Troubleshooting Workflows

Workflow not doing what you expect? Use this guide to find the issue quickly. The most important place to start is the Output tab in the workflow editor.

Workflow Not Running

  • Is it published? Workflows in draft mode do not process live posts. Click Publish when ready.

  • Is it enabled? A disabled workflow will not run even if it is published.

  • Is the trigger configured? Check that the trigger is connected to the correct saved search or event source.

  • Is Listening finding posts? If the saved search has no new matching results, there may be nothing for the workflow to process.

Use the Output Tab

The Output tab shows workflow runs and node-level results. Use it to inspect:

  • Which runs succeeded or failed.

  • Which node failed.

  • The input and output for each step.

  • Structured error messages.

  • Loop iterations and nested outputs where available.

If a run failed, open the failed run and go directly to the failing node. The error message is usually the fastest route to the fix.

Agent Node Not Filtering Correctly

  • Review your prompt. Vague prompts produce inconsistent outputs.

  • Specify the output format. If an If Condition expects yes, tell the agent to return exactly yes or no.

  • Test with real examples. Use posts from the attached search so the test reflects real data.

If Condition Not Branching Correctly

  • Check the variable path used by the condition.

  • Check casing and exact values.

  • Inspect the prior node’s output in the Output tab.

Integration Step Failing

  • Reconnect the integration if authentication has expired.

  • Check required destination fields.

  • Use the Output tab to confirm the workflow is sending the value the integration expects.

Credits Being Used Too Fast

Expensive actions such as enrichment and engagement lookups can burn credits quickly inside loops. Filter first, then enrich only the posts or people you care about.

Use AI to Fix the Workflow

If the Output tab shows an error, use Fix with AI or Edit with AI and include the error message. This gives Jarvis the context it needs to suggest a useful change.