Coach before you answer

Someone brings you a problem in your one-to-one and you already know the answer. Saying it costs you ten seconds.

Asking costs you five minutes, an awkward silence, and the urge to just tell them before you get there. What they walk out with, though, is a decision they made themselves. That's a different thing entirely from a decision you handed them, even if the two decisions look identical on paper.

Coaching is the slower path, taken on purpose. Not because slow is a virtue. Because someone who works an answer out keeps the muscle that produced it, and someone who's handed the answer keeps only the answer, which won't be there next time the situation is slightly different.

That doesn't mean questions are always the right move. Getting that call wrong in both directions is the actual skill here, more than any technique for running a good conversation.

Four questions carry most of it

You don't need much structure for this, and the setups that get over-engineered tend to stop happening at all. Four questions cover most of it, roughly in order, though real conversations wander and that's fine.

What are we actually trying to accomplish. People show up with a complaint or a half-formed frustration more often than a clean goal, and just saying the goal out loud is sometimes half the conversation.

What's actually happening. Facts first, then the assumptions sitting under them, then whatever's not being said about how it feels. Most people jump straight to solutions because reality is the uncomfortable part, and it's usually where the real information lives.

What are the options, and what does each one cost. Not the option. A couple of genuinely different ones, each with its trade-off named. Handing over a single sanctioned choice is just you deciding again with extra steps.

What will they actually do, by when, and what do they need from you. Skip this and the conversation was interesting and changed nothing.

Goal, reality, options, way forward. None of that is complicated. Running it well is the hard part, and that's most of this lesson.

Knowing when to stop asking

Here's the failure mode nobody warns you about: turning every conversation into a Socratic exercise no matter what the person in front of you actually needs.

Someone who's missing a fact doesn't need three open questions circling it. They need the fact. That's it. Someone about to make an expensive mistake because nobody ever told them the constraint doesn't need to discover it through inquiry, either. They need to hear it, plainly, this once.

I used to treat "always ask, never tell" as the mark of a good coach, on the theory that telling was a shortcut and shortcuts were cheating. That's backwards. The real skill is diagnosing which mode the moment calls for. Coach when someone already has the judgment and just needs room to use it. Teach when they're missing information you have and they don't. Direct when the stakes are high, the clock is short, or the decision genuinely isn't theirs to make. All three are legitimate moves. What isn't legitimate is picking one and running it on everyone, every time, regardless of what's actually in front of you.

The test is simple even when applying it isn't: whether they can make the next version of this decision without you. If questions get them there, ask. If it's just going to burn twenty minutes before you tell them anyway, tell them, and say why.

Running it without turning it into an interrogation

Start by saying the purpose out loud. "I want to spend ten minutes on the retro feedback" beats drifting into it sideways, because the person knows what they're walking into and can bring something real instead of guessing what you're fishing for.

Ask open questions, then let the silence sit there. Both matter. The second one's harder. Most managers ask a good question and then, three seconds in, rescue the person from the silence with a hint or a follow-up or the answer itself. Those three seconds were the valuable part. Let them think.

Slower than answering, and the first several times it'll feel like you're bad at your own job, watching someone flounder toward something you already have sitting right there.

Listen for what's actually being said, not for the gap where you get to respond. Curiosity, not your internal checklist of what you'd have done differently. Reflect it back in your own words before reacting to it. "So the real blocker is you don't trust the estimate, not the estimate itself." Half the time that one sentence does more work than anything after it, because it's the first time they've heard their own situation said back to them cleanly.

Ask before you shift from questions to advice. "Want a thought, or do you want to keep working it?" is one sentence, and it changes everything about how what comes next lands. Advice someone asked for gets used. Advice that just arrives gets defended against.

Challenge the pattern, not the person, when something genuinely needs challenging. "You've said this same thing about three different projects now" is a fact you're both looking at together. "You always do this" is an accusation, and it produces defensiveness instead of insight even when it happens to be true.

Call out real progress and real courage when you actually see them. Specific enough that it's obviously not a management reflex. And close on a commitment with a date on it, not a vague sense that the conversation went well. Vague conversations feel productive and change nothing at all.

Before it ever reaches a review

The best use of coaching has almost nothing to do with problems. It's ordinary work, caught early, run as a two-minute exchange instead of a scheduled event.

Nothing in a formal review should ever be new information, and coaching is the mechanism that makes that actually true. A pattern shows up, you say something that same week in five minutes, they adjust or push back and you learn something either way. By the time anything would count as review material, it's already been handled, or it's already been said out loud more than once. Save it for a scheduled conversation months out instead, and a small correctable thing turns into a surprise that damages trust on top of whatever the original problem was.

Growth nobody can put into words yet

Most engineers don't know what they want next, and treating that as a gap in them, rather than a normal starting point, changes how the whole conversation goes.

Skip "where do you see yourself in five years." Nobody has a real answer to that one, only a rehearsed one. Ask what skills they actually want more of. What kind of problem gives them energy versus the kind they're good at but dread. What meaningful progress would look like six months out, specific enough that you'd both recognize it if you saw it. Those questions produce something real, because they're answerable from the work someone's already doing, not from some hypothetical future self.

A goal that lives only in a development document loses to whatever's on fire this week, every single time, because urgent delivery comes with a deadline and growth doesn't, unless you give it one. So tie it to real work. If someone wants technical leadership, make them the one running the next incident review, not the person waiting on a training course that keeps sliding a quarter. Specific, achievable without depending on someone else's calendar, checked in the same one-to-one that covers everything else, and revisited out loud if priorities genuinely change instead of left to quietly expire.

For someone working toward a level change, write the review they'd want to receive at the end of it. What they'd have shipped. What they'd have learned. How they made the people around them better. Then work backward into the actual opportunities between here and there. And say plainly that the plan isn't a promise. It's a path, and you can walk a path without a guaranteed destination, but only if nobody's pretending otherwise.

When they don't know if they even want it

Some people genuinely can't tell whether they want to manage or stay technical, and no amount of talking resolves that. Only doing the thing does. Reading about it doesn't count.

Build a real, small, reversible version of it. Mentor someone for a quarter. Run a project with three people on it. Lead the planning session. Own one technical decision end to end and live with what it costs. Then debrief what it actually felt like, not what they expected it to feel like beforehand, because those two are reliably different, and the gap between them is the information you're actually after.

Becoming a manager is a real choice with a real alternative, not the thing that happens to you next because you happened to be good at the job before it. Treat "I tried it and it wasn't for me" as data arriving on schedule, not a failure that needs correcting.

A lot of this lesson started in a leadership guide a former colleague wrote at CircleCI.

Everything above is just that idea run through a few hundred more conversations since. Ask first. Know when to stop asking. And make the growth conversation as ordinary as the status update, so it never has to become a special occasion to actually happen.