Innovation Week Day 2: Observability for AI, and Observability With AI

AI is significantly transforming the Software Development Life Cycle (SDLC) by enabling faster code shipping with reduced human oversight and introducing non-deterministic behaviors in production environments through agents and large language models (LLMs). This shift has led to a compression of the traditional multi-stage development process into rapid cycles of intent and validation, often occurring in live settings rather than pre-production stages. Honeycomb addresses these changes with two key approaches: observability for AI, which enhances visibility into agent and LLM workflows in production, and observability with AI, which utilizes advanced tools to tackle complex problems quickly. The introduction of Agent Timeline provides a comprehensive view of agent workflows, while deeper LLM insights incorporate prompts and quality scores as essential telemetry. The Canvas tool facilitates collaborative and automated investigations when alerts are triggered, incorporating team knowledge through custom skills to streamline problem-solving. These innovations reflect a broader trend towards integrating AI capabilities into observability practices, allowing teams to respond more effectively to the unique challenges posed by AI-driven software environments.