AI for Observability:

What Actually Works in Production

AI is becoming an integral part of modern observability workflows, especially when it comes to investigating incidents and performing RCA. But a critical question remains:
what actually works in real production environments, and what breaks down under complexity and scale?

In this session, Tomer and Shevach from Logz.io will examine how AI is being embedded inside observability platforms, as an agent that can reason over logs, metrics, and traces to support production-grade RCA.

They’ll separate hype from reality and look at the practical constraints of using AI in live systems, including where human oversight is still essential.

Key discussion points:

• The core observability challenges engineers face in real production environments
• Why AI is positioned to address these challenges when it operates inside observability platforms
• Public LLMs vs observability-optimized AI: why simply connecting an LLM to logs or dashboards works only for basic cases
• The reality of AI agents in production: current capabilities and unresolved challenges
• What “production-grade” AI reasoning over telemetry actually requires
• Real examples of AI agents supporting RCA during live incidents