Large Enterprise Requirements to Scale Agentic AI
Agentic AI isn’t being held back by model capability.
It’s being held back by a lack of control.
Enterprises are deploying autonomous agents into environments that were never designed to govern them.
ONUG’s Agentic AI Overlay Working Group, led by IT executives from Global 2000 enterprises including eBay, Cigna, and Bank of America, produced a series of two papers to address large enterprise requirements for scaling agentic AI.
The Agentic AI Overlay introduces a fundamentally different approach. Instead of embedding trust inside individual models or agents, it establishes an independent control layer that governs how agents communicate, access data, and act across cloud, data center, edge, and external environments.
Part A: Agentic AI Overlay Architecture
The blueprint for enterprise‑grade Agentic AI.
A vendor-neutral reference architecture for agent-to-agent communication, identity and policy enforcement, and zero-trust foundations.
Part B: Scaling, Deployment, and Operationalization Strategies
The path to deploying and scaling agentic AI in production.
Guidance for cross-domain deployment, runtime governance, continuous monitoring, and failure handling at scale.