I Taught an AI to Be Our On-Call Engineer (medium.com)

A new AI tool named Scooby has been developed to assist engineers with on-call incidents by automating root cause analysis. This plugin, built on the Claude Code framework, connects to various internal data sources such as Grafana, Loki, and Confluence, allowing it to analyze alerts without the traditional manual effort of navigating multiple dashboards. The creator, inspired by the limitations of existing tools like HolmesGPT, focused on teaching the AI rather than writing extensive playbooks. By using historical incident data and allowing the AI to self-improve its skills, Scooby has been trained to handle specific company contexts, accommodating unique naming conventions and logging standards. The significance of Scooby lies in its ability to drastically reduce the time engineers spend diagnosing issues during incidents, transforming a once cumbersome process into a quick and efficient investigation. By integrating it with Slack, engineers can invoke Scooby for real-time assistance without switching tools, leading to quicker resolutions and preventing prolonged outages. This innovative approach not only enhances operational efficiency but also builds a knowledge base that evolves and matures with each incident, ultimately increasing the AI's effectiveness in future investigations.