Many AI systems in incident response are black boxes, requiring engineers to trust their conclusions without understanding the reasoning behind them. This can be problematic when an AI claims to have found a root cause without providing clear evidence. A glass-box AI for site reliability engineering (SRE) addresses this issue by making its reasoning transparent. It walks engineers through each step of its analysis, linking directly to relevant dashboards, logs, and documentation. This integration allows for immediate follow-up questions and actions within existing communication tools like Slack. The glass-box AI offers several key features. It provides a detailed investigation process, connecting incidents to their causes with evidence-backed timelines. Engineers can access specific data on demand, avoiding information overload during critical moments. The AI also adheres to established runbooks, ensuring its actions align with operational protocols while capturing any corrections to improve future performance. This type of system supports safe autonomy, requiring human approval for high-risk actions and maintaining an audit trail for accountability. Moreover, the AI is designed to seamlessly integrate with existing tools like Datadog and Jira, allowing engineers to work within familiar environments. Specializing in the unique terminology and practices of a given organization, it pulls relevant information accurately from the start. Its iterative planning process refines recommendations while tracking confidence levels for each action taken. This transparency and adaptability help reduce alert fatigue, improve mean time to resolution (MTTR), and ensure that valuable insights are shared and retained within the team.