As agentic AI begins to assume active roles within cyber defense and SOC workflows, many organizations are deploying these capabilities faster than they are testing or training them. This discussion will examine how agentic AI can/should be prepared, evaluated, and governed before being trusted with active roles in cyber defense—whether sourced from vendors, built internally, or deployed in hybrid models. Discussion will explore readiness criteria, evidence of safe operation, parallels to human training and certification, and the defensibility of AI-related decisions under post-incident or regulatory scrutiny. We will discuss whether approaches to agentic AI meet the same operational standards applied elsewhere in cyber defense.