Duvo Manager Playbook

This is the operating manual for people who manage someone at Duvo. You still do the job. Management sits on top of craft. We hire senior people who manage themselves, we stay a small remote-first company, and we won't invent a manager class to paper over a span problem, an underperformance problem, or a person who shouldn't manage. Read this once. Use it when the work gets ambiguous.

Contents

1. Who this is for
2. Fundamental resources
3. The manager standard
4. Where your authority ends
5. Hiring
6. Onboarding
7. How we judge performance
8. Performance reviews
9. Giving feedback
10. Underperformance and exits
11. Difficult conversations
12. 1:1s
13. Coaching
14. Recognition
15. Promotion
16. Multiplier managers
17. Conclusion

Who this is for

Duvo builds AI agents that run enterprise operations: supply chain, finance, logistics. The work is live customer work. Agents in messy systems. Last mile on a real account. A feature that doesn't move the customer's company is not a win.

We're about fifty to sixty people, remote-first, across Czechia, the rest of the EU, and the US. Product, engineering, GTM, and people who sit with operators share one company. That size is a choice. It only works if the people we hire can run their own work.

We have no dedicated management track. Management is layered on craft. A senior engineer still ships. A senior operator still owns the customer outcome. A GTM lead still owns the relationship. Hitting a senior title doesn't make you a manager. Doing management well doesn't, by itself, get you promoted.

Most managers here oversee a small number of people and keep doing the job. Your primary work is still the work. If management is eating the craft, that's a people-ops problem. The span is too wide, someone is underperforming and you're carrying them, or you shouldn't be managing. Tell people ops. Don't quietly become a full-time manager and call it growth.

You're responsible for the performance, growth, and communication of the people you manage. That means:

• Setting a clear bar for the work and for how we show up.
• Being available for judgment, not for status theater.
• Writing honest reviews in Humaans when the cycle fires.
• Catching slips early and escalating them before they become a story.
• Modeling the values yourself. People copy what you do, not what you post.

This still matters, even though we don't reward it as a career path. You're shaping a colleague's week, their skill, and whether they stay. Managing well is craft. Sloppy management shows up as quiet people, late surprises, and work that needed you in it to ship.

Fundamental resources

You don't need a second handbook. You need the tools we actually use, and this page. If something here is wrong, say so. We would rather fix the page than let folklore win.

Humaans. HRIS. Profiles, documents, PTO, and the live performance cycles. Reviews live here. So do time-off records. If it's a people record, it belongs in Humaans, not in a private doc you will lose.

Ashby. ATS. Jobs, candidates, interview plans, scorecards, and work samples. Sample job ads live in Ashby. Use them as a starting point, not a paste. Every role we open should describe the work we actually need this year.

Slack. How we talk day to day. Work in public. Write a proposal with a point of view, not a discussion thread that goes nowhere. Private channels are for personnel and for things that would harm someone if they were public. Everything else can stand the light.

The values, plus the job bar. We don't score people on a secret rubric. We score them on how they live the values and whether they get the job done. Values without output isn't excellence. Output without the values is a short-term win that turns into a people-ops conversation.

The five values:

1. Own it. Don't wait for problems to land. Make them yours. Ambiguity is a reason to move, not a reason to pause.
2. Obsess over the customer. Know the customer's actual day better than the process doc. Say we shouldn't build this when it wouldn't help. Connect what you shipped to a specific person's work getting easier.
3. Grow and compound. Seek feedback and use it. Don't wait for a cycle. When you're wrong, say so and update. When you solve something, leave it where someone you have never met can find it. One-off work is a failure mode.
4. Prove, don't argue. Kill your own idea when the evidence says so. Measure before you propose. Follow up after you ship. Make feedback specific enough that the other person knows what to change.
5. Say it straight, now. Share bad news, blockers, and disagreements when you have them. Speak up about problems outside your area. It's still your company. Better wrong out loud than right too late.

The review cycles in Humaans. There are three. There is no separate annual review sitting on top of them.

• First 90 days Check-In (manager only)
• 6-month Check In (manager and self)
• Talent Review at Duvo (bi-annual, manager and self, for people past the first ~90 days)

Between those moments, feedback is continuous. A cycle is a snapshot. It's not the relationship.

Org structure lives in Humaans. Job ads and interview kits live in Ashby. Review forms live in Humaans. Look there for samples. We're not attaching PDFs to this page.

The manager standard

We judge managers the same way we judge everyone else: values, plus getting the job done. The job, for you, is that the people you manage do excellent work and know where they stand.

Craft

You use your expertise to guide approach, execution, and taste. You can tell a solid path that's not yours from a path that will miss. Stay close enough to the work to have an opinion. Don't manage from a dashboard.

At Duvo that means the customer work, not a status tool. Can you see the process, the exception, the last mile? Did the agent actually finish, or did a person quietly finish it? Would the operator's day get easier? Cool work that doesn't move the account is not craft here.

You keep a real picture of how they're doing. Not a vibe. The work, the customer outcome, how they write, how they show up, how they take a note and change. If they sell, you have seen the motion. If they sit with operators, you have seen the account. If they build, you have seen it run in a messy system, not only in a demo.

You write things down. Light notes on 1:1s. Specific examples when you give feedback. Honest reviews when Humaans asks. If the only record of a problem is in your head, you don't have a record.

You make the bar visible. They should know what "exceeds" looks like in this role, on this team, this quarter. When they're off it, you say so early.

Values

You run 1:1s that aren't status. You recognize work that matches the values, in public when it helps and in private when it should be private. You give feedback that's specific and that you follow. You don't wait for Talent Review to mention a problem you have been watching for months.

You make it safe to raise a concern, ask a dumb question, and admit a miss. Safety isn't softness. It's the condition for "say it straight." If people only tell you good news, you built that.

You work in public yourself. You write the proposal. You own the miss. You don't ask them for a standard you don't meet.

Coachability

You ask other managers and people ops when you're stuck. You offer advice that's grounded in work you have actually done, not in a thread you liked.

You ask the person you manage how to support them better. When they tell you, you try it. When the note is corrective, you get curious instead of defending your process.

A manager who can't take feedback won't grow a team that can.

Where your authority ends

Some of this work sits next to management. You're in it. You're not alone in it. Knowing the line keeps you from freelancing policy across three legal systems, and from promising things you can't close.

Time off

PTO at Duvo is inform, not request. They log it in Humaans and tell you and the team. You treat them like an adult. You cover the work. You don't interrogate the holiday. You don't ask what they plan to do with the days.

You should almost never block time off. The exception is a true critical event where the work can't move without them and there is no cover. Even then, you're rearranging, not granting permission. If coverage is a chronic problem, the team is too thin or the work is too concentrated. That's a hiring or scope problem, not a PTO problem.

Parental leave, sick leave, long leave

We operate in Czechia, the EU, and the US. Leave rules aren't the same, and they're not something you should interpret in Slack. When someone tells you they're having a child, they're going to be out sick in a way that's more than a few days, or they need a longer leave, bring people ops in. Don't advise them on eligibility, stacking policies, or what they "should" take. Cover the work. Stay human. Let people ops run the policy.

If frequent short absences start to break the team, escalate that too. Name the operational impact. Don't diagnose the person.

Complaints, harassment, discrimination

If someone brings you a serious conflict with a colleague, stop and bring people ops in. Same day. They will help you investigate, keep the circle small, and decide what happens next. You don't run your own inquiry in DMs.

If you hear harassment, discrimination, retaliation, or a hostile environment, or you hear that a complaint might go to a regulator, stop the conversation and say you need people ops in the room before you go further. Then do that. Don't promise outcomes. Don't tell them to "keep it between us." Don't wait until you have more of the story.

These are some of the most serious hours you will have as a manager. Speed and people ops, not instinct.

Pay and equity

Duvo managers have input on variable pay and equity. You can make a case at the 6-month Check In and at Talent Review. You know the work. Your recommendation matters.

You don't set base pay unilaterally. You don't promise a number, a percentage, or a grant in a 1:1. People ops and leadership close offers and close cycles. If they ask what they will make next year, the honest answer is that you will make a recommendation and you don't control the close.

If you have already said a number, tell people ops immediately. A promise you can't keep is worse than a quiet recommendation.

Don't invent a formula or quote someone else's package. Comp conversations leak. Be precise: the work, the scope change if there is one, your recommendation. People ops follows up.

Hiring

Hire when the work is stuck. Not to "build a team." Not because a headcount appeared. Not because a strong person is on the market and you're afraid to lose the slot. If it's not a backfill, you should be able to point to a sustained lack of momentum that a senior person, doing the job we already have, can't absorb.

People ops and leadership approve the open role. People ops runs the process. You own the bar and the work sample.

What you own

The definition of excellent for this seat. The must-haves versus the nice-to-haves. The work sample or job trial: a piece of work that looks like the job, at the seniority you're hiring, with enough room to see how they think. You will interview the work. You will decide if anyone clears the bar.

The trial has to be this job, not a studio puzzle. Forward-deployed and delivery: a messy live process, evidence, a last mile. GTM: a real motion, a written point of view, judgment under a customer constraint. Product and engineering: something that has to run in a messy system, not a clean repo. Same bar across seats: can they get the job done, and do they show the values while they do it.

Design the trial so a strong person can show craft, judgment, and communication. Pressure-test the brief. If the prompt is vague, you will grade on mind-reading. If the prompt is a puzzle, you will hire puzzle-solvers. If the prompt is the job, you will hire people who can do the job.

What people ops owns

The job post in Ashby. Candidate communication. Scheduling. The process staying fair and complete. Closing rejected candidates like adults. The offer. You may have a point of view on the offer. You don't send it.

Walk people ops through the real bar before they screen: hard skills, ownership, the writing you want to see, the work you won't train from zero. A fuzzy bar produces a fuzzy pipeline, and then you lower it because the calendar is moving.

How it should feel

You may hire no one. That's a successful search. An empty seat is cheaper than a person who adds drag, needs you in every piece of work, or has to leave in the first year. If you're uncertain, you're not ready to offer. Ask another question, run another work sample, or stop.

Don't hire for potential you can't see in the work. "They will grow into it" is how you spend a year managing someone.

Offers come from people ops. Once someone accepts, you switch to onboarding. Keep the rejected people you would gladly reopen in Ashby. Don't keep a private list.

Volume isn't a reason to drop the bar. A loud pipeline can make "good enough" feel like a gift. It's not. If you wouldn't be glad to sit next to them in six months, keep looking.

Onboarding

We hire people who can operate as managers of one. After they're up to speed, they shouldn't need you on them. The first weeks are different. They deserve a clear picture of how we work, how we talk, and how you will judge them. Most new-hire pain here isn't the craft. It's the operating system.

Before day 1

People ops handles accounts, Humaans, and the first-week logistics. You handle the work. Know what they will own in week one. Know who they should meet. Know the first real problem, not a fake training project. At Duvo that is a live process, a live account, or a live motion. Write it down so you're not inventing it on the call.

Day 1

Get on a welcome call. The point is human, not a tour of tools.

• Welcome them. Start getting to know each other as colleagues.
• Walk the first days and weeks. What good looks like at the end of week one and week four.
• Point them at people worth meeting, on the team and off it.
• Set weekly 1:1s. Weekly to start. You can loosen later if the work and the relationship can take it.

The first weeks

Use the 1:1s. Don't save this for a doc they won't read.

How we work. The tools. How feedback actually happens (in the work, in the proposal, in the 1:1). How to get stuck out loud and ask for help before the deadline is a surprise. What "done" means on your team, including the last mile: verified, visible, handed off, and real on the customer side. We use AI on our own work. Walk the walk. A teammate who won't try the tools we sell is not ramping.

How we communicate. Slack in public. A written proposal beats a long thread. We are good at discussing and worse at proposing. Don't add to that. Put the work where the team already works. When to ping someone and when to write it where the next person can find it. We don't have a shadow process in DMs.

How you assess the work. Be plain about what you need to see immediately, and what you expect to see as they settle. Craft first. Then judgment: picking the few things that matter, shipping a thin version, not waiting for a perfect plan. Show them a piece of work you consider excellent and say why. If the seat is customer-facing, show them an account done well, not a deck.

How you assess the rest. Walk the values with examples from your own work, not from a poster. Own it, customer, compound, prove, straight. Tell them you will look at whether they get the job done, not whether they sounded busy.

The first-year bar. Mention hell yeah more than once. At 90 days you should already want them to stay. At 12 months you should be enthusiastic, not merely fine. Fine is a conversation. They should hear that early, while there is time to do something about it.

Give live feedback on the work in front of you. Talk through your own decisions so they can steal the judgment, not just the answer.

If they're lagging

Escalate to people ops as soon as you think there might be a problem. Don't wait for the 90-day form to make it real. The 90-day Check-In is a stay-or-concern moment, not the first time you noticed. Early is kinder. Late is how people get a surprise at probation.

How we judge performance

We use one model for everyone. Role and level change what "good" looks like. They don't change the axes.

Values. How they show up. Own it. Obsess over the customer. Grow and compound. Prove, don't argue. Say it straight, now. This is coachable in the small (a specific behavior) and much harder in the large (they don't want to be this kind of colleague). A repeated miss on values isn't a style difference.

Getting the job done. They have to produce. Customer outcomes, not just functional output. Focus and judgment. Dependable delivery through the last mile. Role craft at the standard the work requires. Making the people around them better, not busier. Values without this isn't excellence.

For a quick scan, you can still think HIGH / MID / LOW on three things:

1. Craft. Can they do the work?
2. Values. How do they show up?
3. Coachability. Do they take feedback and use it?

Coachability is the hinge. A person who can do the work and won't be coached won't stay excellent. A person who is still ramping and will be coached is a bet you can often take. A person who is low on values and low on coachability is a people-ops conversation even if the craft is high.

Scans are fast on purpose. People move around. Someone can be low-energy on one project and high on the next. That's human. Don't make a career decision on a single week. Do make a people-ops conversation when the scan keeps coming back ugly.

A pattern of high craft, low values, low coachability is the classic miss. They were hired for the work. They go dark. When you name it, they point at the output and treat the rest as noise. Name the scan so you don't get talked out of a real problem.

Use the model when you level someone, write a cycle, coach, or try to say why they're off. If you can't point at craft, values, or coachability, you have a feeling. Get more evidence or let it go.

Loop people ops when a low on values or coachability repeats, when craft is stuck below the job after a fair chance, or when you're about to write down a serious concern for the second time.

Performance reviews

You assess performance all year. The Humaans cycles are where you write it down. Skipping a cycle, or writing fog, leaves them with a false picture and the team with drag you have been ignoring.

There is no separate annual review. Use these three.

First 90 days Check-In

Manager only. It fires around the end of probation. You have 14 days. This is a stay-or-concern check, not a novel.

Score how they live the values and whether they're getting the job done. Give one or two concrete examples. Name a single next action (continue growing, give focused support, or address a performance concern). Rate overall Below expectations / Meets expectations / Exceeds expectations, the same language as Talent Review.

Hell yeah starts here. If you're not glad they're here, escalate now. Don't wait a year to discover you were being polite. Thin evidence is fine. You will have enough to know if the bet is working.

6-month Check In

Manager and self, around month five. You each score the five values as Demonstrating or Developing, with a short reason. Fill in your side before you read theirs. The useful part is often the gap.

Developing on a value at six months is information. It's not a verdict. A wall of Developing with no movement, or a value they insist they demonstrate and you can't see, is a conversation.

There is a manager-only recommendation on the form: no change, comp, role, or both. That field is for people ops, not for the 1:1. It's not a promise. Don't preview a number. People ops and leadership close it. If scope has grown and they're demonstrating the values, this is the moment to make the case. If nothing has changed, say no. A ritual yes trains everyone to expect a raise for surviving the calendar.

Talent Review at Duvo

Bi-annual. Manager and self. People past the first ~90 days. Same Below / Meets / Exceeds expectations language.

You score each value. You also score getting the job done: customer outcomes, focus and judgment, dependable delivery, role craft, and whether they raise the people around them. Then one next action (continue growing in the role, give greater scope or a stretch assignment, consider for another role, provide focused support, or address a performance concern), an owner, and a date. Evidence first. One or two examples that explain the scores. Not a paragraph of adjectives.

At twelve months you should be enthusiastic. Not "fine." Fine is a conversation. Hell yeah means you want them on the team as they are, and you want to see what they do next. If you can't say that, don't paper it with Meets.

Evidence, not vibes

Support what you write. PRs. Customer outcomes. Shipped work. A proposal that changed the plan. A mess they cleaned up that wasn't their job until they owned it. If you can't link it or describe it so a third person would nod, you're grading a feeling.

When you prepare, look at the work. How much of the final thing is them, and how much is you or a teammate holding the pen? How much correction did it take? Did they pull their weight?

Peer feedback

Ask two or three specific questions of the people who actually work with them. Not "any thoughts?" Recency, one change, more or less next time:

• In the last six weeks, when did they make your work better? One example.
• What is one thing they could do differently that would help you?
• If you worked with them again next month, what should they do more of, or less of?

Ask for candor. Never paste a negative comment into the review or the 1:1. Summarize. Protect the speaker.

Goals

Light. Few. Tied to the work in front of them. Two goals they can actually hit, maybe one stretch that would be impressive and isn't load-bearing. Write them down. Follow them in 1:1s. Don't build a goal program.

Giving feedback

Reviews are snapshots. The job is the weeks in between.

Everyone here is expected to take feedback and use it. That's coachability. Your side of that deal is to give feedback that's useful: observed, specific, aimed at a change.

Three words, because they get mixed up:

• Positive names what worked so it happens again.
• Negative ranks the person. Don't.
• Constructive uses what you saw and what you know to steer the work. Critical, and helpful.

You're aiming for constructive plus positive, all year, in both directions.

Skip the compliment sandwich. It trains people to wait for the "but." Say the thing. If there is also something working, say that as its own sentence, some other minute.

Give it when you see it. Waiting for Humaans is how a small miss becomes a personality.

Approach

People here get a wide lane. If their plan is solid and it's not yours, leave it. That's not management. That's taste enforcement.

Step in when the approach is burning time, quality, the customer, or the team. Be concrete.

Tactics

Most feedback is on the daily work. A PR, a customer note, a runbook, a proposal.

Be short. Name what you see and what should change. Don't dump a dozen nits and call it a standard. Offer the fix, or the example, not just the frown.

Engagement

Sometimes the work is fine and the way they're showing up is not. Blunt review comments. Silence in the places we work in public. A new sharpness with a teammate.

Talk about the behavior, with examples. Not the soul.

Be a colleague, not a therapist. If something personal is leaking into the work, listen, then bring it back to the job. If you're out of your depth, people ops.

And say the good thing when it's true. Specific, tied to a value or an outcome. People can't repeat what you never named.

Underperformance and exits

When someone is doing well, management is light. You stay available. You keep the bar visible. You get out of the way.

When you start to doubt they're meeting the job, you need more precision and less hope. A short window to show they can recover is fair. Half a year of "maybe the next project" is not. You're not being kind by waiting. You're spending the team's patience and their reputation.

Tell them you're looking, then tell them what you saw. If the last month of work needed you to finish it, say that. If it was excellent, say that too. Silent evaluation is how people get ambushed.

When the work is good, you're mostly a peer. When it's not, you take the authority. Uncomfortable in a company this small. Do it anyway, with evidence, and stand behind the call.

Recovery

Most underperformance shouldn't end in an exit. Caught early, named directly, coached in the work, it often turns.

Be explicit. What does better look like. By when. How you will look. Check in often enough that they can feel you're invested, not only present for bad news. When you see the change, name it. "I saw you own the last-mile on [X], and it changed the outcome" is part of the work.

When they're back, say so. Tell them the extra monitoring is done. Tell people ops. If you still feel a knot, use the model, not the knot. Craft, values, coachability, job done. People ops can be a second set of eyes. Trust takes a while to rebuild. That's normal. Don't confuse leftover worry with a new case.

When to escalate

Document early. One clear conversation, with examples, is management. A second write-up of the same problem is a people-ops conversation, and you have it before you deliver that second write-up.

Severe underperformance is falling below the job, or a single miss that actually broke something important. If you don't know which side of the line you're on, ask people ops. Default to asking.

Bring people ops in before you tell someone their job is in danger. You align on the path, the timeline, and what "recovered" means. Then you talk to them. You don't freelance a performance plan in a 1:1 and loop people ops later.

If the case is eating your craft, say so. People ops can take more of the process. The load is supposed to be sheddable. That's the point of not having a manager class.

Exits

Termination is never a surprise and never a manager solo act. If it's a surprise, you failed the months before it. If you do it alone, you have stepped outside the line. People ops partners on the formal process. Leadership is in the decision. You're in the truth-telling.

We're in more than one country. Follow people ops. Don't apply a speech you read on the internet. Don't talk about "at-will," notice formulas, or what you think they're owed. Humane and clear. The logistics belong to people ops.

Broad shape, not a script:

1. You see a serious lapse. Pattern or one-time. One-time gets more room. Pattern gets less.
2. You go to people ops. Together you set a path with a timeline and visible marks of success. The timeline can shorten if nothing moves.
3. You tell them, straight, that the work is below the job, what has to change, and that the job is at risk. You offer help. You don't pretend this is a casual coaching chat.
4. You work the plan. People ops stays close. If you lose hope of recovery, you say so then, not after another month of theater.
5. You make a retain or exit call. If you can't, people ops and leadership won't let the case float.
6. If it's an exit, people ops runs the process. You're usually the one who says it, because you have been the one saying it. If you truly can't, people ops can. In that case, you should expect to stop managing.

After you have said it, you're done. People ops handles last day, access, and paper. You handle the team and the work.

Difficult conversations

The performance chapters tell you when. This is how.

Don't delay. Hoping it blows over is how the rest of the team starts carrying the cost, and how the person loses the chance to fix it.

Tell people ops (and your own manager, if you have one) before a high-stakes conversation. They will help you set a fair path.

Prepare

Write talking points, not a script.

• Where they're off, in the language of the values and the job.
• Two or three examples from the work.
• What it's costing the team or the customer.
• What back on track looks like, and by when.
• What happens if it doesn't change, including that the job may end.

If you sound rehearsed, you will miss the actual conversation.

Imagine how it lands. Then manage your own temperature. Impatience and the urge to soften the point until it disappears are normal. The facts still have to be the facts.

In the room

Open with intent. You want them to succeed, and you need the work to change. Ask questions. Don't deliver a monologue and call it feedback.

Hit the points. Then stop. Repeating the same charge in new words doesn't make it truer. It just grinds.

You come with the non-negotiables (what must change, the marks, the date). You can build the steps with them. They should have a say in how they get there. They don't get a vote on whether the bar exists.

After

Write down what you agreed, including dates. Share it with them. Share it with people ops if this is already a formal path. Then look when you said you would look. If you named the end of the month, you don't discover flexibility on the 29th.

Check the work and the person:

• How is [the thing] going?
• What would help?
• Any questions now that it has been a week?
• How are you doing?

When the date arrives, decide. If they recovered, say why, and that they need to stay there. If they did not, say you're moving to an exit and why. Don't open a new, unofficial timeline because the conversation is uncomfortable.

Saying it's over

Same preparation. Shorter. This isn't a debate.

You have decided to end employment. Last day is X. The reason is the performance you have already discussed. People ops will follow up on the rest. Is there anything they need to say.

Then stop talking. Silence isn't your job to fill. Don't narrate your guilt. They're having the worse day. If it turns into argument, negotiation, or heat, point them to people ops and end the call.

1:1s

1:1s are how you keep the relationship honest. They're not a status meeting. Status can live in the work and in Slack.

They get regular feedback, so a miss can't grow in the dark, and a place that's not the public channel to say they're stuck. You get the same in the other direction, plus time to practice judgment while the stakes are still small.

"No news is good news" works for a few people. It fails most. If they only hear from you when something is wrong, you trained that.

Cadence

Weekly for the first stretch, 30 to 60 minutes. After they're up to speed and the work is clean, move to whatever you both need. Some people want to keep weekly. Some are fine less often. Use your judgment. If you're drifting past a month with nothing, you're not managing, you're hoping.

Either of you can put topics in. A shared note is enough. No project around the meeting.

What to talk about

Prepare a few bullets. Invite theirs. Then go into a real piece of work, a real decision, or a real snag.

Useful questions:

• How did the last stretch go?
• What is working? What is in the way?
• Walk me through a choice you made this week.

That last one is the job. We hired them to decide. Judgment is a muscle. Review the decision, not just the artifact. Sometimes walk one of yours so they can see the steps, including the ones you skipped.

Now and then, ask about growth. What they want more of. What is giving them energy. What they want to get better at, and what you can put in their way that's real work, not a side quest.

Notes

Light. If you will need it later (a goal, a miss, a commitment), write it and share it. If it was just a good conversation, you don't need a transcript. Put anything that matters for performance where you will find it when Humaans fires, not in a chat you will never search.

Coaching

When they need you, think like a coach who still plays. You know the work, and you know that different people, and different weeks, need a different amount of you.

Coaching is how you teach people to decide. If every hard call comes back to you, you didn't build a teammate. You built a queue.

Directive

You tell them what to do. Use it when they don't have the skill yet, when time is actually short, or when there is one correct move and a Socratic circle would be theater.

It's fast. It's also how you become the bottleneck. It doesn't grow them. It assumes you're right. Sometimes you are. Don't make it the default.

Non-directive

You ask, you listen, you don't arrive with the answer. Open questions. What have you tried. What would you do if I were out next week. What does the customer actually need. You reflect back what you heard until they can hear their own plan.

This is slower in the moment and faster by month three. It respects that they may see something you do not. Your job is to open the path, not to walk it for them.

You don't need to be the expert. You need to shut up long enough for them to think.

Situational

You switch. New person, production fire, one right answer: directive. Capable person, messy product problem, time to think: non-directive. The skill is noticing which room you're in.

Nine times out of ten you will know a decent answer. Saying it every time trains them to bring you puzzles instead of decisions. Coaching is how a company of our size stays fast without you in the middle.

If non-directive feels fake, practice on a peer, or in a 1:1 when the outcome can tolerate exploration.

Recognition

People want to be paid fairly. They also want the work to matter, and to have that noticed by someone who knows the difference.

Recognize the thing that matches the values or the outcome. Not "you're nice." Nice is pleasant. Owning the last mile on a customer mess is the job. Name that.

A simple shape: thank you. The specific thing. The effect.

"Thanks. You killed the approach when the evidence did not hold, and we did not waste the next two weeks. That is prove, don't argue."

Public when the team should see the standard. Slack is enough. A note in the channel where the work lives beats a kudos program nobody believes. Private when the person would rather have it private, or when the content isn't for the room. A short DM that names the thing isn't a lesser form. Sometimes it's the only honest one.

Don't invent a ritual so we can say we recognize people. A few minutes on real work is fine. A quota is not.

Ignore what you don't want more of. Reward what you do. That's culture.

Promotion

Two gates, both required.

They're already doing the next job. Not a promising week. A sustained stretch, usually months, where they exceed the current role and meet most of the next one. You can point at work. You would defend the call. You're not stretching the language until it fits because you like them. Liking them is good. It's not a level.

The business needs that seat. Another title has to change how the work runs: ownership, scope, a hole we actually have. Promotion isn't a thank-you. It's not a retention sticker. It's not a prize for being a good manager. If the work doesn't need the next level, find another way to recognize them (scope, compensation recommendation, public work) and be honest about why the title is staying put.

Ask yourself:

• Will this title improve how the team operates, or are we over-leveling relative to the work?
• Why is promotion the right move, rather than a different assignment or no change?
• What happens if we do not? If the answer is "they might be disappointed," that's not enough.

If you can't see better outcomes, clearer ownership, or more capacity, pause.

Talk with them about the path. Be generous and specific about what you would need to see. Never promise a date. Never promise the outcome. Promises you don't control are how trust dies.

When you're ready, write a case. Work samples. A narrative on craft, values, coachability, and the job done. The business reason. Send it to people ops. People ops and leadership decide. You recommend. You don't confer.

Multiplier managers

We don't have managers so the org chart looks grown-up. We have them because even strong people can't always see themselves, and because a team needs a person who will name the standard and keep it.

Done well, this compounds. Done badly, quality fades slowly, then all at once.

You keep good people. People leave work. They also leave managers who won't say the thing, won't grow them, and won't deal with the person making the week worse. Retention isn't a mood. It's operational stability. Replacing a senior person here is slow and expensive, and the customer doesn't pause while you do it.

You protect the work. Excellent has to be sayable. If you won't describe it, it erodes. Ownership blurs. Apathy gets into the craft. One person below the bar pulls the team toward them unless you hold the line. Quality control is a team sport, and you're the one who can't look away.

You push decisions to the edge. If every call comes up, you're the constraint. Teach them to decide. Then you both move.

You reduce risk. The quiet kind. Document concerns when they're small. Escalate to people ops before you write the same problem down a second time. That protects the person (they get a real chance, not a file built in secret) and it protects the company. Avoiding the conversation isn't kindness. It's how a messy exit gets messier, in any country we operate in.

You're the culture. Values on a page do nothing. What you reward becomes the target. What you ignore becomes allowed. Stay decent when you have to pass along a decision you dislike. Broadcasting your dissent to the people you manage doesn't make you authentic. It makes the team smaller.

Conclusion

You won't get this from vibe or from title. You get it from repeating a few unglamorous things: clear expectations, written enough to be fair, said soon enough to be useful.

Come back to this page when the case is messy. Change it when it's wrong. When you're unsure: are they clear on the bar, are you applying the same process you would apply to anyone else, and are you saying the true thing, kindly, now.

Clarity. Consistency. Courage. That's the work.

Inspired by the 37signals Manager Playbook (MIT License, Copyright 2026 37signals).