Most serious AI work starts as a pile: systems that do not talk, decisions that live in people, data that is almost good enough, and a demonstration that looked convincing on Tuesday. Complexity is not the volume of information. It is the absence of a course of action that the organisation can own.
A chatbot on top of that pile does not resolve it. Neither does a strategy slide that names every possible use of a model. The work is to find the bounded problem, the interfaces it already has, and the failure modes that would make a system unsafe to run.
Linear Horizon starts there. Consulting finds whether there is a problem worth building for. Platform is the runtime when the work has a product shape. Build is for the systems that still have to be designed around the estate you already run. The outcome is not a fluent answer. It is a path from the current mess to something that can be operated.