Hybrid Meetings: How Copilot Knows Who Is in the Room

For Microsoft 365 Copilot to know the people in the room by name, not just the people dialing in, you need three things: a room with Teams Rooms Pro and recognition turned on, a voice profile for everyone in the room, and a scheduled meeting where everyone is invited and transcription is running. Miss one, and everything said around the table gets credited to the room.

You’ll notice it in the recap, if not sooner. Remote attendees show up by name, the tasks are assigned, and everything said in the room is attributed to “Conference Room 1.”

Why Copilot doesn’t know who’s in the room

Because Teams can’t tell them apart. Remote attendees join with their own account and their own microphone, so Teams knows who said every sentence. In the meeting room, everyone talks into the same microphone, and the one signed in is the room device, not the people in front of it. Without speaker recognition, Teams attributes everything said there to the room.

Everything Copilot does is built on that. Microsoft calls the identity of each speaker the most essential input for Copilot in Teams: summaries, tasks, and answers all come from the transcript. Ask afterwards what a colleague in the room said about the budget, and the best you’ll get back is “Conference Room 1 said …”

Here’s what ends up in the transcript:

What the meeting room needs

A Teams Rooms device with a Teams Rooms Pro license on the room account. Teams Rooms is the built-in Teams system in a meeting room, with a console on the table. The free Teams Rooms Basic license (for up to 25 rooms) can’t identify speakers; Microsoft lists speaker identification as a Pro feature.

On top of that, there’s a switch on the room device that IT should check once. Microsoft doesn’t quite agree with itself on the default: the table on Microsoft Learn says off, the hint in the portal says on. So check rather than assume. You set it in the Teams Rooms Pro Management Portal or right on the console in the room.

You don’t need a new microphone. Speaker recognition now works on all Teams Rooms devices, Windows and Android alike. According to Microsoft, certified “intelligent speakers” with multiple microphones are more accurate, and for rooms where every word counts, Microsoft suggests taking a look at them.

What everyone in the room does once

Create a voice profile. Think of it as a fingerprint of your voice that lets Teams recognize you in the room. In the Teams desktop app, go to Settings and more (…) > Settings > Recognition and look under Voice recognition. You have two options:

• In the background: You opt in, and Teams builds your profile from what you say in your next meetings. Microsoft says this typically takes a couple of meetings.
• Right away: You read a short passage into your microphone. Pick a quiet spot and a good mic.

Only people in the room need a profile. Teams already knows remote attendees by their account. If Recognition is missing from your settings, your IT team has turned it off; by default, it’s available to everyone.

How to set up the meeting so it works

This part is up to whoever sends the invite:

1. Schedule it; don’t start it on the fly. Teams matches the voices in the room against the invite list. A meeting started on the spot from the room console has no list to match against.
2. Invite the room and every person individually, including the people who’ll be sitting in the room. A forwarded invite counts too. Anyone who isn’t on the invite won’t be recognized in the room, even with a voice profile.
3. Start transcription. No transcript, no recognition.
4. Bring your laptop, with the mic and speakers off. Attribution only starts once someone other than the room device has joined. In a hybrid meeting, the remote attendees take care of that. If you’re all in the room, have at least one person join from a laptop as well. Microsoft recommends that everyone in the room join from their own device anyway, for chat and reactions, with the mic and speakers off so there’s no echo.
5. Keep an eye on size. Microsoft says recognition is most accurate with up to 10 people in the room; more can join. There’s a hard limit on the invite: if more than 50 invitees have a voice profile, recognition stops working. So it’s not the tool for your department all-hands.

What changes for Copilot and the recap

With voice profiles, everything said in the room lands under the right name: in the live transcript, in the AI notes, and in what Copilot tells you afterwards. Microsoft says so explicitly for transcripts, meeting recaps, and Copilot. A task like “send the proposal by Friday” shows up next to the person who took it on at the table, not next to the room.

It helps to keep two things apart:

According to Microsoft, the full recap after the meeting, with AI notes, tasks, and a timeline of who spoke when (intelligent recap), requires Teams Premium or Microsoft 365 Copilot. The speaker timeline only shows up if the meeting is recorded, too.

What about names on the video?

That’s the optional second part: face recognition. Your voice makes sure what you say lands under your name in the transcript. Your face puts your name on the video and in the participant list, for everyone watching remotely. Here’s what it takes:

• Teams Rooms on Windows; Teams Rooms on Android doesn’t support face recognition
• a supported camera, either with Cloud IntelliFrame (which zooms in on each face in the room for remote attendees) or a certified multi-stream camera
• a face profile on top of the voice profile, created in the same place in Teams (nine photos)
• a sign outside the room letting people know recognition is in use, which Microsoft explicitly calls for

The limits are tighter here: no more than 12 people in the room and no more than 64 invitees.

Does everyone have to take part?

No. Voice profiles are opt-in: each person decides whether to create one and can delete it at any time. Anyone without one still shows up as “Speaker 1.” Teams deletes profiles that haven’t been used for a year. They’re stored in your organization’s tenant, and Microsoft says it doesn’t use them to train any models.

During the meeting, you can correct course in two ways. You can turn recognition off for the current meeting on the room console, and anyone can flag a misidentified person in the participant list, which turns that person into a guest for the rest of the meeting. If you show up as “Speaker 1” in the transcript, you can claim your own lines afterwards, but only your own.

Since these are biometric profiles, talk to your privacy team, and to employee representatives if you have them, before you turn this on. It’s less of a big deal than it sounds, and the key answers are already above: profiles are opt-in, can be deleted at any time, stay in your own tenant, and aren’t used to train models. On top of that, only people on the invite get recognized. The conversation is still worth having, for a practical reason: recognition only works for people who choose to create a profile, and people are more likely to do that when they know what happens to their voice.

And without Teams Rooms, just a laptop and a speakerphone?

That can work too, with a different requirement. Without a Teams Rooms device, the speakerphone is plugged into one person’s laptop via USB; Microsoft calls that person the room host. The room host needs Teams Premium or Microsoft 365 Copilot. Everyone in the room joins from their own laptop and picks Room audio before joining. Anyone with a voice profile then shows up by name in the transcript and AI notes.

One caveat: Microsoft’s documentation isn’t consistent here. The same page that lists the requirements for these bring-your-own-device (BYOD) rooms also says elsewhere that BYOD scenarios aren’t supported for identification. Try it once in your own room before you rely on it.

What IT sets up once

For whoever implements this, here’s everything in one place:

• License: Teams Rooms Pro on the room’s resource account.
• Room device: Turn on voice recognition (and face recognition, if you want it) in the Pro Management Portal under Rooms > Settings > Recognition, or on the console. Microsoft gives conflicting defaults, so check. Microsoft recommends configuring it through device settings; the older meeting-policy route is slated for retirement.
• Voice profiles: The Teams AI policy CsTeamsAIPolicy controls whether people can create profiles (EnrollVoice, PassiveVoiceEnrollment for the background option, EnrollFace). All three are enabled by default and can only be changed in PowerShell.
• Transcription: Allow it in the Teams meeting policy, and allow recording too if you want the recap to show speakers.
• Calendar: Recognition only works if the organizer’s mailbox is in Exchange Online. It doesn’t work with on-premises Exchange or third-party calendars like Google Calendar. In hybrid Exchange setups, Teams has to be able to reach the organizer’s calendar in the cloud.
• Patience: Policy changes can take up to 48 hours to kick in.

What you can do this week

1. If you sit in meeting rooms, create your voice profile: Settings > Recognition.
2. Open the recap of your last hybrid meeting and check whether it shows people’s names or a room name.
3. Ask your IT team whether your rooms have Teams Rooms Pro and recognition turned on.
4. Schedule your next hybrid meeting with the room and every person individually on the invite, then start transcription.

If you want to set up hybrid meetings in your team so Copilot knows afterwards who committed to what, team coaching is the place to start.

As of September 2026. Sources: Microsoft Learn — Enable People Recognition on Teams Rooms, Manage voice recognition for Intelligent Speaker, Overview and configuration of voice and face enrollment, Microsoft Teams Rooms licenses, Cloud IntelliFrame, Multi-Stream IntelliFrame, Intelligent recap for Teams calls, meetings, and events, Teams meeting participant best practices, Set-CsTeamsAIPolicy; Microsoft Support — Use Microsoft Teams Rooms to identify in-room participants in a meeting transcription, Create your Microsoft Teams recognition profile. Retrieved September 2026.