Everyone is building AI agents. Few are operating reliably at scale. The agents reason probabilistically, take autonomous actions, call APIs, spend money, coordinate with other agents. Traditional DevOps and MLOps were designed for deterministic systems. AI agents break that contract — they fail on a spectrum, not in binary, and their failures look like bad judgment, not crashes.

Join the session to learn the 8 tenets of AgentOps — the operational discipline for running AI agents reliably at scale on AWS and with Datadog AI Observability and Security solutions. We cover decision-path observability, agent identity and scoped permissions, continuous evaluation, reasoning-based incident response, token cost management, fleet governance, human-in-the-loop escalation, and self-healing automation. Attendees will leave with a practical AgentOps maturity framework.