How Ontology Shapes High‑End ERP Architecture: Unified Semantics and Business Object Hierarchies

Ontology has become a buzzword in China's AI era, prompting many superficial articles; this piece conducts a deep analysis of high‑end ERP architecture, focusing on how ontology provides unified semantics and business object hierarchies.

Business vs Technical Architecture

While business architecture (organization, capabilities, processes) is essential, the author argues that implementation hinges on technical, data, application, security, and ontology layers. The complexity of high‑end ERP lies in balancing these layers.

What Is Ontology?

In the semantic‑web context, ontology defines classes, properties, constraints, and inference rules to enable machine‑readable knowledge sharing. Standards such as OWL (Web Ontology Language) and RDF Schema are cited as typical representations.

Ontology in Knowledge Graphs

Within knowledge‑graph practice, “ontology” refers to a practical, scalable schema that organizes entities rather than a logically complete model. It supports use cases such as search engines, recommendation systems, and enterprise knowledge management by providing a type system and attribute templates.

Static Ontology vs Dynamic Entities

Ontologies are relatively stable—once defined they rarely change unless the domain knowledge undergoes a fundamental shift—whereas entities continuously evolve with time and context to reflect the real world.

Palantir’s Ontology Model

Key concepts include:

Object : business entities like Vehicle, Order, Person, each with a unique identifier and lifecycle states.

Property : attributes classified as static (VIN), dynamic (fuel level), or derived (order total = price × quantity − discount), supporting temporal validity.

Link : relationships between objects (assignedTo, contains) that may carry their own attributes and cardinality constraints.

Action : executable operations on objects (AssignDriver, Refuel, CancelOrder) that can modify properties, create/delete links, invoke external APIs, and trigger workflows.

Function : derived logic for non‑stored attributes, allowing nesting and reuse.

Time Window : validity periods attached to objects or links to query the world state at a specific moment.

Permissions : fine‑grained, object‑level access control based on roles, users, or dynamic conditions.

SAP’s Semantic Unification

SAP’s “clean core” philosophy aligns business objects across OLTP and OLAP layers, effectively implementing an ontology‑like unified semantic layer, though SAP does not label it as ontology.

The author concludes that merely adding ontology layers does not solve the fundamental challenges of Chinese ERP systems; the real difficulty remains the stability of core architecture and the gap with leading vendors such as SAP and Palantir.