When autonomous agents take action at scale, failures are inevitable. A misinterpreted intent, an incorrect refund, or a missed escalation can quickly turn into a customer-facing issue with operational and reputational impact. This discussion focuses on how organisations are structuring accountability when AI systems make decisions, how incident response processes are evolving, and what happens when customers need answers fast. Leaders share how they investigate failures, communicate under pressure, and rebuild trust after AI-driven errors, including the lessons many teams only learn after their first major incident.