Greenlight Financial has deployed Luna, a tiered multi-agent system built on Amazon Bedrock AgentCore, to handle security and risk operations. According to the post, Luna resolves more than 90% of up to 4,000 monthly security and risk cases without human intervention. The write-up covers the system’s architecture, its use of episodic memory for continuous learning, and its expansion from security operations center (SOC) tasks into financial risk triage.
Why it matters
Autonomously resolving the majority of a high volume of cases points to a practical application of multi-agent systems in security operations. The use of episodic memory for continuous learning and the move beyond SOC work into risk triage suggest the approach is being extended to additional operational areas.
Who should care
Teams working on security operations, financial risk, and enterprise agent deployments may find the described architecture and memory design relevant to their own work.