Two Weeks to Download a Brain: Knowledge Capture When Someone Resigns

The resignation letter is gracious, the notice is the standard two weeks, and somewhere in the back of your mind a meter starts running: fourteen working days to capture whatever's in that head that exists nowhere else. The standard response is to assign “documentation” — write down everything you know before you go — and the standard result is a document that reads like a table of contents for a book that was never written. Not because the person didn't try, but because the assignment is impossible: experts are reliably the worst judges of what others don't know. The things most worth capturing are precisely the things they no longer notice they know.

If you ran the vacation test, you already know which questions only this person can answer — and if you didn't, their notice period is the vacation test running with real stakes and a hard deadline. Either way, the two weeks reward a completely different method than “go write.”

Day one: triage, not documentation

Spend the first hour making a list, not a doc — and make it with them plus the person or people inheriting the work. Three columns: what only they know (processes no one else has run, relationships no one else holds, systems no one else has touched); what they know best but others know somewhat; what's already written down or reconstructible. Column one is the whole game — columns two and three get whatever time is left, and usually should get none. The fastest triage question, asked of the leaver directly: “What's going to break in the month after you leave, and what will nobody even know to check?” People answer that question with startling accuracy, because they've been quietly carrying the list for years.

Capture by interview, not by assignment

Knowledge comes out of experts through questions, not through blank pages — it's the same reason tribal knowledge never wrote itself down in the first place. So the successor (not the manager — the person who will actually do the work) runs short interviews against the column-one list, one topic at a time: walk me through it; what goes wrong; what did you try that didn't work; who do you call, and what does that person care about? The successor writes as they go — naturally recording what a newcomer needs, because they are one, and skipping what's obvious, because it's obvious to them too. Record the calls; the transcript catches what note-taking drops. The leaver's job is to talk, correct, and point at things — a division of labor that produces in an hour what solo brain-dumping doesn't produce in a week.

Watch them work the real thing

For anything operational in column one — the month-end run, the renewal dance with the difficult vendor, the system only they've administered — add a watch-them-work session: the leaver drives once while narrating, then the successor drives while the leaver watches. The second half is where the real knowledge surfaces, because experts narrate the happy path and only remember the landmines when someone is about to step on one. “Oh wait — before you click that…” is the sentence the whole session exists to trigger. Relationship handoffs get the same live treatment: a joint email or call introducing the successor to the key contacts — the vendor rep, the partner, the customer who only trusts them — because relationships transfer warm or not at all.

The handoff test

Here is the standard that separates real capture from documentation theater: before the last day, the successor performs each critical task without the leaver's help — docs allowed, questions not — while the leaver is still around to repair what fails. A green run means the knowledge actually landed. A stumble means you found the gap while it costs a conversation instead of a crisis. This is the same completion bar as everywhere else in your operation: done is demonstrated, not described. Whatever the test produces goes into the knowledge base as real articles — owned, dated, findable under the words people will search — not into a “Departure — [Name]” folder, which is where handoff docs traditionally go to die unread.

The bottom line

Two weeks is enough — but only for the right method. Triage to what only they know; extract it by interview and observation, with the successor doing the writing; hand off relationships warm; and prove the transfer with a handoff test while repair is still cheap. Skip the impossible brain-dump assignment entirely. The person leaving is usually glad to help — nobody wants their years of judgment to evaporate at the door. Give them a method that can actually catch it.

— Tom