The article looks at how agentic AI systems are assessed using specific metrics. It explores methods to measure their performance and behavior effectively.