One of your strongest engineers tells you they are thinking about leaving. What do you do?

Two different conversations, and the order matters

"Thinking about leaving" and "here is my notice" are not the same event and should not get the same response. The first is an invitation, usually a deliberate one — people rarely disclose this to a manager they have written off — and it is the only point at which you have real influence. The second is mostly a logistics and knowledge problem with a small retention tail.

In either case the first conversation is diagnostic and it contains no offers. A manager whose opening move is money, a promotion, or a project reassignment has committed to a remedy before knowing the disease, and the engineer notices that they were handled rather than heard. What you are trying to establish is what changed and when, because almost nobody decides to leave for the reason they state first. The stated reason is frequently compensation and the underlying one is frequently not.

Find out what it actually is

The distinction that governs everything afterwards is whether the cause is something you can change, something you can influence, or something you cannot touch.

Underlying reason	What you can honestly do
Work has stopped being interesting; no growth in sight	The most fixable of all, and the cheapest — real scope on something they have not done before, with a date
Passed over for promotion, or no visible path	Be exact about the gap and the timeline, or admit there is no headroom here and say so
A relationship — you, a peer, a difficult stakeholder	Fixable if it is a structural conflict, sometimes not if it is you; either way naming it is required
Compensation genuinely below market	Influenceable, slowly, and usually only within a band you do not control
Burnout, on-call, or a life change	Load and shape are within your gift; act on them regardless of whether they stay
They want a domain, a location, or a company stage you do not have	Nothing, and pretending otherwise wastes their time

The questions that get you there are open and unhurried: what changed, what would have to be different for this to be the right place in a year, when did you first start thinking about it. That last one is the most informative and the most uncomfortable, because the answer is usually months before you noticed.

Then stop talking and let the silence run. The first answer is the socially convenient one, and the real one arrives afterwards.

Money is a narrow instrument

A counter-offer is legitimate when the cause is genuinely compensation and you have been underpaying someone against your own band. In every other case it buys a few months and leaves the original problem intact, and there is a second cost worth stating in an interview because it shows you have thought past the individual: a raise granted only when someone threatens to leave teaches your team the mechanism by which pay changes here. The people who will not threaten — often the quietest and most loyal — learn that they are paying for their loyalty.

The honest version is also more persuasive than a bidding war. Something like:

"I would rather not have this be about matching a number, because if the number is the only thing that changes you will be having this conversation again in six months. Tell me what the last year here has been like and what you want the next one to look like, and I will tell you honestly what I can and cannot do about it. If part of that is pay, I will find out what is genuinely possible and come back with a real answer rather than a hint."

And where you cannot deliver, say it. "There is no staff opening on this team in the next year, and I am not going to pretend otherwise" costs you the retention and earns you a reference, a boomerang candidate, and a reputation that helps you keep everyone else.

When they are going anyway

Two pieces of work start the same week, and interviewers listen for whether you volunteer them.

The first is knowledge. Notice that you have just been handed a measurement of your own concentration risk: whatever this person is the only one who can do is now visible, and the notice period is the cheapest time you will ever have to fix it. Concretely, that means naming the two or three systems where they are the single point of failure, pairing a named successor onto each rather than asking for documentation in the abstract, and having them walk the team through the failure modes of anything they own on call. Ask for the undocumented things specifically — the manual step before a release, the customer who always calls at quarter end.

The second is the team. People notice a strong colleague leaving before the announcement and draw conclusions from silence, so tell the team quickly, in a way you have agreed with the person leaving, and be honest that you did not want it. Then expect the follow-on conversations, because one senior departure prompts several people to ask themselves the same question. Those conversations are the best information you will get all quarter and the strongest candidates use them, rather than treating attrition as a single event to be absorbed.

The exit conversation itself is worth running properly. A departing engineer has nothing left to lose and will tell you things nobody currently employed will, provided you ask about specifics rather than for a general verdict, and provided you do not argue with any of it.

The part that reflects on the manager

The strongest answers get here without prompting: if a resignation from a key person was a complete surprise, that is a management gap before it is a market event. The signals almost always existed — withdrawal from things they used to care about, a drop in discretionary effort, no longer arguing in design reviews, a sudden interest in cleanly finishing everything, declining the stretch project. Weekly one-to-ones with actual content are what surface these, and a manager who cannot say when each of their engineers last talked about what they want next has no early warning system at all.

There is also a category question worth raising, because it stops this becoming a panic response to every departure. Not all attrition is regretted, and a manager who tries equally hard to keep everyone is not exercising judgement. Some departures are healthy: someone who has outgrown what the team can offer, someone whose strengths no longer match where the work is going. Saying which category this one is in — and why — is the difference between a manager and someone who fears turnover.