Many organizations can measure how often an agent runs. Far fewer can confidently measure whether it is succeeding. As agentic systems take on more complex tasks, defining success becomes increasingly difficult. Teams must balance technical metrics, user outcomes, operational reliability, and business impact while accounting for behaviour that can vary across different situations and environments. This panel examines how organizations are approaching measurement in agentic AI, the signals they trust, the metrics they ignore, and the frameworks emerging to evaluate performance beyond uptime and cost alone. The conversation will explore what meaningful success looks like and how teams are building the visibility needed to measure it consistently.