Back to Blog

Your First Skip-Level Meeting: How to Run One Without Undermining Your Managers

EngineeringLeadership

Last updated: October 14, 2026

By Kris Drouet, Engineering Executive, in partnership with KORE1

A skip-level meeting is a one-on-one or small-group conversation between a leader and their managers’ reports, without the manager present, and it works only when the leader listens for patterns and routes every decision back to its owner. Run it as a listening session, never a decision venue. The quickest way to undermine a new manager is to say yes to something in that room that was theirs to decide.

You have managers now. Maybe three of them, maybe seven, and somewhere below each one sits a team you used to talk to every day and now mostly hear about secondhand. Somebody, probably HR or your own boss, has told you to start doing skip-levels. Nobody has told you how to do them without knocking the legs out from under the managers you just spent a year building up. That part matters more.

I got my first one wrong in about four minutes.

I had just moved from running a single team to running managers on a mortgage-tech platform, and I booked thirty minutes with each engineer one level down. The second conversation was with a senior engineer who had a sensible complaint. Our Jenkins pipeline was flaky, builds were failing for reasons nobody could reproduce, and she wanted the team to move to GitHub Actions. I had wanted that for a year myself. So I said it sounded reasonable and she should go for it. Felt great. Felt like leadership.

Her manager had decided two weeks earlier to keep Jenkins until our SOC 2 audit window closed, because changing the build system mid-audit meant re-evidencing every change-management control we had. He had good reasons. He had explained them to the team. And now the team had learned, from me, that there were two places to get a yes, and one of them outranked the other. He found out the next morning, from her, in standup. That was the real cost. Not the pipeline. The fact that every engineer on his team now had a reason to go around him.

It took me most of a quarter to repair. This piece is the version of skip-levels I wish I had run instead, built on the same idea I use for moving your best IC into a manager seat, which is that every decision has an owner and your job is to keep it with that owner, especially when you could make it faster yourself.

Silver-haired engineering director in a charcoal blazer listening with her hands on a closed notebook while a software engineer in a gray hoodie explains something across a small table, burnt orange wall behind them

What a Skip-Level Meeting Is For

A skip-level meeting is a scheduled conversation between a senior leader and employees two or more levels below them, held without the employees’ direct manager in the room, so the leader hears how work actually feels on the ground and the employees get direct access to context from above.

The definition is the easy part. The reason it exists is more interesting, and it changes how you should run one.

Information gets filtered on the way up. Always has. Everybody knows this in the abstract. The research is less abstract. Frances Milliken, Elizabeth Morrison, and Patricia Hewlin interviewed 40 employees for a 2003 study in the Journal of Management Studies and found that most of them had been in situations where they were worried about an issue and did not raise it with a supervisor. The most common reason was fear of being labeled negatively and damaging a relationship they valued. Not fear of being fired. Fear of being seen as the complainer.

Then there is the finding that should make every new director sit up a little. James Detert and Linda Treviño ran 89 interviews inside a high-tech multinational for a 2010 paper in Organization Science titled, plainly enough, “Speaking Up to Higher-Ups: How Supervisors and Skip-Level Leaders Influence Employee Voice.” Employees named skip-level leaders, those two to five levels above them, as reasons to see speaking up as risky or futile nearly as often as they named their own boss.

Take that one seriously. You are already shaping whether people talk, whether or not you ever hold a skip-level. The meeting just decides which direction.

A skip-level does two jobs. It lets you hear what the filter strips out, and it shows people, in person, that speaking up two levels above their manager is safe and not futile. It does not do a third job, and this is where most first-timers go wrong. It is not a place where decisions get made.

Three Ways a Skip-Level Undermines Your Managers

None of these come from bad intent. All three come from a senior leader being helpful in the wrong room.

The on-the-spot yes is the one I did. Most new directors do. Somebody brings you a reasonable request that sits inside their manager’s authority, you agree because it is reasonable, and you have just overruled a decision you did not know existed. Even when you happen to agree with the manager, saying “I’ll look into it” in a voice that sounds like a promise does the same damage. The engineer leaves with a commitment from the wrong person.

The back channel is slower and worse. Much worse. After two or three rounds, people notice that problems raised with you get fixed faster than problems raised with their manager. Rationally, they start bringing things to you. Your calendar fills with what used to be your managers’ work, your managers lose the information they need to manage, and you have built yourself the exact approval queue that turns a leader into the bottleneck, one level higher than last time. It shows up in Slack first. DMs to you that start with “quick question” and should have gone to someone else.

Then there is the performance review in disguise. “How’s your manager doing?” sounds like care. To the engineer it sounds like an invitation to evaluate their boss, on the record, to their boss’s boss, with no idea where the answer goes. Most people deflect. A few take the invitation, and now you are holding an unverified opinion about a manager who was not in the room to answer it. GitLab’s public handbook, which is one of the more thorough written policies on this, is blunt about it. Its skip-level guidance lists checking in on a manager’s performance “by going around them” as something skip-levels are not for. It also rules out using them to manage priorities or careers, which are the manager’s job.

All three failures have the same shape under the hood. A decision that belonged to the manager quietly moved up a level. Nobody announced it. It just happened in a thirty-minute meeting with a nice cup of coffee.

Before the First One, Brief Your Managers

Do this a week ahead. In writing, then out loud in your staff meeting, so nobody hears about it from their own team first. Your managers, who will hear about these meetings within an hour whether or not you tell them, are the audience new directors forget, and they spend the whole round deciding whether you are building them up or quietly building a case against them.

What you are giving your managers is a contract. It tells them exactly what you will do with what you hear, so the skip-levels stop being a threat and start being a tool they get something out of. Mine has four terms.

  • I will not make decisions in these meetings. If something comes up that is yours to decide, it comes back to you, and I tell the person that in the room.
  • You hear themes from me, not quotes with names attached, within two business days. One exception, which I cover below.
  • Anything that is genuinely mine to fix, like a cross-team blocker or a budget question, I fix, and I tell you before I tell your team.
  • I am not evaluating you through your people. If I have a concern about how you are managing, you will hear it from me directly, in our one-on-one, with specifics.

Then ask them what they want you to listen for. This part matters more than it looks. A manager who has been trying to get a sense of whether the team understands the reorg will happily hand you that question. Now the skip-level is partly theirs. They stop bracing for it.

Expect at least one manager to be nervous anyway, usually the newest one. That is fair. If they are four months into the role, they are in the stretch where a promoted IC is most likely to look like they are struggling when they are actually on track, and a skip-level that goes badly can tip a fragile team into routing around them for good. Slow down for that team. Do their skip-levels last, after you have practiced on everyone else.

The Routing Table: Who Owns What You Hear

Nobody gave me this before my first round. It applies the same logic as my decision-ownership map, which sorts calls by how reversible they are and how far they reach, to the specific things people say in a skip-level. Almost everything you hear falls into one of six rows. In most of them, your whole job in the room is to say out loud whose call it is.

What you hearWhose decision it isWhat you say in the roomWhat you do afterward
A call inside the team, like tooling, process, standup format, or who reviews whatThe manager“Tell me why it matters to you.” Ask, then make no promise.Pass the theme to the manager. Their call stands whichever way it goes.
A blocker that crosses teams, like a platform dependency stuck for three weeksYou“That one is mine. I’ll have an answer by Friday.”Fix it. Tell the manager first, so the team hears about it through them.
Career, promotion, compensation, “am I on track for senior?”The manager“That belongs with your manager, and I’ll make sure it happens.”Flag to the manager that the question is live. Check a month later that it was answered.
Strategy and context, like where the AI roadmap is going or why the reorg happenedYouAnswer it fully. This is the best use of your thirty minutes.Tell your managers what you said, so the story matches next week.
The same concern about the manager from three people, on separate datesYou, as the manager’s coachThank them. Ask what happened, not who. No commentary on the manager.A coaching conversation with the manager within a week. Never with the team.
Harassment, discrimination, a safety issue, a fraud or regulatory concernYou and HR, nowSay plainly that you have to act on it, and what happens next.Escalate the same day. This is the one row where nothing waits for the manager.

Look at the third column. Mostly questions. Ownership said out loud. The rows where you commit to something on the spot are the ones where the problem already sits above the manager’s authority, or cannot wait for anybody.

The hardest row is the first one, because it is where your expertise itches. If you came up through the code, and most directors in engineering did, you will hear a tooling complaint and know the answer before the engineer finishes the sentence. Keep it to yourself. Trust me. On that mortgage-tech platform, the first-row complaints were almost all tooling, Jenkins again, a code-review rule that required two approvals on a five-line config change, a standup that had drifted to forty minutes, and every one of them was something the manager already knew about and had a reason for. If the manager’s call is actually wrong, that is a conversation for your one-on-one with the manager, where they can push back with context you might not have. I did not have the SOC 2 context. I would have found out in about ninety seconds if I had asked him instead of answering her.

Row five needs a number, because the instinct is to act on one strong comment. Don’t. One person with a complaint about their manager is an anecdote, and sometimes it is a performance conversation that person is trying to get ahead of. Three people, on different dates, describing the same behavior in their own words, is a pattern. Show me the data before you go into a manager’s one-on-one with a concern built from a single conversation.

Hand sliding a folded blank note into one slot of an old wooden mail-sorting cabinet holding cream notes and burnt orange folders

Running the Skip-Level Meeting Itself

Thirty minutes. Book it yourself, from your calendar, with an agenda in the invite, so it does not look like something HR made you do. Make the first one closer to a coffee chat than a meeting. Keep it light. GitLab’s handbook suggests exactly that for a first one-on-one skip, then a clear agenda after.

Format depends on how many skips you have. GitLab requires skip-levels at least quarterly for any leader whose reports have reports, allows groups of no more than 10 to 15, and never records them, because the point is trust. I run smaller groups than that. Six is about where I stop hearing from the quiet engineers. Past about thirty skips, one-on-ones stop fitting in a quarter, so I rotate through groups, a new set each quarter, which GitLab also suggests for large organizations.

Open by saying what the meeting is and what it is not, in roughly the language you used with your managers. Something like “I’m here to listen and to answer questions about where we’re going. I won’t be making decisions about your team in here. That’s still your manager’s call.” It costs fifteen seconds. It also takes most of the pressure off, because now nobody has to wonder whether a complaint becomes an action item.

Then ask questions that are about the work, not about the boss. These five consistently get real answers for me.

  • What slowed you down the most in the last two weeks?
  • If you could change one thing about how this team decides things, what would it be? (That one tells me a lot about decision ownership without anyone having to name a person.)
  • Is there anything you’ve stopped bringing up because nothing happens when you do?
  • What do you not understand about where the company is going?
  • Who on another team made your life easier lately?

The third question is the one that earns the meeting. It goes straight at the futility perception Detert and Treviño described, and the answer often points at you rather than the manager. Good. That is a row-two or row-four problem, and it is yours to fix.

What not to ask. “How’s your manager doing?” covered above. “Is everyone happy?” gets you a yes. Anything that requires them to rank teammates. And don’t spend the whole half hour on your strategy deck. Elizabeth Morrison’s 2014 review of employee voice and silence in the Annual Review of Organizational Psychology and Organizational Behavior defines voice as communicating suggestions, concerns, or information about problems to someone higher up. Upward is the point. If you talk for twenty-five minutes, you’ve held a town hall with fewer chairs.

Take notes on paper. A laptop screen between you and an engineer who is deciding whether to tell you something reads like a transcript is being made.

The 48 Hours After

This is where skip-levels earn their keep or lose it, and it is where almost nobody spends any effort. I didn’t, at first.

Within two business days, every manager gets a short note from you. The note does not need to be long (mine are usually five or six bullets in a Slack DM), but it does need to land before that manager’s next one-on-one with anyone on the team, because otherwise they walk into it blind. Themes, not names. What you heard that is theirs, flagged as theirs. What you heard that is yours, and what you are doing about it. And anything that surprised you, framed as a question. “Two people mentioned they’re not sure who approves schema changes on the shared topics. Is that settled and I missed it?” A manager who gets that note feels briefed. A manager who gets nothing spends the week wondering what was said about them. Every single time.

Then close the loop with the people who talked to you. Not individually, and not in a way that reveals who said what. When you fix a row-two blocker, say so in your next all-hands or staff update and credit the team that raised it as a team. When a row-one question went back to the manager and the manager kept their decision, the manager explains it. Not you.

Track it. A simple sheet, one row per theme, columns for date, row in the routing table, owner, and status. Mine had 23 themes after my first full round across five teams on that mortgage-tech platform. Eleven belonged to managers. Seven were mine. Three turned out to be the same issue, a shared Kafka topic with no clear owner, heard from three teams who each thought it was somebody else’s problem. Two never resolved. Those numbers are from my own notes, not a study. The proportions have stayed about the same every time I’ve done this since.

The second round is the real test. Watch the room. If people bring up things that came from the first round and got fixed, you have built trust. If the room is quieter than last time, something you did in those 48 hours told them the meeting was theater.

When the Skip-Level Surfaces a Manager Problem

Sometimes row five fills up. Three people, separate conversations, same story. The manager is never available, or takes over every hard ticket, or has stopped running one-on-ones.

The temptation is to fix it from below. To start meeting with the team more often, to quietly reassign the hard tickets, to become the manager’s backup. Every one of those moves teaches the team that their manager is no longer really in charge, and once they learn that, they do not unlearn it.

Go to the manager. Bring the pattern without the names and ask what they are seeing. Most of the time a new manager knows something is off and has not found words for it, and that becomes the coaching conversation that catches a wobble before it becomes a wash-out. If the manager is newly promoted and you need structure around it, KORE1’s engineering manager coaching and advisory engagement was built for that stretch. And if the issue is less about one manager and more about you and your peers learning to lead leaders, which is the stage this whole piece is about, engineering leadership transition coaching is where I would start.

Occasionally the answer is that the seat needs a different person. It happens. That is a hiring problem, and treating it as a coaching problem for a third quarter costs everyone. For that case, KORE1 runs engineering manager searches as direct hire placements, and KORE1’s placements hold a 92% one-year retention rate. A lot of that comes down to the brief, and the brief is better when a director already knows, from a full round of skip-levels, exactly what the team needs that it was not getting.

Director in a navy sweater and an engineering manager in a cream blouse holding a coffee cup walking down a bright office hallway in conversation, burnt orange wall behind them

What Directors Ask Me Before Their First Skip-Level

Quarterly, Monthly, or Whenever There Is Time?

Quarterly is the floor I set for every engineer two levels down, with the first full round finished within your first sixty days of managing managers.

Monthly starts to feel like supervision, and the managers feel it first, usually as a slightly too-casual question in their own one-on-one about whether everything is okay with the team, which is the sound of a manager wondering what you are hearing. Whenever-there-is-time means never, because skip-levels are always the meeting that gets moved.

Do I Warn My Managers First, or Does That Kill the Candor?

Brief them first, always, because a skip-level your managers learn about from their own team looks like an investigation no matter what you say in the room.

Candor doesn’t come from surprise. It comes from people seeing, over two or three rounds, that what they said led somewhere and did not get anyone in trouble. A briefed manager can even tell their team the meetings are a good thing, which no memo from you will ever match.

One-on-One or a Group?

One-on-one for the first round if you have fewer than about thirty skips, then small groups of four to six, rotating, once people know you.

Groups are efficient. People are braver with peers nearby, too. They also make it harder for the quietest person to say anything. I mix it. A one-on-one with each new hire inside their first quarter, groups for everyone else.

Somebody Used the Meeting to Vent About Their Manager. Now What?

Thank them, ask what happened rather than who did it, and change nothing until you hear the same thing from at least two other people independently.

That is not dismissive. It protects the person who talked, because acting on a single complaint points straight back at them. If it never shows up again, it might still be true, and you watch it. If it shows up twice more, it goes to the manager as a pattern, in your one-on-one, and the venting engineer is never named.

I Used to Sit Next to These People. Does That Make It Awkward?

It usually helps, as long as you stop being the person they bring technical decisions to, which is the habit former peers will test first.

They will. Someone will ask your opinion on an architecture choice their tech lead already made, half because they trust you and half to see what happens. Give the routing-table answer. Kindly. You will have to do it three or four times before it sticks, and it is worth every one of them. My view on how much engineering leaders should still code has not changed. Staying technical is how you ask good questions. It is not permission to answer the ones that belong to someone else.

What If Nobody Says Anything Real?

A quiet first round is normal and is itself useful information, since it tells you how risky speaking upward has felt on that team.

Keep showing up. Close the loop visibly on the one small thing somebody did mention. By the second round, someone tests you with something slightly bigger. That is the meeting starting to work.

Your Skip-Levels Belong to Your Managers Too

The best compliment I ever got on a round of skip-levels came from a manager, not an engineer. He told me his team brought him more problems after I started doing them, not fewer, because they had figured out that things raised in either room ended up with the right person. That is the whole design. You listen two levels down so your managers can lead one level down, with better information and their authority intact.

If you are about to run your first round and want to compare notes on the routing table or the manager brief, connect with me on LinkedIn. I am always curious which row fills up first on other people’s teams. And if what your skip-levels keep telling you is that a team is short a person, not short a process, talk it through with KORE1’s engineering recruiting team.

Related reading: The Three Skills a Senior IC Has to Unlearn to Lead.