Alert when one article keeps failing to deflect

Detects a sudden spike in failures for one specific article and treats it as an incident. When the same article is shown to customers but escalated several times inside a rolling window, that usually means the answer just broke, went stale after a product change, or is plain wrong, so the workflow pages the owner immediately rather than waiting for the weekly review.

Use this around product launches and pricing or policy changes, when a previously fine article can go wrong overnight. It catches the fast-moving failures that a weekly digest would miss.

1. 1An Intercom webhook fires on each human escalation.
2. 2The workflow records the article-failure event in Postgres and counts failures for that article over the trailing 60 minutes.
3. 3A logic step compares the count to a per-article threshold; below it, the run stops silently.
4. 4When the threshold is crossed, it triggers a PagerDuty incident routed to the docs on-call.
5. 5It posts the article, the recent failing questions, and the failure count to the support Slack channel for context.