Skip to content
⌘K
InboxAsk a human

Outbound webhooks for Escalation Shield

Updated by Vaishnavi

Outbound webhooks let Escalation Shield send each alert to an external system you own. For example, send alerts about high-risk feedback to your ticketing system or incident queue.

To send feedback into Enterpret, use the Webhook Integration.


Before you start

You need:

  • An Escalation Shield automation in Enterpret

  • An HTTPS endpoint that can accept POST requests with Content-Type: application/json

  • Any token, API key, or custom header value your endpoint requires

  • Permission to edit the automation

Your endpoint must acknowledge accepted requests with a 2xx response. The legacy sender times out after five seconds per attempt, so acknowledge the request first and process longer work asynchronously.


Set up an outbound webhook

  1. Open Automations in Enterpret.

  2. Open the Escalation Shield automation you want to configure.

  3. Click Edit.

  4. In delivery settings, select Webhook and choose Add webhook when offered. Older editors may show this under Alerts.

  5. Enter the webhook URL. The URL must start with https://.

  6. Choose an authentication method: None, Bearer Token, API Key (X-API-Key), or Custom Header.

  7. If you chose Bearer Token, API Key (X-API-Key), or Custom Header, enter the credential value. For Custom Header, also enter the header name. If you chose None, leave credentials blank.

  8. Click Add webhook.

  9. Click Save changes.

You can use a webhook destination alongside Slack and email destinations. Credentials are encrypted at rest.

If you need to send the same alert to several external systems, point the webhook at a relay endpoint that forwards the alert to each system.


Authentication options

Option

Header sent

None

No authentication header

Bearer Token

Authorization: Bearer <credential>

API Key

X-API-Key: <credential>

Custom Header

<custom-header-name>: <credential>


What your endpoint receives

Enterpret sends a POST request with JSON.

Each request includes:

  • Content-Type: application/json

  • Idempotency-Key: <opaque delivery key>

  • The authentication header you configured, if any

The User-Agent value can vary by delivery path. Do not use a fixed value to authenticate requests.

Use the Idempotency-Key header to deduplicate retries. The payload contains the automation ID, stable escalation levels, and a compact record summary.

Example payload

{ "automation_id": "automation-uuid", "org_id": "your-org", "escalated_at": "2026-06-27T12:00:00Z", "levels": { "escalation_level": "extreme", "emotional_intensity_level": "extreme", "sentiment_shift": "sentiment degraded" }, "record": { "id": "record-id", "url": "https://dashboard.enterpret.com/your-org/record/record-id", "source_url": "https://your-source-system.example.com/tickets/12345", "content_excerpt": "The customer message that triggered the alert", "created_at": "2026-06-27T11:58:00Z", "source": "ZendeskSupport" } }

Field reference

Field

Description

automation_id

The Escalation Shield automation that fired.

org_id

Your Enterpret org slug.

escalated_at

When Enterpret constructed the escalation payload; this is not proof of successful delivery.

levels.escalation_level

Discrete escalation level: low, mild, moderate, high, or extreme.

levels.emotional_intensity_level

Discrete emotional intensity level: low, mild, moderate, high, or extreme.

levels.sentiment_shift

Conversation sentiment movement, when available.

record.id

Enterpret feedback record ID.

record.url

Link to the feedback record in Enterpret.

record.source_url

Link to the source record, when available.

record.content_excerpt

An excerpt of the triggering feedback text. Agent-powered payloads can truncate long text and metadata to fit their size limits.

record.created_at

Source timestamp for the feedback, when available.

record.source

Enterpret source identifier.


Delivery behavior

Escalation Shield runs in the background and sends webhook alerts when it detects escalation risk.

Return a 2xx status after accepting the request. Legacy delivery can retry failures. Agent-powered webhook sends do not automatically retry an uncertain or unsuccessful POST because your system may already have processed it. Treat Idempotency-Key as an opaque value and deduplicate repeated values; its format differs by delivery path.


Troubleshooting

I do not see webhook alerts

Check that:

  • The automation is active.

  • The automation has a matching feedback filter.

  • The webhook URL starts with https://.

  • Your endpoint promptly returns a 2xx response after accepting the request.

  • The credential in Enterpret matches what your endpoint expects.

I see duplicate requests

Legacy delivery can retry when it does not receive a successful response. Agent-powered sends report failures without automatically retrying an uncertain POST. Duplicate requests are still possible; deduplicate repeated Idempotency-Key values and inspect delivery results before manually retrying.

My endpoint rejects the request

Confirm the authentication method matches your endpoint:

  • Bearer Token expects Authorization: Bearer <credential>.

  • API Key expects X-API-Key: <credential>.

  • Custom Header sends the credential under the header name you provide.

My endpoint receives the alert but cannot route it

Use automation_id, levels, and record.source to route the alert. If you need more fields from the original record, use record.source_url when it is available. Use record.id to identify the feedback in Enterpret.


Frequently asked questions

Is this the same as the Webhook Integration?

No. The Webhook Integration sends feedback into Enterpret. Outbound webhooks send Escalation Shield alerts out of Enterpret.

Can I still send alerts to Slack or email?

Yes. A webhook can be used alongside Slack and email alert destinations.

What fields should I use for routing?

Use automation_id to identify the automation, levels for escalation severity, record.source to route alerts by source, and record.url, record.source_url, or record.id to connect the alert back to a record.

What should my endpoint return?

Return a 2xx response after accepting the alert. If your processing takes longer than a few seconds, acknowledge the request first and process it asynchronously.