At this year's AWS Summit, Ravical CTO Ken Bastiaensen took the stage to share practical lessons from deploying AI agents in production. While much of the industry's attention remains focused on model capabilities, the session explored a different challenge: what happens when agents move beyond prototypes and start operating at enterprise scale. Drawing on Ravical's experience running agents against real workloads, with real users and real spend, the presentation focused on the operational foundations required to build reliable agentic systems.
Ken Bastiaensen, Co-Founder and CTO Ravical
As organisations move from experimentation to deployment, AI agents introduce an entirely new set of engineering challenges. Security, observability and reliability can no longer be treated as afterthoughts. One of the key topics discussed was agentic security and the importance of system-level thinking. Ken highlighted emerging risks that arise when agents interact with tools, external systems and large amounts of organisational context, and why traditional software security approaches are no longer sufficient on their own.
The second half of the session focused on the realities of operating AI systems at scale. Topics included token pricing, cost optimisation strategies and how organisations can manage rapidly growing AI workloads efficiently. The talk concluded with a discussion on evaluation frameworks for agentic systems, including shadow runs, large-scale testing and monitoring approaches that help teams maintain quality and reliability as deployments grow. Together, these practices form the foundation of Ravical's unified agentic platform, helping professional services firms scale advisory workflows while maintaining oversight and control.
