Monitoring AI agent runs

Every time an AI agent is triggered for a ticket, a run record that tracks its full lifecycle is created. Each run record tracks the status, the number of tool calls it made, and, if it failed, the reason.

To review AI agent runs, navigate to Reports > AI Agent Runs.

What a run record shows

By clicking on an AI agent run record, you can see detailed information about the given run. Each record contains:

  • Run ID and its status,
  • details about the AI agent that was executed and the ticket that was processed,
  • the start & completion times and duration,
  • the number of tool calls made,
  • and, if applicable, the failure reason.

Run statuses

StatusMeaning
QueuedThe run has been created and is waiting to be picked up by the AI provider.
SubmittedThe run has been picked up and dispatched to the AI provider. The flow or model is executing and may be making tool calls.
WaitingThe run was paused because the AI provider is temporarily rate-limited. LiveAgent resumes it automatically once the throttle window expires.
CompletedThe AI agent made one of the tool calls that mark the run as completed: Resolve ticket, Transfer ticket, or Submit ticket summary. A run is also marked as completed if the safety-net timeout closes it after at least one tool call was made (see below).
FailedThe run did not complete — for example, it stayed in Submitted longer than 10 minutes without any tool call (see safety-net timeout below). Open the run to see the failure reason (see below).

Throttling: If your AI provider is temporarily rate-limited, LiveAgent automatically pauses the run and resumes it once the provider's throttle window expires — no action is needed on your part. If a paused run cannot resume within 60 minutes of starting, it is timed out.

Safety-net timeout: A run that stays in the Submitted state for more than 10 minutes without finishing is closed automatically. If it had already made at least one tool call, it is marked as completed; if it made none, it is marked as failed with the "Execution timed out" reason.

Possible failure reasons

When an AI agent run fails, its record's details show one of the following reasons:

ReasonWhat it meansWhat to do
Execution timed outThe run did not complete in time and made no tool calls before its time window elapsed — either it stayed Submitted for more than 10 minutes, or a throttled run could not resume within 60 minutes of starting.Check your flow or model configuration and your provider's availability. If throttling is frequent, reduce the number of rules that trigger AI agents simultaneously. The next matching rule trigger creates a new run automatically.
AI provider errorThe AI model returned an unexpected error during processing.Review your provider or flow configuration. If the error is intermittent, the next rule trigger will create a new run automatically.
Tool execution failedThe AI requested a tool call, but the action could not be completed (e.g., attempting to assign a ticket to an agent that no longer exists).Review the flow or agent logic and verify that referenced agents, departments, and tags still exist in LiveAgent.
Tenant rate limit exceededYour account reached its overall AI rate limit.Reduce the number of rules that trigger AI agents at the same time, or spread the workload out. The limit resets automatically over time.
Monthly AI budget exhaustedThe budget selected for this run reached its monthly tool-call cap before the run completed.Increase the budget's monthly limit, or simplify the flow to require fewer tool calls per ticket. See AI Budgets & Credits. Each budget's usage resets at the start of every calendar month.
Credit pool exhaustedYour account's AI credit balance was fully used up.See AI Budgets & Credits for how credits work and the free credits included with your account.
Ticket was deletedThe ticket was deleted from LiveAgent while the AI was still processing it.No action needed. This is expected behavior when tickets are removed during active processing.

×