Nobody Reads 40,000 Vulnerability Findings

Handing the backlog to a chain of AI agents — and knowing when to stop and ask a human. Detection was never the bottleneck. Triage, ownership, chasing and retesting are. This session walks through an agentic pipeline that takes a finding — whether it came from a cloud scanner, a SAST tool, an endpoint agent or a pentest report — all the way to a verified fix. Contextual severity scoring, one agent independently checking another's work, and the hard part: knowing when the automation is confident and when a human still has to decide.