Agents
A plant with 20+ vendors now faces 20+ AI agents, each with its own access model
At the Industrial AI Summit 2026, panellists including GrayMatter's Scott Christensen and Skkynet's Gary Tillery addressed a new governance problem: a plant running 20 or more vendors faces each one deploying its own AI agent, with its own access model and integrations.
Put agent access in the contract before the next vendor ships one into OT: what it can touch, through which integration, and who revokes it.
Agents
Cybus CEO on structuring factory context for AI agents — and where MCP servers fit
In follow-up answers from the Industrial AI Summit 2026, Cybus CEO Peter Sorowka covered how factories should structure context for AI agents and the role of MCP servers in connecting LLMs to plant data. A companion piece takes on the architecture: how to connect an LLM to a factory.
An agent is only as good as the plant context it can reach — budget for structuring that data before you buy the agent.
Frontier
Spatial and agentic AI come with a precondition: data and governance before scale
Automation World argues that spatial and agentic AI are changing industrial automation, but that manufacturers must build data and governance foundations before scaling deployment.
If your agent pilot has no data owner and no governance sign-off, it is not ready for a second site.