Questions to ask an employee before they leave

Someone resigns, and the plan is that they will “write everything down before they go”. They will not. Not because they are unwilling. Describing a job you have internalised over years is a different skill from doing it — and they are still doing the job, for two more months, while handing over live work. What comes back is a folder nobody opens.

The alternative is to ask instead of asking them to write. Below is what to ask, when to ask it, and how to tell which answers you can actually rely on.

When should the handover start?

The day the resignation is accepted. Not the final week.

The last week is spent closing live work, saying goodbye and chasing loose ends. Anything you ask then gets a rushed answer, and rushed answers omit the caveat that mattered. Four short conversations across the notice period beat one long session on the final Friday. They also let you come back with a second question once you have understood the first answer.

There is also an ordering to it. Start with the parts of the job that only they do — knowledge held by one person is the knowledge that leaves with them. The parts shared with colleagues can be reconstructed later; the parts that are theirs alone cannot.

What actually walks out of the door?

Not the procedure. The procedure is usually written down somewhere, out of date but recoverable. What leaves is five things, and it helps to ask for them separately because people volunteer the first and forget the rest.

• The sequence and its dependencies — not the list of tasks but the order they must happen in, and what breaks when one runs early.
• The judgment calls — where they choose between options by reading signals rather than following a rule. When to escalate, which exception is safe, what is worth chasing.
• The traps — what looks right, goes wrong, and is known only to people who have been caught by it once.
• The reasons — why a rule exists, which is what lets the next person depart from it safely instead of “improving” it into a failure.
• The people and places — who actually approves things, which of the four documents is current, who to call when a system misbehaves.

Those five categories are the nugget types Nuggetz uses, and they work just as well as a checklist for a human-run handover.

What questions should you ask?

Twenty questions, grouped. Ask them in a conversation, and follow up on anything that sounds rehearsed. “Usually”, “normally” and “it depends” are signposts. Each one points at knowledge nobody has written down.

The shape of the work

1. Walk me through a normal week. What is on a fixed schedule and what arrives unannounced?
2. Which parts of your job does nobody else do?
3. What do you do that is not in your job description?
4. Which parts of the written process do you do differently, and why?
5. What would stop working the week after you leave, if nobody picked it up?

The judgment

6. What decisions do you make that you have never had to explain to anyone?
7. When you have to choose between two options here, what tells you which one is right?
8. What is the threshold — the number, the timing, the size — at which you change approach?
9. When do you escalate, and who to?
10. What would you push back on if you were asked to do it, and why?

The traps

11. What goes wrong most often, and how do you know it is about to?
12. What looks fine but is not?
13. What has broken badly at least once, and what did you change afterwards?
14. Which system or report do you not fully trust, and what do you check it against?
15. What takes far longer than anyone expects?

The people and the reasons

16. Who do you contact when you are stuck, and what do they usually tell you?
17. Who needs to know when something changes, that is not obvious from the process?
18. Which rule here looks arbitrary but exists for a reason — and what is the reason?
19. What did you get wrong in your first six months that you wish someone had told you?
20. What have you never been asked about, and never written down?

Question 20 is the one to keep for last and not skip. It is unreasonably productive, because it asks someone to notice their own blind spot rather than answer inside yours.

How do you know the answers are right?

You ask somebody else the same questions.

This is the step most handovers skip, and it is the one that turns a transcript into something a successor can rely on. Ask two colleagues the same five questions about the same process and one of three things happens. They agree, and you can be reasonably confident. One knows something the other does not, which tells you who to ask next. Or they disagree. Two experienced people apply different thresholds, each certain they follow the policy, neither ever asked — and that is far more common than anyone expects. A disagreement found during a handover is a finding. The same disagreement found six months later is an incident.

Whatever you do, do not resolve a disagreement by picking the more senior person’s answer and writing that down. Record both, name who said what, and mark it unresolved until somebody with the authority to decide actually decides.

What do you do with the answers?

Write them into something their successor follows, not a transcript they have to read.

Three rules make the difference between a document that gets used and one that gets archived. Keep every instruction traceable to who said it, so a reader can weigh it rather than take it on faith. Only let settled knowledge instruct anyone — anything two people described differently is not settled enough to tell a third person to do. And list what is unresolved at the end rather than leaving it out. A document that hides its own uncertainty is the more dangerous one, because silence reads as “this is fine”.

That is the shape of a handover worth having. Run it in a meeting, in a shared document, or as a knowledge interview that asks the questions and writes the answers up for you. The hard part does not change: ask for the judgment, and ask more than one person.