.
OBSERVED CLUE
An invented report compares a full month with a partial month.
EXPERT CHECK
Compare the period labels with the current client reporting agreement.
DECISION
Hold the explanation until comparable periods are established.
CONTRAST
A second report uses comparable periods but a changed metric definition; check that definition separately.
TEST
A colleague must identify the right check in a third example and explain its source.
A correct stop for missing information is acceptable.
Further reading
Written by Said Lopez, founder of LeverageAI and author of Lean AI. Our guides draw on the method in the book and our work with clients.