Recorded live at the MLOps World | GenAI Summit 2025 — Austin, TX (October 9, 2025) Session Title: Agent Drift: Understanding and Managing AI Agent Performance Degradation in Production Speaker: Kumaran Ponnambalam, Principal AI Engineer, Cisco Talk Track: Agents in Production Abstract: As AI agents become core to production systems, ensuring consistent long-term performance has become one of the most pressing challenges in applied AI. In this session, Kumaran Ponnambalam, Principal AI Engineer at Cisco, introduces the concept of Agent Drift — the gradual performance degradation that occurs as agents encounter evolving data, shifting user behavior, changing tools, or updated model parameters. Kumaran breaks down the root causes behind agent drift and explains how it affects reliability, accuracy, and overall system trust. Attendees will learn how to measure, monitor, and mitigate agent drift through proactive observability, performance tracking, and adaptive retraining strategies. This session provides both the theoretical understanding and the practical toolkit needed to maintain high-performing, resilient AI agents in dynamic, real-world environments. What you’ll learn: • What “Agent Drift” is and why it matters in production systems • How to detect early signs of drift in data, behavior, and model interactions • Practical frameworks for measuring and tracking AI agent performance • Strategies for mitigating drift through retraining, calibration, and adaptive logic • How to build reliable, self-correcting agent systems at scale.