Elastic’s InfoSec team has significantly enhanced their Security Operations Center (SOC) efficiency by implementing an agentic AI pipeline that improved alert triage accuracy from 60% to 92%. Instead of relying solely on raw alert data, the team utilizes Elastic Workflows and Agent Builder to provide AI agents with rich context, including historical case data, internal investigation guides, and user risk information from Workday. This approach transforms AI summaries from untrusted "slop" into actionable intelligence that analysts can verify and act upon with high confidence.
The architecture employs three specialized agents—Pattern Finder, L1 Investigation, and Summarizer—working in a synchronous pipeline to reduce token costs and increase speed. The Pattern Finder identifies historical trends and analyst feedback loops, the L1 Investigator performs targeted external queries to reach a verdict, and the Summarizer formats the final report for Slack and Kibana. This automated ecosystem enables analysts to close cases rapidly, often with a single click, by integrating directly into existing communication channels and providing a transparent feedback loop for continuous model improvement.