ODSP2164 Pre-recorded On demand
Know when AI should yield to human judgment
Gene Moody
Summary
From Microsoft's AI summary
Gene Moody of Action1 discussed where AI should yield to human judgement. Action1 survey data showed 58% of system administrators in 2026 understood how to integrate AI, against 24% in 2024, and 16-20% reported AI-driven patch prioritisation or maintenance. Trust varied by context: 48% trusted AI to patch workstations and 17% production servers. Moody covered least privilege, reversibility and blast radius, human-defined revocable authority boundaries for incident response, five principles for safe autonomy, and decision traceability beyond simple logging.
Session description
Microsoft's catalogue description
AI has become remarkably capable at analysis, correlation, recommendation, and prediction. Yet the closer AI gets to decisions with real operational consequences, the more administrators insist on human authority in the final dispatch. Using 2026 survey data from more than 1,000 sysadmins, this session examines where that boundary exists, why it persists, and what it tells us about the future of autonomous IT.