One of the recurring frustrations customers face is that their systems don't talk to each other well. Data sits in silos. Integrations are expensive to build and even more expensive to keep alive through upgrades. A project that works well at one site turns into a rebuild at the next one. None of this is new, and the industry has never really solved it in a systematic way.
What has changed is the cost of leaving it unsolved. Most of the IT and operations leaders I talk to are under pressure to put AI to work in production this year. AI agents and assistants are only as good as the context they are given, and context is exactly what a fragmented architecture destroys. When the relationship between an asset, a line, a work order and a quality result lives in four systems with four naming conventions, the agent returns a confident answer that happens to be wrong. That is where a lot of AI pilots stall, and it is rarely a model problem. It is a data context problem.
That is a large part of why we have been working with CESMII, the Smart Manufacturing Institute, and why we support what they are building around interoperability.