Listen to Production the Way It Deserves

For years, the way we monitor and alert on production systems has remained reactive. We define assumptions, set thresholds, and wait for alerts to fire, often too late, too early or too often.As systems grow, this model creates noise, operational toil, and burnout, while still failing to capture what really matters in production.

In this talk, we will explore how agentic AI can change the way we approach software reliability. Instead of relying solely on predefined rules and static alerts, AI agents can continuously observe production signals, reason about system behavior, and help engineers detect and understand issues earlier.

Drawing from real-world SRE experience, we will look at the current bottlenecks of observability and alerting, and how AI agents can assist in identifying problems, surfacing weak signals, and supporting incident analysis. We will also discuss the strengths and limitations of AI agents in reliability systems: where they add real value, where they fall short, and why human expertise remains essential.

Finally, we will explore what the future of reliability could look like when engineers shift from reacting to alerts to truly listening to production