---
schema_version: "1"
id: ebc87129cc8df6441ff4afabe525c729f60e93c336cc943de0778f86e4b63404
title: Manage Alerts for Agent Health Monitoring
collection: admin
version: current
locale: en-us
portal_url: https://docs.salesforce.com/admin/current/en-us/ai/264-0-0/agent_monitor_health_alert.md
canonical_url: https://help.salesforce.com/s/articleView?id=ai.agent_monitor_health_alert.htm&release=264.0.0&type=5
source_site: help.salesforce.com
content_type: documentation
content_hash: 08f5099c5445fa820d39f4885dcc9f1cc4d846d3ca8bd5806fda297545856482
description: Set up an alert to monitor a metric for a particular agent. You receive an email notification when a performance data breaches an alert threshold.
products:
  - Agentforce
  - Cross Cloud
release: "264"
---
# Manage Alerts for Agent Health Monitoring

Set up an alert to monitor a metric for a particular agent. You receive an email notification when a performance data breaches an alert threshold.

## View Alerts

1. In **Agentforce Studio**, click **Agents**.
2. In the Explorer under **Observe & Optimize**, click **Alerts**.
3. Click the **Alerts** subtab.
    
    The Alerts subtab shows a list of any alerts that you've created.

## Create an Alert

Create an alert to receive email notifications when a performance data breaches an alert threshold.

1. Click **Create Alert**.
    
    ![Create an alert to monitor agent health](https://sf-zdocs-cdn-prod.zoominsoftware.com/tdta-ai-generative_ai-264-0-0-production-enus/f8ef6498-d7b2-44e3-9b34-ffd5a850aae9/generative_ai/images/agent_monitor_health_alert_create.png)
2. Select the metric that you want to monitor.
    
    Metrics are sorted by agent type. Select an available standard metric or a custom metric.
3. Select the agent that you want to monitor.
4. If available, select **At or Above** or **At or Below**.
    
    For some metrics, such as Error Rate, operator selection occurs automatically.
5. Specify a threshold value for the selected metric.
    
    The threshold value represents the selected metric: quantity (for example, Unique Sessions), time period (for example, Agent Response Latency), or percentage (for example, Escalation Rate). See "Best Practices for Determining Alert Thresholds" in this article for suggestions.
6. Save your work.

For most metrics, the system evaluates alerts every 10 minutes based on the previous 15 minutes of data. For Error Rate and Agent Interaction Duration, the system evaluates alerts every minute based on the previous 15 minutes of data. After sending a notification, the system pauses notifications for the same alert for 30 minutes.

![Note](https://sf-zdocs-cdn-prod.zoominsoftware.com/tdta-ai-generative_ai-264-0-0-production-enus/f8ef6498-d7b2-44e3-9b34-ffd5a850aae9/images/icon_note.png)

Note Alert email notifications are sent only to the user who creates the alert. You can't select a different recipient, multiple recipients, or a distribution list.

## Edit an Alert

Change the threshold value that triggers an alert.

1. On the **Alerts** tab, find the alert that you want to edit.
2. Select **Edit** from the drop-down menu.
    
    You can't change the alert's metric or agent selections. Create an alert instead.
3. If available, change the **At or Above/Below** selection.
4. Change the threshold value.
5. Save your work.

Future notifications are triggered based on the revised threshold.

## Remove an Alert

Delete an alert to stop future notifications and remove existing associated notifications. Deleting an alert removes it permanently.

1. On the **Alerts** tab, find the alert that you want to remove.
2. Select **Delete** from the drop-down menu.
3. Click **Delete** to confirm.

## Best Practices to Determine Alert Thresholds

To determine optimal alert thresholds, refer to historical data combined with business-defined acceptable limits.

Step 1: Analyze historical data.

For the metric on which you want to create an alert, pull at least 30 days of history. This window is sufficient for you to understand the normal range and identify anomalies.

Example: Agent Interaction Duration (seconds) over the last 30 days ranged from 1.75 to 6.24 seconds.

Step 2: Define what's acceptable.

Historical data alone doesn't tell you where to draw the line. It’s a business decision. Discuss with stakeholders to agree on an acceptable user experience.

Example: Is a 5-second response time acceptable to users, or does it start to feel like the conversation is stalling?

Step 3: Combine both data points to estimate the threshold.

Overlay the acceptable-response target on top of the historical range:

Example:

- Historical range: 1.75 seconds - 6.24 seconds
- Business-acceptable ceiling: ~5 seconds (longer than this, users start to feel disconnected from the conversation)

Step 4: Decide the value for the threshold level.

Example: Set the alert threshold at 5 seconds, at or above this value, to trigger a notification. Beyond 5 seconds, users start to perceive the agent as unresponsive mid-conversation. A significant delay becomes a real UX problem rather than normal variance.

Step 5: Configure the alert.

Example:

- Metric: Agent Interaction Duration (sec)
- Agent: Car\_Rental\_Service\_Agent
- Condition: At or Above 5
- Evaluation: Alerts are evaluated every minute based on the previous 15 minutes of data.
- Notification cooldown: After sending the alert notification, the system pauses redundant alert notifications for 30 minutes.
- Delivery: Email notification.