LLM Observability at Scale: Governing, Monitoring, and Securing AI Agents in Production

Scaling AI agents without observability introduces cost risk, blind spots in behavior, and increased complexity in controlling data access and actions. Rodrigo Moreno and Willians Aguiar share their observability maturity journey and how that foundation enabled them to build Titan, an internal platform to standardize and govern AI agents in a regulated financial environment. They embedded Datadog’s LLM Observability from the start to monitor behavior, track token usage, enforce guardrails, and debug production issues. With more than 30 agent-powered applications in production and over $11 million in value captured in 2025, they’ll share what it takes to move from experimentation to a governed, production-grade AI platform. Learn practical framework to build and operate AI agent programs at scale, covering how to monitor agent behavior and costs while enforcing safety and compliance guardrails.