Scaling AI agents in production without strict observability is a recipe for runaway costs, unpredictable behaviour, and compliance nightmares. Moving from AI experimentation to a governed, production-grade enterprise platform requires a robust foundation. In this session, we will explore the essential observability journey for standardising and governing AI agents, even in highly regulated environments. We’ll discuss how to leverage tools like Datadog’s LLM Observability from day one to track token usage, monitor agent behaviour, enforce guardrails, and debug production issues in real time. You’ll walk away with a practical framework to build, monitor, and scale AI agent programs safely, securely, and cost-effectively, drawing on proven industry patterns that manage dozens of live applications and drive millions in value.
Sagar Utekar is a Site Reliability Engineer at CrowdStrike, Docker Captain, CNCF Ambassador, and Kubernetes community leader with deep expertise in cloud infrastructure, observability, platform engineering, and DevOps. A former VMware SRE, he is an active open source contributor, CNCF project maintainer, educator, and content creator who helps engineers build practical skills across Kubernetes, cloud, and reliability engineering.
