Reference Architecture for Agentic AI

The agentic AI landscape is facing an architectural crisis due to framework fragmentation, bottlenecking the creation of truly reliable autonomous systems. 

This session pivots from individual agent design to the critical challenge of systems architecture and coordination, presenting the essential blueprint and reference architecture for agentic AI – apps that infuse AI and continuously build trust. We will detail how agentic systems are designed and built and how a future global control plane enables disparate, multi-framework agents—operating across diverse networks, clouds, and regulatory regimes—to be dynamically orchestrated and reliably composed into collaborative teams. 

The session will dive into the core engineering hurdles, including cross-framework durability protocols, state management in a distributed world, and providing the necessary compliance and audit trails, offering a vendor-neutral, blueprint-level look at the architecture required for enterprise-grade Agentic AI.