How to leverage Open Telemetry in an AI world

One of the challenges we see too many times is that not everyone is doing basic monitoring. At a certain point the AI wave arrived and now companies are deploying LLM-powered applications without any observability/security in place. How do you track token usage? How do you debug a hallucinating agent? How do you optimize costs when you don't know which model calls are eating your budget?
OpenTelemetry—the second most active CNCF project after Kubernetes—has an answer: the new GenAI Semantic Conventions. This talk explores how cloud-native teams can leverage OpenTelemetry to bring the same observability rigor we apply to traditional microservices into the world of AI agents, LLMs, and generative AI workloads.