AI Observability Must Evolve for the Agentic Era
Check Point, Wednesday, September 2nd, 2026
Check Point argues agent observability must capture decision traces and be able to stop unsafe actions before they execute.
Steve Giguere of Check Point's Office of the CTO argues that traditional observability answers whether software worked, which is the wrong question for AI agents.
The harder question is whether the system made the right decision, and the telemetry that answers it must also be capable of stopping unsafe actions before they execute.
He proposes agents need a full decision trace covering goal, context, plan, tools, credentials, actions and outcomes. The piece frames observability for agentic systems as an enforcement point rather than a purely passive monitoring layer.