Everyone is talking about AI transforming industries. In rail, I think we're missing the real problem.

It's not AI. It's data. More specifically, it's labeling.

Rail is one of the most complex, fragmented networks in the world, thousands of facilities, services, capabilities, and constraints, but no consistent way to describe them. No shared language. No structure that systems can reliably read. A transload at one site can mean something very different at another; storage might include switching at one facility and exclude it at another. Capabilities, limitations, and operational nuances are often implied rather than defined. Most of that detail lives in people's heads, not in systems.

So we end up trying to layer new technology on top of inconsistent, human-described data, and then wonder why accuracy and efficiency fall short. This isn't a tooling problem. It's a structure problem.

Why Structure Comes First

Before anything can be optimized, automated, or intelligently matched, it has to be clearly defined. If the underlying data isn't consistent, systems can't reliably interpret what a facility actually does; matching supply and demand becomes inconsistent; and visibility across the network stays fragmented.

You can build better software. You can apply better models. But without a shared structure, you're still working from slightly different versions of reality, and that limits everything.

A Different Approach: Building a Shared Language

What we're working on is a public ontology for the rail industry: a structured framework for how rail services, facilities, and capabilities are labeled and understood. Not as a layer on top, but as a foundation. The goal is simple: create a system where both humans and machines can interpret what actually exists across the network, consistently. The challenge is making that structure standardized enough to be useful while remaining flexible enough to reflect real-world operations.

Who Defines the Data Matters

The people closest to the operations understand the nuance. They know what their facility can and cannot do, the edge cases, the details that don't show up in a generic description. But instead of leaving that knowledge unstructured, the system provides a defined framework for how it is expressed. Operators don't invent their own labels. They select how their operations are represented within a shared structure. That balance keeps the data consistent across the network while still allowing each operation to be accurately described.

What Changes When the Structure Exists

Once that layer is in place, things start to shift. Information becomes interpretable instead of implied. Matching becomes more precise because services and capabilities are defined the same way across locations. Visibility improves because the network is described in a consistent language. And over time, the system gets stronger as more participants contribute to that shared structure.

This Isn't About AI

AI will play a role in the future of rail, but it won't be the starting point. Without structure, systems are forced to guess; with structure, they can actually understand. The difference between those two states is what determines whether technology improves the industry or just adds another layer of complexity.

If there's a real unlock in rail, it starts here, with a shared language and the structure to support it.