Who can use this feature?
Plan: Call Center AI
Managed from: Admin Portal
User type: Admin
Agent voice authentication is a Voice Security feature that helps you detect agent substitution: cases where someone other than the assigned agent handles live calls. When it is on, Krisp builds a voice baseline for each agent and monitors voice consistency across calls, flagging an agent when an unexpected voice change is identified.
The feature is turned on by an Admin and applies to the whole team.
Info
Agent voice authentication uses voice biometric data to monitor voice consistency. Agents are asked to consent before any voice data is collected. Review your team's consent and data handling requirements before turning it on.
What Agent voice authentication does
Once enabled, Krisp establishes a voice baseline for each agent and compares every call against that baseline. If an unexpected voice change is identified, the agent is flagged so your team can review it. Monitoring runs in the background and does not interrupt the agent during a call.
How to enable Agent voice authentication
- In the Admin Portal, go to Team settings >>> Security >>> Voice security
- Find Agent voice authentication and turn the toggle on.
How agents are enrolled
When an agent first logs in after the feature is enabled, Krisp asks them to consent to voice collection. If the agent declines, no voice data is collected and the feature stays inactive for that agent.
For agents who consent, enrollment happens in one of two ways:
- Automatic enrollment: Krisp builds the agent's baseline passively from their calls. After enough consistent calls, the baseline is committed and monitoring begins. The agent's status becomes Auto-enrolled.
- Admin-verified enrollment: for agents who need stronger verification, an Admin can run a supervised verification session to capture a confirmed baseline. The agent's status becomes Admin-verified, the highest trust level.
Verification statuses
Each agent has a verification status, shown in the Verification status column under Directory >>> Teams (and Directory >>> Users).
| Status | What it means |
|---|---|
| Pending | Krisp is collecting voice data but has not committed a baseline yet. |
| Rejected | The agent declined voice collection. The feature is inactive for that agent. |
| Auto-enrolled | A baseline was committed automatically from the agent's calls. Monitoring is active. |
| Admin-verified | The baseline was confirmed through a supervised verification session. This is the highest trust level. |
| Possible mismatch | A voice deviation was detected that should be reviewed. |
Reviewing voice mismatch alerts
To review alerts, go to Directory >>> Teams and open a team, then select an agent to open their details. On device-based teams, agents appear in the Usernames tab; on email-based teams, they appear in the Users tab.
The agent's details panel shows their verification status and a list of voice mismatch alerts, each with a date and a Call ID. The Call ID appears when Krisp Bridge is enabled.
For each alert, an Admin can take one of two actions:
- Not a mismatch: marks the alert as a false positive. The voice from that call is added to the agent's baseline, and monitoring continues against the updated baseline. Use this when the change is legitimate, for example a new headset or the agent being unwell.
- Mark as resolved: dismisses the alert without changing the baseline. Monitoring continues against the existing baseline.
You can also use Reset in the agent's details panel to clear the current enrollment and restart voice collection for that agent.
Important
Marking an alert as "Not a mismatch" adds that call's voice to the baseline. Only do this when you are confident the voice belongs to the assigned agent, since adding an unauthorized voice would weaken future detection.
Audit logs
For a full compliance trail across the team, go to Team settings >>> Security >>> Audit logs. This records Agent voice authentication events, such as the setting being turned on or off, consent decisions, enrollment, mismatch alerts, and the actions Admins take on those alerts. To learn more about this section, refer to Audit Logs.
FAQ
Agents are asked to consent before any voice data is collected. They are not notified when a specific mismatch alert is raised.
No voice data is collected for that agent and the feature stays inactive for them. Their status shows as Rejected, so an Admin can follow up directly.
An Admin can mark the alert as "Not a mismatch". This adds the new voice to the agent's baseline and prevents repeated false positives for the same voice going forward.
Yes. Voice data is scoped to the team where consent was given. If an agent is on multiple teams, the feature applies only on the team where it is enabled and consent was collected.
An Admin can use Reset to restart voice collection for that agent, or run a supervised verification session to set a confirmed baseline manually.
The Call ID appears when Krisp Bridge is enabled. Without Bridge, alerts still show the date but not a Call ID.