Automation output routing

​

Definition

Automation output routing decides which work surface owns each execution result. Custom Automations default to Chat output. Tracked Issue output remains an explicit choice when repeated work needs Issue fields and lifecycle state.

​

States and modes

An Automation Run uses the exact states received, running, coalesced, skipped, issue_created, completed, and failed. issue_created records that tracked Issue output was created before that Issue follows its own lifecycle. The exact terminal state remains on the Run even when no new Chat or Issue is created.

​

Constraints

• Every real Chat-output execution gets its own conversation. Rudder does not reuse an arbitrary human Chat as a permanent output sink.
• Skipped or coalesced Runs do not create empty conversations.
• A failed Chat-output Run keeps visible partial or fallback evidence.
• Tracked Issue output follows normal assignment, status, Agent Run, Review, and close-out rules after the Issue is created.

​

Boundaries

Output mode does not decide whether work is important, expensive, or long-running. Choose the surface based on the coordination record people need. A governed policy may still require an Issue even when the result could be read in Chat. The Automation Run remains the dispatch record in both modes. A Chat transcript does not replace the Run, and an execution Issue does not replace the Automation definition or schedule.

​

Examples

• Send a daily digest to Chat when operators will read it and ask follow-up questions in the resulting conversation.
• Track a weekly compliance check as an Issue when it needs an assignee, acceptance criteria, artifacts, and a reviewer decision.
• If concurrency policy coalesces a duplicate tick, inspect the Automation Run for the reason. Do not expect an empty Chat or Issue placeholder.