The automation that broke because nobody owned the API key

A meaningful share of automation outages I've diagnosed had nothing to do with the workflow logic itself. The API key was created under a personal account that got deactivated when someone left the company, or the integration's OAuth token expired and nobody was subscribed to the expiry notification, or a shared credential got rotated for a security reason without anyone checking what depended on it.

This is a governance problem wearing a technical costume. Automations are usually built by whoever's available at the time, using whatever credentials are easiest to grab in the moment, and the question of who owns that credential long-term, and what happens when they leave, rarely gets asked until it's already a production incident.

The fix is a simple registry, not a sophisticated system: every credential an automation depends on gets logged against a service account, not a personal one, with an explicit owner and an expiry-monitoring alert, reviewed whenever someone leaves the team or a credential rotation happens for any reason.

This is unglamorous infrastructure work that never shows up in a demo and consistently prevents the specific failure mode that takes down automations with no warning: not a logic bug, but a credential nobody was watching, quietly expiring on its own schedule.