Skip to content
PraxivaraPraxivaraHelp Center
AI employees Work, runs and teamwork AI employee's job failed

Why did my AI employee's job fail?

Understand why an AI employee's job ended with a problem, where to look in Activity and Needs you, and the quickest ways to get it working again.

Updated Nov 16, 2026 at 12:54 AM ET 3 min read Ask AI about this article

When an AI employee's job doesn't finish the way you expected, Praxivara shows you what happened. Most problems come down to a few common causes, and nearly all of them are quick to fix.

Start in Needs you on your AI employees page, where problems show up. Then open the employee and go to its Activity tab. Activity lists every run, what happened, and any problems. It replaces the old Errors tab.

Common reasons a job fails

  • A disconnected app. If the employee needs Gmail, Slack, a CRM, or another app that lost its connection, that step can't run. This is the most frequent cause.
  • Not enough credits. Runs stop when your workspace balance reaches zero.
  • A missing file or detail. If a skill expects a key, password, or reference file that isn't set up, the step can't complete.
  • A trigger fired with unexpected data. An incoming email or webhook that doesn't match what the employee was set up to handle can send it off course.
  • An outside service was down or slow. A connected app timed out or returned an error of its own.
  • It is paused. When the Working switch at the top of the employee's page is off, it does no unattended work. You can still chat with it.
  • It is archived. An archived employee does no work. On Your AI employees, open the Archived tab and choose Restore.
  • It isn't set to work then. Open its Setup tab and check the section named after it, such as When Maya works, for its schedules and triggers.

Not failed, just waiting for you

Sometimes a job hasn't failed at all. It is waiting for your OK or an answer. You choose what waits for your OK. Anything waiting shows above the chat box in the employee's chat, and approvals and questions also reach you on your connected text channels.

If two linked employees go back and forth without progress, the conversation pauses and lands in Needs you, for example "Leo and Maya are going in circles". You can choose Read the thread, Let them carry on, or End the job.

How to find the cause

  1. Open the problem. Select it in Needs you, or open the employee and go to Activity. You'll see what happened and where it stopped.
  2. Read the run in Activity. The last part that worked tells you where things broke.
  3. Fix the root cause. Reconnect the app, top up credits, or add the missing file or detail, depending on what Activity shows.
  4. Try again to confirm. Ask the employee in its chat to try again, or wait for its next scheduled or triggered run, and check that the result lands in Work. To check without sending or changing anything, choose Do a practice run from the More button (three dots) at the top of its page.

Tip. If a job partly finished, check Work and Activity before trying again so the employee doesn't repeat something it already did.

When the setup itself is the problem

If a job fails the same way every time, the issue may be in how the employee is set up rather than a one-off glitch. Tell the employee in its chat what went wrong, in plain words. It can update its job, when it works, the apps it uses, or its skills. Each change shows in the chat as a change note with See the change and Undo. You can also edit its Setup tab directly. Earlier versions are listed under History in Setup, where you can choose Restore this version.

Note. If a scheduled job keeps failing, turn off the employee's Working switch while you look into it. That stops unattended work, and you can still chat with it.

Next steps

Was this helpful?

Related articles