Automating incident resolution with a multi-agent self-healing SRE solution on AWS

This article describes how Daimler Truck built a multi-agent self-healing SRE solution using Amazon Bedrock AgentCore, AWS DevOps Agent, and AWS Lambda to automate incident resolution.

• Automatically detects SRE errors, performs root cause analysis, and generates pull requests with proposed fixes
• Reduces incident resolution time by approximately 88%, from 2-3 hours to 15 minutes
• Includes human-in-the-loop review and approval before deployment for safety
• Uses safeguards including de-duplication, sandbox validation, temporal policies, and cost circuit breakers
• Frees engineering teams from reactive firefighting to focus on higher-value innovation work
• Plans to expand to additional incident types including application-level errors and predictive blocking

The solution demonstrates how to combine AI agents, monitoring, policy controls, and human oversight to automate repetitive incident triage while maintaining operational safety and reliability.