"What If I Train Them and They Leave?" Here's the Math Nobody Runs

There are four possible outcomes when you consider developing someone. Leaders spend nearly all of their emotional energy on one of them.

	They stay	They leave
You develop them	The outcome everyone pictures when they picture this working. More common than fear allows.	The box that haunts you. Real, costly, and the only one you can see.
You don’t develop them	The box that feels safest and quietly costs the most.	The box nobody counts. They leave anyway — often because nothing was growing.

Look at where the fear lives. One box out of four. And the choice it pushes you toward — the bottom row — contains the two outcomes that are actually worse.

The form that stayed open for eleven days

A manager I’ll call Camille had an approval form open for eleven days before she closed it without signing.

It was a modest request. Terrence, who ran the scheduling desk and had quietly rebuilt half the intake process on his own initiative, wanted the company to cover a project-management certification. Six weeks, evenings, a few hundred dollars.

Camille liked Terrence. She trusted his judgment more than she trusted some people two levels above him. And every time she opened that form, she thought about Alina.

Two years earlier she’d done exactly this for Alina — paid for the certificate, rearranged the schedule, let her lead a live project as a capstone. Alina had been terrific. And five months after finishing, she’d used that credential to land a role somewhere that paid better.

Camille said all the right things at the goodbye lunch. Privately she felt something closer to betrayal.

So the form sat open, and closed, and open again. Camille told herself she was being prudent. What she was actually doing was protecting herself from a repeat of Alina by making sure Terrence never became someone worth losing.

She wasn’t a bad manager. She was a wounded one, doing arithmetic that felt like wisdom and was actually fear.

Let’s do the arithmetic for real

The fear presents itself as financial. So let’s treat it financially.

Gallup estimates that replacing an employee costs somewhere between one-half and two times their annual salary — and they describe that as conservative.

Take Terrence at $65,000. Replacing him costs somewhere between $32,500 and $130,000.

The certification Camille wouldn’t sign cost a few hundred dollars and six weeks of evenings.

Even under assumptions maximally hostile to development — assume the training causes the departure, assume it happens immediately, assume you capture nothing while he’s there — you would have to lose roughly a hundred Terrences to that certification before it cost you what losing one of them to boredom does.

That is not a close call. It isn’t close by two orders of magnitude.

And Camille, who is good at her job and reads a P&L every month, could not see it — because the thing being weighed was never money.

Why the fear is louder than the math

Losses register more intensely than equivalent gains. That’s the core of prospect theory, Kahneman and Tversky’s 1979 work.

(One honest caveat, since business writing routinely overclaims here: you may have heard losses feel “about twice as large” as gains. Recent meta-analytic work suggests that ratio is far less stable than it’s usually presented. Take loss aversion as a real tendency, not a constant you can calculate with.)

Even the modest version explains Camille perfectly. Losing Alina registered as sharp and specific — a name, a date, a goodbye lunch, a story she could tell. The slow cost of an undeveloped Terrence five years out — the ideas he stops offering, the ceiling he quietly accepts — registers as almost nothing, because it has no single moment attached. It just feels like Tuesday.

But it is a loss, distributed across years rather than concentrated into an afternoon. And distributed losses are exactly what loss-averse minds are worst at noticing.

The box that feels safest

They stay. And something in them stops.

They stop bringing ideas, because ideas that go nowhere start to feel foolish to keep offering. They stay dependent on you for decisions they could have learned to make. And you experience that dependence as a burden rather than recognizing it as something you built.

You get years of a person’s presence without years of their growth, and you call it stability.

It is not stability. It’s a slow leak you’ve decided not to look at.

What’s changed since you first heard this question

Two things, and neither is small.

The value of a specific skill is now genuinely uncertain. The old version of this question asked whether the person would stay. The new one also asks whether the skill will. That’s a fair question — but be careful, because it can become a sophisticated new costume for the same old fear. Two tests: Would I still want this person to have this capability if I knew they were staying forever? And: Am I declining to develop, or declining to develop this specific thing? Only the second is a judgment.

And the ready-made talent pipeline is thinning. Researchers at Stanford’s Digital Economy Lab, working from payroll data on millions of American workers, found that workers aged 22 to 25 in AI-exposed occupations sit about 19% below expected employment — driven almost entirely by reduced hiring, not layoffs. Employers are opening fewer entry-level doors.

Think about what that means if you’ve already hired someone. The supply of people trained at somebody else’s expense is not getting deeper. It’s getting shallower. Which makes the ability to develop the person already at your scheduling desk less of a nicety and more of a competitive position.

Ten minutes, on paper, before you decline

1. Which box am I actually choosing? Mark it. Write one sentence describing what that box looks like three years from now.

2. What does turnover cost me here? Half to twice annual salary. Write the number. Compare it to the request. Say the ratio out loud.

3. What do I get while they’re still here? List three things that get better in the next ninety days. If you can’t list three, the request may genuinely be wrong — and now you have a real answer to give them.

4. What am I actually afraid of? For your eyes only: I’m hesitating because I’m worried that ______.

5. The reversal test. If I knew this person was staying five years, would I approve it? If yes, my objection is about departure, not about the request. Departure is not a reason. It’s a fear.

6. Say it out loud. Whatever you decide, tell them the real reason within a week. A soft no that both of you understand means never costs more than an honest no ever will.