Enabling Physical AI: Connecting Edge Devices and Virtual Co-Workers in a Microsoft Architecture

Microsoft customer and partner engagement product manager Jason Farmer explains what it takes to enable physical AI, which requires connecting physical things on the edge. He walks through a reference architecture from an architectural perspective and outlines the benefits of putting such an infrastructure in place, covering both cloud-connected and on-site connected devices.

■ Architecture Overview
- Cloud-connected devices connecting directly to Azure IoT Hub vs. on-site devices communicating over a proximal protocol
- Edge infrastructure running on Azure Local or an Arc-enabled server, hosting Arc-enabled Kubernetes
- Azure IoT Operations for on-prem data and Foundry Local for running AI workloads on-prem
- Both data streams reaching Azure Resource Manager and Microsoft Fabric, then exposed to services such as Teams, Dynamics, and Work IQ

■ Virtual Co-Workers
- An OT virtual co-worker agent running in Foundry Local that keeps functioning on-prem if the cloud connection is lost
- An enterprise virtual co-worker running in Microsoft Foundry or Fabric, communicating through solutions like Dynamics and Teams

■ Closed-Loop Operations & Demo
- How virtual co-workers link agentic solutions down to edge devices
- A link to a demo for seeing the setup in action

IT/OT professionals, digital operations teams, and anyone involved in industrial edge deployments will come away with a clear map of how physical AI is structured end to end, and with a reference point to explore the architecture further through the provided demo.