When automation breaks.
WHY IT MATTERS
Automation fails silently, and silent failure is worse than manual work.
WHAT CAUSES FAILURES
A connected system changing Credentials expiring Data in an unexpected format Volume limits reached Service outages Changes to the process nobody reflected
WHAT TO ESTABLISH
Notification when an automation fails.
WHY
Otherwise you discover it through its consequences.
WHO SHOULD BE NOTIFIED
A named person who will act.
WHAT TO BUILD IN
Handling for predictable problems: missing data, invalid formats, unavailable systems.
WHAT TO DO ABOUT UNPREDICTABLE ONES
Fail visibly, and route to a person.
WHY VISIBLY
A failure that stops quietly leaves work undone with nobody aware.
WHAT TO AVOID
Automations that retry indefinitely Automations that continue with partial data
WHY PARTIAL DATA
It corrupts records in ways that are difficult to find later.
WHAT TO ESTABLISH
What the manual fallback is.
WHY
Systems fail, and the work still must happen.
WHAT TO DOCUMENT
How to perform the task manually.
WHY
Nobody remembers once it has been automated for a year.
WHAT TO MONITOR
Whether automations are actually running.
HOW
Periodic checks that expected outputs appeared.
WHAT TO REVIEW
Failures, by cause.
WHAT REPEATED FAILURES INDICATE
A fragile design or an unhandled exception.
