Question

Your team runs a multi-agent customer-support application in Azure AI Foundry. Finance reports that monthly token spend has doubled, but overall request volume is unchanged. You need to identify which specific agent operations and prompts are consuming the most tokens so you can optimize them, without disrupting production traffic. Which approach in Foundry observability should you use first?

Accepted Answer

Enable tracing and inspect span-level attributes to correlate per-operation prompt/completion token counts, then aggregate cost by operation in the monitoring dashboards

Answer

Switch the deployment from serverless (pay-as-you-go) billing to provisioned throughput units to cap monthly spend

Answer

Configure a groundedness and relevance evaluation run on production samples to reduce unnecessary model calls

Answer

Reduce the model's maximum context window in the deployment settings to force lower token usage globally