> For the complete documentation index, see [llms.txt](https://developer.harness.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://developer.harness.io/ai-sre/3.0/ai-sre-for-administrators/set-up-on-call-management/define-escalation-policies.md).

# Configure Escalation Policies

Escalation policies control what happens when an on-call responder does not acknowledge a page.

By setting up multi-level escalation chains, you ensure that incidents always reach someone who can respond, even if the primary responder is unavailable.

### Create an escalation policy <a href="#create-an-escalation-policy" id="create-an-escalation-policy"></a>

{% embed url="<https://app.tango.us/app/embed/004cfc98-17f2-4947-84a3-75dfbb3aeeaf?skipCover=true&defaultListView=false&skipBranding=false&makeViewOnly=false&hideAuthorAndDetails=true>" %}
Create Escalation Policy
{% endembed %}

1. Navigate to **On-Call** → **Escalation Policies**.
2. Click **Create Escalation Policy**.
3. Configure the policy:
   * **Name:** A descriptive name (for example, "Payments Team Escalation").
   * **Levels:** Add one or more escalation levels. Each level defines:
     * **Target:** A specific user or group to notify.
     * **Timeout:** How long to wait (in minutes) for acknowledgment before escalating.
     * **Retries:** Number of retry attempts at this level before moving to the next.
4. **Attach a schedule** to the policy so the system knows which on-call rotation to page at the first level.
5. Click **Save**.

***

### How escalation works <a href="#how-escalation-works" id="how-escalation-works"></a>

When a page is triggered:

1. The **current on-call responder** (from the attached schedule) is notified through their configured notification channels. If the first channel produces no acknowledgment, AI SRE automatically falls back through the responder's remaining channels before the escalation level times out. Go to [Configure Notification Fallback](/ai-sre/ai-sre-for-administrators/set-up-on-call-management/notification-fallback.md) to understand the multi-channel fallback order and retry logic.
2. If no acknowledgment is received within the **timeout period**, the system retries based on the retry count.
3. If the first level is exhausted, the page **escalates to the next level**, typically a team lead, manager, or secondary on-call group.
4. This continues through all configured levels until someone acknowledges.

***

### Best practices <a href="#best-practices" id="best-practices"></a>

Follow these practices when you design escalation policies:

* **Always configure at least two levels:** A primary responder and a backup prevents incidents from going unacknowledged.
* **Set appropriate timeouts:** Too short and you escalate before someone has time to respond; too long and incidents sit idle. A common starting point is 5 to 10 minutes per level.
* **Use groups at higher levels:** Escalating to a team or group at level 2 or higher increases the chance of a quick response.
* **Attach the right schedule:** Make sure the escalation policy points to the correct on-call rotation. A misconfigured attachment pages the wrong team.
* **Test the full chain:** Run a test page through every level of the policy before relying on it in production. Verify that notifications reach each target correctly.

***

### Next steps <a href="#next-steps" id="next-steps"></a>

* Go to [Configure Notification Fallback](/ai-sre/ai-sre-for-administrators/set-up-on-call-management/notification-fallback.md) to control how each responder is notified across channels.
* Go to [Route Alerts](/ai-sre/ai-sre-for-administrators/set-up-on-call-management/configure-alert-rules.md) to invoke an escalation policy when monitoring alerts fire.