Platform policy · Microsoft 365
Teams noise-suppression policy: what admins actually control
Search for “Teams noise suppression” and the internet hands you client-settings tutorials. Useless from the admin chair. The real question is which parts of Teams' audio cleanup are governable per policy group from your tenant — and the answer is more specific, and more limited, than most rollout plans assume. Here is the surface as Microsoft documents it today.
The lever that doesn't exist (start here)
There is no meeting-policy setting that pins the classic noise-suppression level on your users' desktop clients. As of our 20 September 2026 check of Microsoft Learn's “Manage meeting policies for audio and video,” the Audio & video policy pane governs IP audio/video modes, media bit rate and video effects — not a tenant-enforced suppression level for regular participants. Users choose their own level in the client; the tenant cannot force it.
Write that into your rollout doc, because it reframes the project: on Teams, your governable levers are voice isolation, voice enrollment, and dial-in suppression — plus, if you need a guaranteed org-wide floor across every app, a deployed suppression layer that doesn't depend on per-user client choices at all.
The Teams surface: three levers and a gap
- Lever 1 · Voice isolationCsTeamsMeetingPolicy
-VoiceIsolationDefault: Enabled - Lever 2 · Voice enrollmentCsTeamsAIPolicy
-EnrollVoice-PassiveVoiceEnrollmentDefault: Enabled — isolation depends on it - Lever 3 · Dial-in suppressionCsTeamsMeetingPolicy
NoiseSuppressionForDialInParticipantsMicrosoftDefault, behaving as Enabled - No lever · Desktop suppression levelNo meeting-policy settingUsers choose it in the clientThe tenant cannot force it
Lever 1: voice isolation, per policy group
Voice isolation is the strong option in the Teams stack — an AI model keyed to the individual user's enrolled voice profile that passes that voice and rejects other sounds and other voices. For open-plan offices and shared home workspaces it addresses the case plain suppression can't: a second human talking nearby.
It's governed by -VoiceIsolation on CsTeamsMeetingPolicy, default Enabled, so your real decisions are scoping and enrollment (next section):
# Allow (default) or block voice isolation for a policy group
Set-CsTeamsMeetingPolicy -Identity "CallHeavyRoles" -VoiceIsolation Enabled
Set-CsTeamsMeetingPolicy -Identity "SharedDeviceUsers" -VoiceIsolation DisabledMicrosoft's documentation notes the voice profile is stored on the user's local device and used only for this purpose — the sentence your privacy review will want, with the caveat that they should read it from Microsoft Learn (“Manage voice isolation for your users' calls and meetings”), not from us.
Lever 2: enrollment, the dependency everyone forgets
Voice isolation only works for users who have enrolled a voice profile, and enrollment has its own policy surface — on CsTeamsAIPolicy, not the meeting policy:
# Both default to Enabled
Set-CsTeamsAIPolicy -Identity "CallHeavyRoles" -EnrollVoice Enabled
Set-CsTeamsAIPolicy -Identity "CallHeavyRoles" -PassiveVoiceEnrollment Enabled-EnrollVoicecontrols whether users can capture a voice profile through the Recognition tab in client settings.-PassiveVoiceEnrollmentadditionally covers the low-friction “express enrollment” path. Disable both and voice isolation is dead in practice regardless of the meeting policy — the failure mode to check first when “the feature doesn't work” tickets arrive.
Rollout note: because enrollment is a user action, your deployment artifact here is a comms line (“record your voice profile this week — two minutes”), scoped to the same groups whose policies you enabled. Policy makes it possible; the nudge makes it real.
Lever 3: dial-in participant suppression
The newest genuinely admin-shaped control, rolled out from October 2025 (Message Center MC1135397): noise suppression applied to PSTN dial-in participants — the leaf blower behind the person who joined from a car. The policy property ships with a value of MicrosoftDefault, which Microsoft moved to behave as Enabled during the rollout — so treat it as on by default unless you have explicitly set it, and read the current value rather than assuming. It's managed in the admin center under Meetings → Meeting policies → Audio & video → “Noise suppression for dial-in participants,” or in PowerShell via the NoiseSuppressionForDialInParticipants setting on CsTeamsMeetingPolicy.
Our advice: leave it on everywhere, and note it in the change log rather than running a change ticket — a default-on Microsoft rollout you merely acknowledged is not a change you own.
Scoping: the part that makes it a policy, not a setting
All three levers ride Teams' standard policy machinery, so the deployment pattern is the one you already run for meetings governance:
- Create named policies for real populations —
CallHeavyRoles,SharedDeviceUsers,Default— rather than editing the Global policy reflexively. - Assign by group where populations are dynamic; direct assignment for the exceptions.
- Record the intended state in your config repo. Six months from now, “why is voice isolation off for the training room accounts?” should have a written answer.
# Sanity-check what a given policy actually says
Get-CsTeamsMeetingPolicy -Identity "CallHeavyRoles" |
Select-Object Identity, VoiceIsolation, NoiseSuppressionForDialInParticipantsWhere Teams' levers stop
Honest boundary list, because rollout plans built on imagined levers fail in week two:
- No enforced suppression level for desktop participants — covered above; the users own that dial.
- Teams-only scope. Nothing here touches the softphone, the recording app, or any other place your organization talks. Multi-app estates are exactly where an OS-level layer starts justifying seats.
- Enrollment friction is real. Voice isolation's dependency on user enrollment means its adoption curve looks like a comms campaign, not a policy push. Plan for that or plan to be disappointed by the usage numbers.
- Docs drift. Microsoft reorganizes these policy panes regularly — verify parameter names against Microsoft Learn during change planning, the way we did before publishing this page. (No affiliate relationship exists between this site and Microsoft; the free levers are recommended because they're free.)
When policy isn't enough
If your audit of this page ends with “so the tenant can't actually guarantee clean mic audio in every app” — correct, it can't. That guarantee is what a deployed OS-level layer is for: one package, every app, no dependence on user-side toggles. Krisp's 7-day trial (verified 20 September 2026) lets you test that claim against your configured Teams baseline in a week.
Referral link, declared: when an organization licenses Krisp after clicking this button, Krisp pays RolloutDesk a fee, and the quote you receive is the one you would get going direct. The Teams facts above come from Microsoft's docs and say what they say either way.
Related: the company-wide playbook (where these levers slot in), Zoom's admin audio surface (a usefully different philosophy), and Intune deployment for pushing the layer policy can't provide.