Executives require a high-level report on how often agents run into errors, how quickly most...

Comprehensive and Detailed Explanation:

For executive summaries,Median Correctnessprovides a measure of agent response accuracy, whileError Rate and Median/P99 Latencyoffer direct insights into agent reliability and responsiveness. These are core metrics for tracking agent quality and operational performance over time.

Oracle documentation explains:

“Median Correctness, Error Rate, and Latency metrics are presented in dashboard visualizations, providing business leaders with quick, actionable summaries of AI agent effectiveness and user experience.”

(Reference: Oracle Fusion AI Agent Studio Executive Reporting, 2024, p. 17)

Prompt injection and LLM requests track security or operational events, not overall performance quality.

Practical takeaway:Use correctness, error, and latency metrics for executive reporting.