Scale One AI Agent Across Sites (Part 2)
Christoph Berlin, Microsoft’s Vice President of Engineering for Cloud and AI, builds on Part 1’s root cause example to explain how manufacturers can deploy an agent across many sites without rebuilding it for each location.
A hybrid approach to agent deployment
Manufacturers do not have to choose a single deployment location for every AI agent. The video describes a hybrid model in which one agent framework supports execution in the cloud, at the edge, or in a fully disconnected environment.
The shared foundation consists of:
- Common data contracts across deployments.
- Common guardrails for the agent framework.
- Repeatable blueprints that support reuse across sites.
- A single management interface for the deployed agent instances.
The focus is on making the agent repeatable across hundreds of sites rather than building it over and over for each location. Berlin closes with two lessons for scaling agentic AI in manufacturing.
Video chapters
- 0:00: Why manufacturers cannot run every AI agent in the cloud
- 1:10: Whether AI agents should run in the cloud or at the edge
- 1:49: How a blueprint makes AI agents repeatable
- 2:25: How one AI agent runs across hundreds of sites
- 3:04: Two lessons for scaling agentic AI in manufacturing