Choose the right record
Do not treat a Task being queued, a function being requested, or an Agent saying “done” as proof
that the intended outcome occurred.
Investigate one interaction
- Open Logs at the Agent or organization level.
- Narrow the time window and select the columns that matter.
- Open the matching log and review its summary, details, transcript, reporting fields, and channel information.
- Open the complete conversation when the summary does not explain the behavior.
- If a Task started the work, inspect its attempt or run history as well.
- Verify external actions in the destination system.
Investigate a pattern
Export or filter a bounded time range rather than relying on one anecdote. Compare successful and unsuccessful interactions under the same Agent version, channel, audience, and workflow. Useful questions include:- Did the Agent receive the required information?
- Did it select the expected function, app, helper, or transfer path?
- Did the external system accept and persist the action?
- Was the reported outcome supported by the conversation?
- Did the problem begin after a publish, integration change, or routing change?
Build a useful support report
Include:- the affected Agent and account
- exact timestamps and timezone
- Task, log, and conversation references
- expected behavior and observed behavior
- customer impact
- evidence from the destination system
- changes already attempted
Protect sensitive information
Conversations and logs can contain customer data. Export only what is needed, share it with the appropriate people, and remove downloaded files when the investigation is complete.Investigating programmatically? Use the
incident workflow.