Your best people are leaving with the answers.
When an experienced person retires, the organisation loses more than their time. It loses why things are done the way they are, and usually finds out too late.
Every organisation has a few people everyone goes to. The estimator who knows why that client always gets a different margin. The engineer who remembers which supplier’s parts failed in 2019. The contracts manager who can tell you, without looking, which clause will cause trouble on a job like this one. Nobody wrote down what they know, because nobody needed to: they were always there.
Then they retire, or move on, and the questions keep arriving. For a while, people ring them. After that, the organisation quietly starts making the same mistakes they used to prevent, and nobody can quite say what changed.
Why handover projects fail
The usual response is a knowledge-capture exercise: a few weeks of interviews, a handover document, perhaps a wiki. It rarely works. Experienced people don’t hold their knowledge as a list of facts; it comes out when they meet a specific problem. Ask them to write it all down in the abstract and you get a tidy document that misses the parts that mattered.
The timing is wrong too. By the time a handover is scheduled, the person has one foot out of the door, and the organisation is asking them to summarise thirty years in a month.
Keep the reasoning while the work is done
The better answer is less dramatic. Don’t try to extract knowledge at the end; keep it as a side effect of ordinary work, from the start.
Keep the approved version, not just the final file. When an expert approves a price, a report or a response, that approval and the evidence behind it are kept together. The next person can see what was decided and on what basis.
Keep the corrections. The most valuable knowledge is often the edit: the paragraph an expert rewrote, the figure they changed, the clause they struck out. Those changes say more about judgement than any interview.
Make it findable by the next person. Past work is only useful if a newcomer can ask a question in plain language and be shown the relevant decisions, from evidence they are entitled to see, with the sources attached.
You can’t interview thirty years out of someone in a month. You can stop letting their judgement evaporate every day they work.
This is the practical side of an argument we made in an earlier paper: an organisation’s accumulated judgement is worth more than any model it rents. AI makes that judgement easier to use than ever before, but only if it was kept in the first place. The organisations that keep it will still have their best people’s answers long after those people have gone. That is what we mean by knowledge that stays on the record.