Settings and troubleshooting

Diagnose a workflow or AI agent

Inspect triggers, conditions, and run logs when the expected action does not happen.

Published · Updated

When a workflow or AI agent behaves unexpectedly, trace one event from start to outcome instead of changing several settings at once.

Before you start

You need access to workflow or agent configuration and run logs, plus a suitable test conversation.

Steps

  1. 1

    Pick one case

    Record the conversation time, triggering event, expected result, and actual result.

  2. 2

    Check activation and scope

    Confirm the workflow or agent is enabled and attached to the right inbox or channel.

  3. 3

    Inspect conditions and steps

    Compare conversation data with trigger conditions, then check variables, actions, and external dependencies.

  4. 4

    Read the run log

    Find the last successful step or error, change one factor, and repeat the test.

Check the result

You can identify the stopping point or unmet condition and verify a fix with a new test.