Why we use a copilot for every 1:1 (and what it actually catches)

The first time I suggested putting a meeting copilot in my 1:1s, my reports' reactions split exactly in half.

Half of them said "sure, makes sense." The other half went visibly tense and asked questions like "who else can see this?" and "does it go to HR?" Both reactions were correct. A 1:1 is the most psychologically intimate work meeting on most engineers' calendars, and putting an AI listener in the room is not a neutral act.

Three months in, every one of my reports has opted in. Not because I pushed — I didn't — but because the value showed up faster than the discomfort. This is what I learned about why, and what the copilot actually catches that I didn't.

The 1:1 hygiene problem

I manage eight engineers. Each gets a 30-minute weekly 1:1. That's four hours of my week spent in 1:1s.

For most of my management career, I took notes in a Notion page during each call. The page had columns for "topics discussed", "action items", and "themes to revisit". I was diligent about it for the first month with any new report. After that, the quality degraded predictably.

By month three, my notes for any given report looked like this:

2025-12-08 1:1 with Alice

Talked about the migration

She's frustrated with the design review process

Action: I'll follow up with Mike about the staging environment

Three observations about that note:

"Talked about the migration" is a fact about the meeting. It tells me nothing about what Alice said or what I committed to.
"Frustrated with the design review process" is a theme, but I didn't capture why. Next 1:1, I'll have to ask her to re-explain.
"I'll follow up with Mike" is an action item without a date, and the historical evidence is that I'm going to forget about it.

Multiply that note quality by eight reports across thirty weeks, and you have 240 1:1 notes of mediocre quality. None of which are queryable. None of which surface patterns. None of which catch the things I commit to and then forget.

This is the manager hygiene problem: notes are work, work decays, and notes about people are too important to decay.

What we set up

We rolled this out cautiously. Three rules from the start:

The copilot only runs if the report explicitly opts in. No default-on. No pressure. If you decline, your 1:1 runs as it always did.
Transcripts and summaries are visible to the report. They get the full output, immediately after the call. If they want to delete it, they can.
Nothing is shared with HR, skip-levels, or anyone else. The data lives in the 1:1 pair's private space and goes nowhere else.

This took a real conversation to land. Two of my reports asked for a written guarantee. I sent one. They opted in.

We also picked a copilot that supports a "summary only" mode — the transcript is held briefly to generate the summary, then deleted. Some reports chose this. Others chose to keep the full transcript. The optionality was important. The thing being measured is trust, and trust is built by giving people more control than they expect.

What the copilot catches that I missed

After a quarter of weekly 1:1s with the copilot running, I can list the specific categories of things it catches that I didn't.

"I'll get back to you" commitments

This is the biggest one. In a 30-minute 1:1, I commit to roughly three things, on average. "I'll follow up with Mike." "I'll send you the doc on architecture decisions." "I'll think about the team structure question and come back next week."

Without the copilot, I remembered maybe one of those three commitments. The other two got dropped. My report would either re-raise them (annoying for them) or quietly stop expecting me to do them (much worse for the relationship).

With the copilot, those commitments get extracted as action items with my name on them, and they show up in my Monday morning planning. I now follow through on roughly 95% of them. The remaining 5% I consciously decide not to do and tell the report. That's a real change in how trustworthy I am as a manager.

Themes across reports

This one snuck up on me. After two months, the tool surfaced a quiet pattern: three different reports had mentioned, in unrelated 1:1s, that the on-call rotation felt brittle. None of them had escalated it. None of them had filed a ticket. But three out of eight reports had said something like "on-call has been rough lately" within a four-week window.

I would have missed this. Each individual mention was a small thing. The pattern across reports was a real problem. We rebuilt the on-call rotation a month later, and it's measurably better. But I wouldn't have caught the pattern without the copilot aggregating themes across my 1:1s.

To be clear: the copilot doesn't share content across reports. It surfaces themes — anonymized, aggregated patterns — to me. So I see "on-call rotation" as a theme that appeared in 1:1s with three reports, without seeing which three. If I want to know more, I have to either ask people directly or look at my own notes.

The questions I forgot to ask

The copilot has a feature where, before each 1:1, it surfaces "open threads" from previous sessions. These are topics that were discussed but never resolved — questions I asked but didn't get a clear answer to, or things the report said they'd think about and never came back on.

In a typical 1:1, two or three open threads from previous weeks surface. I'd guess I would have remembered maybe one of them organically. The other two would have stayed quietly unresolved for weeks.

A specific example: in October, Bob mentioned he was considering taking a sabbatical in the spring. We talked about it for five minutes and moved on. In November, December, January — I forgot about it entirely. In late January, the copilot surfaced it as an open thread before our 1:1. I asked. He'd been waiting for me to bring it up. The conversation we had was the most important conversation we'd had in months.

That conversation would have happened in March, if at all, without the copilot.

The first time someone says something

There's a particular pattern the copilot is unusually good at: the first time a topic is raised. When Alice first mentioned that the design review process was slow, it was a passing comment buried in a discussion about something else. The copilot extracted it as a flagged concern. Three weeks later, when she raised it again more directly, the system already had the prior context. I could see this had been on her mind for almost a month before she said anything concrete.

That changes how I respond. "Yeah, you mentioned something about this back in early February — sorry I didn't follow up" is a very different opener than "wait, this is new, tell me more."

What the copilot doesn't help with

I want to be honest about the limits. Three things the copilot is bad at:

Reading the room. When someone says "things are fine" with a tone that means "things are not fine", the copilot transcribes "things are fine." Tone, body language, the long pause before a sensitive question — these are not in the transcript. The human in the room still has to pay attention.

Knowing when to stop talking. The copilot can summarize what was said. It can't tell you when the right move was silence. The art of a good 1:1 — making space, sitting with discomfort, letting someone arrive at their own conclusion — is still entirely on the manager.

Replacing trust. If your report doesn't trust you, no amount of structured notes will fix that. The copilot is a multiplier on a working relationship; it's not a substitute for one. If your 1:1s are bad, an AI listener will make them faster but won't make them better.

The unexpected effect on the report

The thing that surprised me most was how my reports felt about getting the summary themselves.

I assumed they'd treat it as a manager artifact — useful for me, not for them. Wrong. Several of them told me they re-read the summary later in the week. Two of them now reference past 1:1s in our current ones ("hey, two weeks ago we talked about X and I wanted to update you"). One of them keeps a running document of action items from across our 1:1s, which she uses to track what I've committed to.

That last one is the most healthy outcome. A 1:1 is a two-way meeting; the manager has work to do that the report should be able to hold them accountable for. The copilot, by giving both parties the same notes, equalizes that loop. I find I'm being held to my commitments by Alice and Bob and Carol in a way that I wasn't before. This is a good thing.

The consent moment is the whole game

If there's one thing I'd tell another manager rolling this out: the consent conversation is everything. Not in the legal sense. In the trust sense.

When I sat down with each report and explained what the copilot would do, what it would record, who would see it, and what their veto rights were, I wasn't selling a tool. I was demonstrating a value: I take your privacy seriously enough to ask. The fact that I asked was, for most of them, the reason they said yes.

The two who initially declined didn't decline forever. Both opted in after a month — one because she saw the action-item discipline in her teammates' 1:1s, and one because he wanted to be able to re-read what he'd said about a topic he was still thinking about.

If you skip the consent moment — if you turn on a copilot without asking, or worse, lie about who sees the output — you don't have a productivity tool. You have a betrayal of trust waiting to happen.

Six months of better 1:1s

The tactical wins from a meeting copilot in 1:1s are easy to list: better follow-through, themes that surface, open threads that get closed. But the strategic effect is something quieter. I'm a better manager than I was six months ago, and the copilot is part of why.

I forget less. I follow up more reliably. I notice patterns across my team that used to slip past me. My reports have more leverage in our relationship because they have the same notes I have. The conversations themselves are still human, still messy, still about things that don't fit in a bullet point — but the scaffolding around the conversations is finally as careful as the conversations deserve.

If you manage people, your 1:1s are your most important meetings. Treating their record-keeping as a memory exercise is, I now think, an underrated mistake.