Leading Hybrid Teams Without a Two-Tier Culture

The clearest sign that a hybrid team has split into two tiers is a decision that gets made twice. Once in the room, over lunch or at someone's desk, and again later in a Slack thread for the people who weren't there. The second version is always a little thinner, a little more defensive, and it arrives a day late. I have watched a payments platform team make the same architecture call three times in a week because nobody could agree on where the real decision had happened.

I run engineering at a regulated fintech, and roughly half my people are remote on any given day. I did not set out to write about hybrid work. I set out to stop losing good remote engineers who quietly concluded they were second-class citizens, because they were right, and I was the one who let it happen.

The proximity tax is real and it compounds

There is a tax that gets levied on anyone not physically present, and most managers never see it because they are the ones collecting it. It shows up as the promotion that goes to the person the director bumps into, the stretch project handed to whoever was standing near the whiteboard, the benefit of the doubt extended to the engineer whose face is familiar. None of this is malice. It is just how human attention works, and attention is the currency of career progression.

The compounding is what makes it dangerous. A remote engineer who misses one hallway conversation is fine. One who misses two hundred of them over a year has a materially different understanding of where the org is heading, which projects have air cover, and which battles are worth picking. By review season the gap in context reads as a gap in judgment, and it gets written up as one. I have signed off on calibration ratings that, looking back, were partly a measurement of who had bandwidth to be in the office on Thursdays.

Naming the tax is the first honest thing you can do. I now say it out loud in leadership meetings: if a remote person is underperforming, prove to me it is about the work and not about the seat. Most of the time the manager can. Sometimes they go quiet, and that silence tells me everything.

Every decision needs a written address

The single highest-leverage change we made had nothing to do with tooling. It was a rule: if a decision was not written down somewhere durable, it did not happen. Not in a DM, not in a meeting nobody minuted, not in someone's head. A durable address means a linked document, a ticket comment, an ADR, an RFC. Somewhere a person joining on Tuesday can find it.

This sounds bureaucratic and I braced for people to hate it. The opposite happened. The engineers who liked it most were the ones in the room, because they were tired of being asked the same question five times by five people who had missed the conversation. Writing the decision down once was cheaper than re-litigating it, and it turned out the in-office folks were paying a tax too, just a quieter one.

If the only record of a decision is that some people remember being convinced, you do not have a decision. You have a rumor with good attendance.

Pick a lane: fully in the room or fully remote

The worst meeting format ever invented is the one where six people sit around a conference table and two others dial in from home. The remote pair become disembodied voices that everyone forgets to unmute for, staring at the back of a laptop propped at the end of the table, watching people they cannot hear laugh at something they cannot see. I banned that format after one of my strongest backend engineers told me she had stopped speaking up in those meetings entirely because the latency of interrupting a live room from a webcam was too high.

Now the rule is binary. Either everyone is physically in the room, or everyone is on their own screen with their own camera and their own square in the grid. No hybrid rooms for anything where decisions get made. When three people happen to be co-located, they still take the call from their own desks. It feels absurd, three people in the same building each staring into their own webcam, and it works, because it collapses the two tiers into one.

For the social stuff, the offsites, the team lunches, the whiteboard jams, we lean hard into being in person and we are honest that those are in-person events. What we do not do is let load-bearing work decisions happen there by accident.

Default to writing, reserve talking for the hard parts

A team that writes well is a team where presence matters less. When the design lives in a document, the debate lives in the comments, and the rollout plan lives in the ticket, then being physically present is a nice-to-have rather than a prerequisite for contributing. I would rather read a tight one-page proposal than sit through the meeting that produced a worse version of it.

This is not free. Writing is slower than talking, and some people are genuinely better in conversation than on the page. So we do not write everything. The rule I use is roughly this:

• Anything that changes an interface, a data model, or a public contract gets written before it gets built.
• Anything reversible and cheap can be decided in a call, as long as the outcome lands in a durable address afterward.
• Anything contentious gets written first, precisely because writing forces the disagreement into the open where remote people can engage with it on equal footing.
• Emotionally loaded or ambiguous conversations happen live, on video, because text is a terrible medium for repairing trust.

Promotion parity is where you find out if you meant it

You can talk about fairness all you like. The place it becomes real is the promotion committee. For two cycles I tracked the office-attendance pattern of everyone who got promoted, and the correlation with in-office days was uncomfortable enough that I stopped pretending it was a coincidence.

So we changed the evidence standard. A promotion case now has to stand on artifacts a stranger could evaluate: shipped systems, design docs, incident write-ups, mentorship that left a trail. "Everyone knows they're great" is no longer an acceptable argument, and I will push back on it in the room even when I happen to agree that the person is great. If the case cannot be made from the record, the case is not ready, and that discipline protects the remote engineer whose work speaks louder than their visibility.

The pleasant surprise was that this raised the bar for the in-office people too. A few coasting careers got exposed once presence stopped counting as a contribution. I lost one person over it. I would make the trade again.

Onboarding is where the two-tier culture is born

New hires reveal the true shape of your culture faster than anyone, because they have no accumulated goodwill to cushion the gaps. A remote joiner who spends their first two weeks unsure who to ask about the deployment pipeline is learning, correctly, that this place runs on tacit knowledge held by people in a building they are not in.

We rebuilt onboarding around the assumption that the new person is remote and has no one to tap on the shoulder. Every setup step is written and tested by the last person who joined. Every team has a documented map of who owns what. Each new engineer gets an assigned buddy whose actual job, measured and asked about, is to answer the questions that a co-located hire would have absorbed by osmosis. We watch one number in particular: the gap between remote and in-office joiners in time to first merged pull request. When that gap closed, we knew the tacit-knowledge problem was actually solved rather than merely hidden.

Managers set the temperature, whether they mean to or not

Culture on a hybrid team is set almost entirely by where managers put their attention. If I schedule my important one-on-ones for the days I am in the office, I have told my remote reports exactly where they stand. If I do my thinking out loud in the kitchen, the people not in the kitchen are thinking alone.

I ask my managers to audit themselves on this quarterly. Who did you give the interesting work to, and was it the person nearest the desk? Whose growth are you actively invested in, and does that list skew toward the faces you see? Nobody enjoys this exercise. It is far more useful than another engagement survey, because it puts the responsibility where the leverage actually sits, on the individual leader making a hundred small allocation decisions a week.

Be precise about what in-person is genuinely better for

None of this is an argument that office time is worthless. It plainly is not. The mistake is treating "in person" as a general-purpose good rather than a specific tool with specific uses. Being in a room together is excellent for the ambiguous, high-bandwidth, trust-building work: the messy early phase of a design where nobody knows the shape of the problem yet, the repair of a strained relationship, the deep review of a serious incident where tone matters as much as content.

It is a poor use of a commute to sit alone at a desk answering Slack, which is what a lot of mandated office days actually are. So we gather with intent. We bring the team together for the hard, human, generative work, and we protect the rest of the time for the focused solo work that a quiet home office does better than an open-plan floor. The office is a tool, not a loyalty test, and treating it as the latter is how you build the two tiers you claim to be avoiding.

Conclusion

If I could give one instruction to a leader inheriting a hybrid team, it would be this: assume the remote experience is the real experience, and design for it. The in-office people will be fine either way, because presence forgives a lot of process debt. The remote people will only be fine if the system was built with them at the center. Build for the person who is not in the room, and you will find, quietly, that you have built something better for everyone in it too.