Instacart Builds Blueberry, an AI-Powered Assistant to Help On-Call Engineers Investigate Incidents
Not a pilot. Instacart deployed AI agents across incident response—parallel subagents generating root-cause hypotheses in real time.
Why it matters
Autonomous agent systems are moving from research into production reliability engineering. Instacart's Blueberry shows how agents can handle high-stakes, multi-step investigative work while keeping humans in control—a template for enterprise agent deployment at scale.
The key facts
12 to knowInstacart's Blueberry uses parallel subagents for incident investigation
MCP (Model Context Protocol) integrations enable agent-to-system communication
System generates grounded root-cause hypotheses from operational data + incident history
Deployed in Slack for on-call engineer workflows
Reduces investigation time while maintaining human oversight
Published August 2026 on InfoQ
Instacart built Blueberry as a live incident response system for on-call engineers
System uses parallel subagents to parallelize investigation
Integrates MCP (Model Context Protocol) for operational data access
Leverages historical incident knowledge for grounded root cause hypotheses
Runs in Slack; keeps engineers in decision-making control
Targets reduction in incident investigation time
Go to the source
InfoQ AI/MLinfoq.com
Publisher excerpt: Instacart introduced Blueberry, an AI-assisted incident response system that helps on-call engineers investigate production issues faster. It combines AI agents, operational data, and historical incident knowledge to generate grounded root cause hypotheses in Slack. It uses parallel subagents, MCP…
