
How Anthropic's Model Hardware Standard Enables Agentic and Physical AI in Manufacturing
What does Anthropic's Model Hardware Standard (MHS) mean for the future of agentic and physical AI in manufacturing?
In this episode of AI in Manufacturing, Kudzai Manditereza speaks with Jan Wokittel, Director of Smart Manufacturing at Roche, about how AI agents could transform the way manufacturers interact with industrial equipment, develop automation applications, and build future-ready manufacturing systems.
The conversation explores Anthropic's Model Hardware Standard (MHS), a proposed approach to enabling AI agents to understand and interact with physical hardware through standardized interfaces. Jan explains why this could significantly reduce the complexity of industrial integration, accelerate innovation cycles, and give engineers, scientists, and business users greater freedom to develop applications without relying on proprietary vendor ecosystems.
They examine how MHS compares with established industrial technologies, including OPC UA, Module Type Package (MTP), and Asset Administration Shell (AAS), and discuss whether these standards will compete, coexist, or evolve to support AI-driven manufacturing. The discussion also explores the role of MQTT and the Unified Namespace (UNS) as a real-time data and event backbone through which AI agents can consume operational information, respond to production events, and coordinate activities across industrial systems.
Drawing on his experience at Roche, Jan shares how manufacturers can build modular, interoperable technology architectures that accommodate emerging AI capabilities while maintaining reliability, regulatory compliance, and long-term flexibility. He also discusses the implications of AI for industrial automation vendors, equipment manufacturers, and the future of software development in manufacturing.
The episode concludes with a discussion of the opportunities and challenges of deploying AI agents in regulated manufacturing environments, including validation, governance, and the potential shift from human-in-the-loop approval toward human-on-the-loop oversight.
Key topics include:
- Anthropic's Model Hardware Standard (MHS) and its relationship to MCP
- Enabling AI agents to understand, access, and interact with industrial equipment
- MHS versus OPC UA, MTP, and Asset Administration Shell (AAS)
- Semantic interoperability and machine-readable equipment information
- MQTT and Unified Namespace architectures for agentic AI
- Open standards, vendor independence, and the future of industrial automation
- Building modular, AI-ready manufacturing architectures
- AI validation, governance, and human oversight in regulated industries
About the guest:
Jan Wokittel is the Director of Smart Manufacturing at Roche, where he works on technology strategies and architectures supporting the next generation of manufacturing facilities. His work focuses on industrial interoperability, digitalization, emerging technologies, and building scalable foundations for smart manufacturing.
About the podcast:
AI in Manufacturing by Industry40.tv explores how artificial intelligence is transforming manufacturing operations, industrial software, automation, and engineering. Hosted by Kudzai Manditereza, the podcast features conversations with industry leaders, researchers, technology innovators, and manufacturing practitioners about the technologies, architectures, and strategies shaping the future of industrial AI.
Learn more at https://www.industry40.tv
More episodes from "AI in Manufacturing"



Don't miss an episode of “AI in Manufacturing” and subscribe to it in the GetPodcast app.








