Question

A financial services company deploys a multi-agent solution in Azure AI Foundry that helps advisors answer client questions. Compliance requires that, for every response the agent produces, auditors can later reconstruct exactly which tools were invoked, which knowledge sources were retrieved, and the sequence of model calls that led to each answer. Which Foundry capability should the team enable to meet this requirement?

Accepted Answer

Enable tracing with OpenTelemetry integration so that end-to-end spans capturing tool calls, retrieval steps, and model invocations are logged for each agent run

Answer

Configure Azure AI Content Safety filters with severity thresholds on all agent inputs and outputs

Answer

Turn on token usage metrics and configure cost alerts in Azure Cost Management for the deployment

Answer

Assign the auditors the Azure AI Developer role so they can inspect the agent configuration in the portal