We’ve enabled the OAC AI Agent and plan to roll it out to business users across multiple countries. As part of our audit and model-risk controls, we need to review interactions to identify potential false negatives and any hallucinated or inaccurate responses.
Specifically, we’re looking to understand:
- Conversation retention: Can OAC store AI Agent prompts and responses as a history for a defined retention period (e.g., 30/90/180 days)?
- Auditability: Can we see who asked what, including user identity (or a unique user identifier), timestamps.
- Access & export: Is there an admin console, logging feature, or API to retrieve these logs for audit review and evidence?
If this is supported, could you please share the recommended approach and configuration steps (e.g., which settings to enable, where logs are stored, and how retention is managed)? If it’s not supported natively, what’s the best-practice workaround (for example, routing prompts/responses to an external logging/SIEM solution)?
Regards,
Astha
