Automation reliability
How do I make an automation retry or fall back when something breaks?
Separate temporary failures from errors that need human review. Retry safe temporary failures with limits, prevent duplicate actions, and route unresolved cases to a fallback queue or manual process.
What to check
- Which failures are safe to retry?
- How are duplicate actions prevented?
- Where do unresolved cases go?