Skip to content
  • There are no suggestions because the search field is empty.

Improve the Accuracy of Decisions AI Meeting Summaries with Microsoft Teams Voice Recognition

Ensure better speaker attribution in hybrid and in-person meetings by configuring your Teams Rooms and BYOD setups for voice recognition.

How voice recognition enhances Decisions AI summaries

Decisions AI generates smart, speaker-aware meeting summaries using Microsoft Teams transcripts. When accuracy matters—like attributing action items or decisions to the right person—Microsoft Teams' voice recognition is a key ingredient.

This article helps you set up voice recognition in Teams Meeting Rooms and BYOD (Bring Your Own Device) environments, so Decisions AI can deliver clearer, more reliable summaries.

1. Understanding your meeting room setup

Knowing your room type is essential for configuring voice recognition correctly.

Teams Meeting Room Setup

  • Room has a Teams console or control panel
  • Participants can join without connecting a laptop
  • The room often appears in the Teams participant list or video feed
BYOD setup:
  • Someone connects a personal laptop to the room’s screen, camera, or mic

  • The meeting starts from that laptop

Each setup has unique requirements for enabling speaker attribution.

2. License requirements for voice recognition

Teams Meeting Room

  • The room's resource account needs a Teams Rooms Pro license
BYOD setup:
  • The person connecting the laptop must have either:

    • Teams Premium license
    • or Microsoft 365 Copilot license

3. How to check and assign licenses

Licenses are assigned in the Microsoft 365 Admin Center:

  • For users (BYOD):

    Users > Active users > [Select user] > Licenses and Apps
  • For Teams Rooms:

    Users > Rooms & equipment > [Select room] > Licenses and Apps

Note: Only users and rooms with the correct licenses will have access to voice recognition features in Teams meetings.


4. Why voice recognition matters for Decisions AI users

When your setup is correctly configured, you get:

  • Voice-based identification of in-room participants

  • Speaker-attributed transcripts - improves the accuracy of Decisions AI summaries when it matters who said what

  • A better experience in hybrid and in-person meetings, where remote speaker identification is limited

💡Even though speaker attribution isn't required for Decisions AI to work, it adds valuable context - especially when tasks or decisions need to be linked to specific individuals.

 

5. Voice profile requirements

Each speaker must set up a voice profile with Microsoft to be recognized.

Without a voice profile:

  • The user will appear as a generic speaker (e.g. “Speaker 2”)

  • They must manually assign the label to themselves during the meeting

🔁 Manual assignment must be done every meeting until a voice profile is set up and consistently recognized.

 

6. Invite participants properly

To enable speaker recognition:
  • Participants must be invited to the meeting for Microsoft to connect their voice to their user profile. 
  • Limit to 20 enrolled users max - speaker attribution is disabled above this limit
  • Optional: In-room participants can also join from their own devices and select the room audio to improve identification:


7. Setup checklists by room type

For Teams Meeting Rooms

  • Use certified Teams Room hardware
  • Join via the console (not a laptop)
  • Ensure the room account has a Teams Rooms Pro license

For BYOD setups

  • Connecting user has a Teams Premium or M365 Copilot license

  • Turn off noise suppression for the connected user to avoid voice loss

  • Only the connected user will be identified if the setup is incomplete


8. Best practices for voice recognition

To get the best results from voice recognition, especially during testing or in new rooms, follow these tips:

✅ Encourage early participation

Microsoft needs more than short answers like "yes" or "no" to identify a speaker.
Start the meeting with brief introductions or quick updates from each participant to help the system build voice recognition context.

✅ Monitor the Transcript Panel

While it's not required for voice recognition to function, it's highly recommended in your early meetings. Keeping an eye on the transcript panel helps the organizer verify that each participant is being correctly identified.

9. Summary of Best Practices

 
Best Practice Why It Matters
Use certified rooms or correct licenses Enables voice recognition
Turn off noise suppression (BYOD) Captures all voices in the room
Encourage early participation Helps Microsoft recognize each speaker
Monitor the Transcript Panel Confirms speaker tagging during setup/testing
Ensure users have voice profiles Enables automatic attribution
Invite all participants Links voice to identity
Limit to 20 enrolled users Maintains recognition quality
Optional: join from personal device Improves in-room voice detection
Assign unidentified speakers manually Keeps summaries accurate

10. Managing unidentified speakers

If someone is labeled as "Speaker 1," they can assign the label to themselves:

  1. Open the transcript panel
  2. Find the correct voice label (like “Speaker 2”)
  3. Assign it to yourself

👤 You can only assign speaker labels to yourself—not to others.

 

 

Helpful Links from Microsoft

For setup support and troubleshooting: