Instacart has launched Blueberry, an AI-driven system designed to accelerate incident response for on-call engineers. The tool integrates AI agents with operational data and historical incident records to formulate grounded root cause hypotheses directly within Slack. By leveraging parallel subagents and MCP integrations, Blueberry aims to reduce investigation latency while ensuring engineers retain final control over the diagnostic process.
- AI agents generate root cause hypotheses using historical incident data and live operational metrics.
- Integration with Slack allows engineers to receive and act on insights without context switching.
- Parallel subagents and MCP tools enable faster data gathering during active incidents.
- The system prioritizes human-in-the-loop control to maintain reliability and trust.
- Goal is to reduce mean time to investigation (MTTI) for production issues.