What Happens When a Run Fails
When a scheduled run encounters an error — such as an AI provider outage, a rate limit, or an agent configuration issue — TeamAI records the failure in the run history and attempts to recover automatically.
Here is what happens step by step:
The failure is recorded — the run appears as "Failed" in run history with an error message describing what went wrong.
A retry is scheduled — instead of waiting for the next regular cadence, the system schedules a retry with increasing wait times (backoff).
Notifications are sent — if you have notifications enabled, you will receive an email or TeamworkFX todo about the failure.
Retry Behavior and Backoff
When a run fails, the system doesn't immediately try again. Instead, it waits progressively longer between retries to avoid hammering a service that may be down:
Failure # | Wait before retry |
1st failure | 1 minute |
2nd failure | 5 minutes |
3rd failure | 15 minutes |
4th failure | Schedule is auto-paused |
If a run succeeds at any point during this sequence, the failure counter resets to zero and the schedule returns to its normal cadence. Failed runs don't count toward your monthly run limit.
Auto-Pause
A schedule is automatically paused when:
4 consecutive failures — the retry budget is exhausted, or
A permanent error occurs — such as the agent being deleted or the workspace losing access to a required resource. Permanent errors pause the schedule straight away, without retries.
When a schedule is auto-paused, you will see a warning banner at the top of the schedule detail page explaining the reason. Auto-paused schedules never resume on their own — you always resume them manually.
Auto-Pause Reasons
Reason | What it means | What to do |
Upstream provider errors | The AI provider (e.g. OpenAI or Anthropic) returned repeated errors such as rate limits or outages, so retries stopped. | Wait for the provider to recover, then resume the schedule manually. |
Repeated failures | The schedule failed multiple times in a row due to temporary issues. | Check the run history for error details. Fix any issues, then resume manually. |
Configuration error | A permanent problem was detected — for example, the agent was deleted or the workspace lost access to it. | Fix the underlying issue (choose another agent, restore access), then resume manually. |
How to Resume a Paused Schedule
Whether the schedule was paused manually or auto-paused by the system:
Go to the schedule detail page.
If auto-paused, review the warning banner and the recent run history to understand what went wrong.
Fix any underlying issues if needed (e.g. check that the agent still exists, or wait for a provider outage to resolve).
Optionally click Run Test to confirm the fix — the schedule stays paused during the test.
Click Resume.
After resuming, the failure counter resets and the schedule returns to its normal cadence.
Common Issues
My schedule didn't run at the exact time
Each schedule starts within a few minutes (up to 5) after its set time, on purpose, to spread load. If it's off by more than that, check the timezone setting on the schedule — a different timezone than you expected will make it run at a different local time.
My schedule ran but the output is empty or wrong
This usually means the prompt needs improvement. Review the completed conversation and compare the prompt against the guidance in Writing Effective Prompts for Scheduled Runs. Use Run Test to iterate on the prompt without using your monthly run limit.
I'm not receiving notification emails
Notification emails are sent from [email protected]. Check your spam/junk folder and ensure this address is whitelisted. Also verify that the correct recipient email addresses are entered in the schedule's notification settings.
My schedule shows "Paused" but I didn't pause it
The system auto-paused it after repeated failures or a permanent error. Check the warning banner and run history on the schedule detail page for the specific reason. See the Auto-Pause section above.
My schedule shows "Completed"
The schedule reached its end date and has no more runs. Edit it and extend or remove the end date to make it active again.
I can't create or resume a schedule because of a limit
You can have up to 20 active schedules per workspace. Pause or delete a schedule you no longer need to free a slot.
My schedules stopped running this month
You may have reached the limit of 100 successful scheduled runs this month. Runs are skipped until the limit resets at the start of next month (UTC); your schedules stay Active and pick up again automatically. You can still use Run Test in the meantime.
The Project or Client field is disabled
Only verified agents can be used with projects or clients. Pick a verified agent, or ask a workspace admin to verify the agent. See Scheduling with Projects and Clients.
"This client doesn't have an active project"
A schedule runs in its client's project, so the client needs an active project. Choose a different client, or ask your team to set up or unarchive the client's project.