Automations
Read and recover failed runs
An automation run is one attempt to carry out the task in its saved schedule. Run history helps you answer three questions: Did it start? What happened? What should change before the next attempt? Read the run before editing the automation or creating a replacement. A single failed run usually points to a specific input, connection, approval, or limit problem.
Open the run history
- Open Automations and select the automation.
- Open its Activity or Run history section.
- Select the newest run. Start with its status, then read the result, error, and activity details.
- Check whether a later run already succeeded before changing anything. A scheduled run may have recovered on its own.
The status tells you what happened:
- Running means the attempt is still in progress. Do not start another copy just because it has not finished.
- Done means the run completed. Read the result to confirm it did the work you expected, not only that it avoided an error.
- Failed means the run stopped with an error. The error and activity usually identify the next fix.
- Needs approval means the agent reached an action that is waiting for your decision. Review the requested action; approve it only if it is expected.
- Held means a scheduled job was deliberately prevented from running. Check the automation's schedule or hold setting instead of treating it as an app failure.
- Unfinished means Otto can no longer see an active process for a run that still says it was running. Treat it as stopped without assuming that work after the last recorded activity completed.
If an action is waiting for approval, approval is the recovery step; do not loosen the automation's permissions just to make the status disappear.
Fix the cause, then test
Use the exact error as your guide:
| Run shows | What to check and do |
|---|---|
| Device offline or unavailable | Bring Otto online. Confirm the automation is still enabled, then choose Run now when you want to test immediately. |
| Connection unavailable or the account needs attention | Open Integrations, reconnect or reauthorize the app, then return to the automation and confirm its saved app access. Run it again. |
| Insufficient access | Identify the action that was blocked. If the task really should send, publish, purchase, or edit, open Edit, raise that app's access only as far as needed, and Save. Read-only access is enough for monitoring and summaries. |
| Needs approval | Read what the agent wants to do and why. Approve the expected action, or edit the task to avoid or narrow that action. |
| Limit reached | Narrow the task, reduce how often it runs, or raise the limit deliberately. A limit failure is not a reason to grant broader app access. |
| The request was unclear or produced the wrong work | Choose Edit and specify the source, desired result, stopping point, and what to do when a decision is needed. |
| The result went to the wrong place | Open the delivery settings, choose the intended destination and notification behavior, save, then run a test. |
| Held | Check whether the schedule was intentionally paused or held. Resume it only when you want scheduled runs to continue; use Run now for a deliberate test if available. |
For a run that failed during an external change, such as sending a message or updating a record, inspect the destination before retrying. The first attempt may have reached the app even if the run later reported an error; avoid creating duplicates.
Rerun safely
- Make the smallest change that addresses the error.
- Select Save if you edited the automation.
- Choose Run now when you need an immediate test; otherwise let the next scheduled run verify the fix.
- Reopen the new run and check its status, result, destination, and activity.
- If it fails again, compare the new error with the original before making a second change.
Do not increase permissions, rewrite the task, and add a new connection all at once. Changing one control at a time preserves a clear explanation of what fixed the workflow. If the automation should not start another scheduled run while you investigate, choose Pause; pausing does not cancel a run already in progress.
When to create a new automation
Keep the existing automation when the owner, schedule, destination, and overall job are still right. Edit it when only the task, access, approval behavior, limit, or delivery setting needs to change. Create a separate automation only when it is genuinely a different responsibility or needs a different owner, schedule, or result destination. Do not duplicate an automation to recover from one failed run.