The layers of intelligence

A vertical AI system needs more than a model. It needs context. For enterprise AI, the quality of the model matters less than the quality of the context it can access. A useful company corpus should therefore not be treated as a document repository.

The first layer in our model is the structured operational layer: the numerical reality of the company, including products, suppliers, customers, orders, inventory, assets, locations and transactions. This is where the knowledge graph becomes essential, because it connects those entities and allows AI to understand not only individual facts, but their relationships. The second is company context: the organisation’s own policies, contracts, procedures, product knowledge, supplier agreements and accumulated human expertise. Our third layer is industry context: the terminology, structures, regulations, processes and decision logic that define the sector in which the company operates.

These three layers make the corpus useful as an intelligence asset. Industry context gives the system a frame of reference. Company context makes it specific. Structured operational data grounds it in what is actually happening.¹

Our approach is built around three connected layers: industry context, company context and structured operational data. Together, these layers create the foundation for company-specific AI that can support real business questions, not just generic conversations.²,³

1. The foundation layer

Highly structured operational data.

This is where the numbers, entities and relationships of the company are organized. Products, suppliers, materials, customers, locations, flows, inventories, service levels, risks, KPIs, costs and constraints are connected through our proprietary graph approach. This layer allows the system to move beyond isolated documents and start understanding relationships, dependencies and consequences.

2. Company context

Data, language and logic of the company.
This is where the system starts to understand how a specific organization works. Contracts, SLAs, process descriptions, strategy documents, supplier files, customer agreements, planning procedures, policies, reports, playbooks, meeting documents and internal knowledge are brought together into a healthy company corpus. This is not just document storage. It is the creation of a usable intelligence base for the company.

3. Industry context

Enrichment of the corpus.
This is the part we bring. We pre-build and continuously develop domain context for supply chain, operations and industrial environments. This includes industry language, typical processes, planning logic, common risks, supply chain structures, decision frameworks, relevant standards, business questions and analytical patterns. It gives the system a strong starting point before it ever touches company-specific information.

4. The company corpus

These three layers make the corpus useful as an intelligence asset. Industry context gives the system a frame of reference. Company context makes it specific. Structured operational data grounds it in what is actually happening. Research into knowledge graphs similarly emphasises the importance of schema, identity and context when integrating heterogeneous information, while retrieval-augmented AI research shows the value of giving language models access to explicit, updateable sources of knowledge rather than relying solely on what is stored inside the model itself.⁴

For organisations building decision-critical AI, this also has a governance dimension. NIST explicitly places organisational context, data selection, documentation, human oversight and knowledge of system limitations within the management of trustworthy AI. A well-governed corpus therefore becomes more than input for an AI model: it becomes part of the organisation’s own intelligence infrastructure.