Skip to main content

Managing Tasks

Once a task is created, you can update its prompt or trigger at any time, check what each run did, and troubleshoot runs that did not complete.

Editing a Task

From the task detail page you can update a task at any time without creating a new one.

The Task tab shows the prompt and the trigger configuration. Edit the prompt directly in the text field and save to apply the change to future runs. Use the Edit button in the Triggers section to change the trigger type, folder, schedule, or source pattern.

The toggle at the top of the page enables or disables the task. A disabled task does not run on its schedule or respond to file events. Use this to pause a task while you adjust the prompt without deleting it.

The Run Now button runs the task immediately regardless of the trigger setting. Use this to test changes or to run the task on demand outside its normal schedule.

Viewing Task Results

Every AI Task run creates an AI chat session. You can see the full transcript of what the AI did, including each step it took, the files it accessed, and the output it produced. The Run Logs tab on the task detail page links to the chat session for each run. You can also find task-originated sessions in Chat Logs, where they appear with the task name as the initiator.

Sessions from in-progress runs can be viewed while they are still running. Use the refresh button on the session detail page to see the latest messages as they arrive.

When a Run Fails

Tasks fail for a few common reasons:

  • The folder path in the prompt does not exist or is spelled differently than the actual folder on the site.
  • The output folder the AI was told to write to does not exist.
  • The file is a format the AI cannot read, such as an image, audio file, or a password-protected PDF.
  • The task does not have permission to read a folder or write to the output location.
  • A File Action trigger fired, but the file was moved or deleted before the task started.
  • The prompt is too vague for the AI to determine what to do, so it stops without producing output.

A failed run may be partially completed. Because AI Tasks work through a sequence of steps, any action the AI took before the failure point happened and is not rolled back. Steps after the failure were skipped. For example, if the AI read a file and moved it to a staging folder but then failed while writing the output report, the file move happened but the report does not exist.

The chat session is still created for every run, including failed ones. Open it in Chat Logs to see exactly where the run stopped, what the AI had already done, and what error it encountered.

Failed runs do not retry automatically. If the task is on a schedule, the next scheduled run fires as normal. To re-run immediately after fixing the prompt or the underlying issue, use the Ad-Hoc trigger.

No alert is sent when a run fails. If your workflow depends on the task completing, set up a NotificationExternal LinkThis link leads to an external website and will open in a new tab on the output folder so your team knows when the expected output did not arrive.