Pick the destination before adding the integration
Choose a channel that has an owner and a clear purpose. A shared operations channel works well for vendor and uptime alerts; incident discussion can happen in a separate channel or thread. Decide who is expected to respond and which changes deserve attention.
Do not send every status update to every team. Route a payment-provider incident to the people who own checkout, and an identity-provider incident to the people responsible for authentication. Recovery notifications are useful when responders need to know when to verify service restoration.
Connect the channel
ServiceAlert supports Slack, Microsoft Teams, Discord, Google Chat, email, and webhooks depending on the plan. Open notification settings in the dashboard, choose the integration, and follow the current setup instructions shown there. The messaging platforms change their administration screens over time, so the in-product flow is the source of truth.
Send a test message before relying on the channel. Confirm that the right people can see it and that the link returns them to useful incident context.
Treat webhook URLs as credentials
A webhook URL can usually post to the destination without another login. Keep it out of tickets, screenshots, source control, and shared documents. If it is exposed, revoke or rotate it in the messaging platform and update the integration.
Limit who can change alert destinations. An audit trail should make it possible to identify when a route changed and who changed it.
Reduce noise deliberately
Start with the services that can materially affect customers or internal operations. Add lower-impact dependencies only when someone has agreed to own the alerts. If a channel becomes noisy, fix the routing or monitor selection rather than asking everyone to pay more attention.
Keep a second route for failures of the primary collaboration tool. Email or another team channel may be enough; paging is appropriate only for events that require an urgent response.
Review the whole path
An integration is working only when an alert reaches a person who knows what to do. Test the path periodically, including recovery. Keep the first response step or runbook near the alert destination, and remove routes that no longer have an owner.