AI Raises the Stakes for Observability Engineering

AI-assisted development has shifted the focus in software engineering from writing code to understanding how the resulting system operates in production. With agents and assistants capable of generating entire services in a short time, teams must now invest in validating changes, investigating incidents, and reducing risk to maintain a fast shipping pace.

This has elevated observability engineering to a central role in the development process, as highlighted by Liz Fong-Jones, a Technical Fellow at Honeycomb, in an interview with Alan Shimel. Fong-Jones distinguishes between telemetry, which refers to the raw data like logs, metrics, and traces, and observability, the broader capability to use this data, combined with human knowledge and processes, to understand the system and make informed decisions.

She views observability as an evolving quality of software, similar to testability or accessibility, and argues that teams never truly complete their observability efforts. The discussion delves into how AI transforms the observability loop, offering significant benefits through agent-assisted workflows that can generate instrumentation code, filter through large volumes of telemetry, and expedite investigations.

However, Fong-Jones emphasizes that AI does not replace human judgment but rather redirects it away from routine tasks to focus on interpretation and decision-making that only humans can perform. The second edition of the O'Reilly book she co-authored shows the shift from questioning whether observability is distinct from monitoring to understanding how it underpins AI-driven software delivery.

Cloud native architecture, characterized by the widespread use of containers and Kubernetes, has become the standard for modern software development, with OpenTelemetry serving as the foundational standard. Within this context, observability engineering is no longer a peripheral consideration but a critical discipline that enables teams to accelerate their delivery while maintaining control over the associated risks.

Organizations that treat observability as a core engineering function are better positioned to continue delivering value as other teams grapple with post-incident challenges.