Cornerstone OnDemand says Orion AI cut database diagnosis time by 78%

AWS published a Cornerstone OnDemand case study about Orion AI on October 7, 2026. The multi-agent system, built with Amazon Bedrock and Strands Agents, was delivered by a three-person team in six months. Cornerstone says database diagnosis time fell from 45 minutes to 10, a 78% reduction.

The post also reports that database lifecycle work went from more than 10 manual steps to a single interaction, a 70% reduction, while the reporting lag from the SRE team to the data team changed from 15 minutes to instantaneous. Orion AI reduced redundant alerts by a median of 65%; the article says about three or four of every 10 previous alerts now reach engineers.

Orion AI uses a hub-and-spoke design, with a meta-orchestrator assigning work to specialized agents. It runs as containerized services on Amazon ECS and uses 13 domain-specific agents across infrastructure monitoring, database diagnostics and lifecycle operations, customer analytics, and knowledge-driven support. For remediation, it can recommend a fix and create a populated Jira ticket for the appropriate on-call engineer. Destructive operations require user confirmation within five minutes; without it, the system defaults to deny.

AWS has previously described a cloud-migration multi-agent framework built on Amazon Bedrock AgentCore; this case focuses on database operations. Cornerstone says it chose ECS because AgentCore runtime was unavailable when the project began and is evaluating it as a possible future migration option.