Signs an Employee Is About to Quit, and the Conversation to Have Before They Go

"Hey, is everything okay? You have seemed a bit checked out lately."

It is meant kindly and it lands as a verdict. You have told your report that you have been watching, that you have reached a conclusion, and that you would like reassurance. The only socially available answer is "yeah, sorry, just tired." The signs an employee is about to quit are usually visible weeks before that sentence gets said.

So they say that. You say okay. The conversation that mattered does not happen.

The signals that actually predict it

Every signal worth trusting is a version of one thing: discretionary effort going away. Not output, which usually holds for months. The work nobody assigned.

They stop arguing. The person who used to push back on the roadmap now says "sounds good" to everything. Disagreement is expensive, and people spend it on a future they expect to be part of.

They stop planning past a horizon. Vague about the thing that lands two quarters out, happy to let somebody else own the piece that would define their next year.

They stop recruiting. They used to send you names and talk the place up to friends. That goes early, because it is the part of the job that takes believing in it.

Their questions change shape. Fewer answers in the channel to questions that were not theirs, and their own move from how do we fix this to who owns this now.

None of these is proof. Each has innocent explanations behind it: burnout, a sick parent, a project they quietly hate, a new baby. What makes a signal worth acting on is never the item. It is the direction, over weeks, in somebody whose normal you know.

Disengagement is not underperformance either. If the work has dropped below the bar you are having a performance conversation, which is a different meeting.

The ones that are only noise

The lists that circulate are attendance trivia. Started dressing better. Blocked time on the calendar. Single days of PTO. Logging off at 5:30 three days running. Camera off.

All of that measures presence, not engagement. Half the time the mystery block is physical therapy.

The real damage is what reading those signals does to you. Once you go looking for evidence you will find it, and then you manage the person you have decided they are. They feel it. People always feel it. Start checking one person's hours, or reading a calendar you would not read for anyone else, and you are running surveillance. Being wrong there is expensive. You will have told somebody who was staying that you watch them and assume the worst, which is its own resignation event.

A clean rule: if you would not look at it for the whole team, do not look at it for one person.

By the time you notice, the decision is mostly made

Something happened first, months ago. They got passed over. A reorg moved the work they cared about. A promise quietly did not happen. A manager they liked left.

Then they sat with it. Then they said it out loud to one person outside the company. Then they updated the resume, not seriously, just to have it. Then they took a call out of curiosity and found it more interesting than expected.

The withdrawal you can see comes after all of that. It is not the start of a decision. It is the residue of one already made. Which has two consequences.

Stop trying to win a single conversation. Somebody most of the way out will not be argued back inside thirty minutes, and pressing for a commitment makes them defend the exit out loud, which makes it more real. Aim for information instead: what happened, and whether anything you can genuinely change would move it.

Move the conversation earlier, permanently. The only variable fully in your control is when you have it.

The stay conversation, and why it is not a counteroffer

A stay conversation is thirty minutes, on the calendar, about why somebody is here and what would make them leave. Not a performance review, not a retention pitch, not a reaction to a resignation. It happens when there is no crisis, which is the only time the answers are worth anything. Two rules make it work.

Everyone gets one. Run it only with the person you suspect and it is an interrogation with a friendly name, which they will read correctly inside a minute. Run it with everyone twice a year and it is just how you manage.

Ask about the past, not the future. "Are you thinking about leaving?" has one socially available answer. "In the last year, was there a moment when you thought about leaving?" is a question a committed person can answer honestly, because the moment has already passed.

That last move is the one that matters. A stay conversation that produces a list you never act on is worse than no conversation, because you have proved the answer changes nothing.

What actually keeps people, and why they name money

Roughly in order of how often it is the real reason:

1. Whether the work is going somewhere they want to go, and whether they believe the next step exists.
2. Their manager. Whether they get told the truth and defended in rooms they are not in.
3. The people around them, and whether the work those people do is any good.
4. Autonomy. Whether they can do the job well without fighting the organization for permission.
5. Money.

Money is almost never at the top, and it is very often the reason people give. They are not lying to you. Money is the only reason that is precise, safe, and about nobody in the room. "I got a better offer" is a clean exit. "I stopped trusting you in March when you did not back me in the planning meeting" is not a conversation anyone wants to open on the way out.

So treat money as the headline, not the story, and test it with one question.

If that comes back fast and yes, money was never it. If it comes back slowly, you have found the real thing.

When they already have an offer

A counteroffer buys time without changing whatever made them look, and it rewrites the relationship. You now know they were interviewing. They now know the reliable way to get paid here is to interview elsewhere. Your team knows both within a month.

A bidding war is a conversation you have already lost. The terms are set by somebody who is not in the room, and every compensation conversation you have with that person afterwards gets compared to the one time you moved fast because you were about to lose them. The honest version works better.

That last line is what makes the rest credible, because it is the only sentence in the script that costs you something. And if you do correct a salary, correct it because it was wrong, then check who else sits in the same gap. Fix it only for the person holding an offer and you have published a policy.

When they leave anyway

Most of the time they leave anyway, and the number you were managing to was never zero.

What you are really managing in the last four weeks is everybody else. The team is watching how a person gets treated on the way out and working out what would happen to them.

The first ten seconds are the whole reaction. Whatever you say when they tell you is what gets repeated, to the team, to their friends, and back to you years later. "Congratulations. Tell me about it." Not a wounded pause, not "wow, okay."

Do not go cold, and do not shrink the job. The common failure is the manager who is gracious for the announcement and distant for the month after, pulling them off the good work and dropping them from planning. Everyone sees it and nobody says anything. Give them the handover as real work with a real deadline instead.

Get the truth later, not in the exit interview. It collects the polite version, because they still need the reference. Ask two months after they have gone, when there is nothing left to protect.

And keep count. One person leaving is a person. Two leaving for the same reason is a fact about the team, and usually about the organization or about you.

This does not work if they have already signed. Once there is an offer on paper, the conversation stops being about what is wrong and becomes a counteroffer, and counteroffers mostly buy months, not years. We did not measure that here: it is the reason to have the conversation in the quiet weeks instead.

Knowing the questions is not the same as asking them

The questions above took two minutes to read. Asking them is different work.

You have to invite somebody to tell you what is wrong with working for you, and then sit still while they do it. No context, no small correction that makes you look more reasonable, no version of "well, to be fair." The reflex to defend yourself arrives two seconds in, and it arrives in your voice before you notice it in your head.

That is a live skill, not a reading skill. It is the same one that makes a one-on-one produce the second answer instead of the first, and the same one that carries a hard conversation with a report.

You find out whether you have it the first time somebody says "honestly, I stopped believing there was a next thing for me here about eight months ago," and you have half a second to decide whether your next sentence starts with "that is not really fair."