Skip to main content

🧠 Ontology and AI

What is Ontology?​

An ontology turns data into understanding — mapping not just what your business knows, but what it means. It provides explanations and insights that go far beyond traditional reports and dashboards, enabling AI to reason about your business processes, configurations, and data relationships.

To be pragmatic, let us walk through how it works in practice — step by step.

Step 1 — Starting with a Real Business Question​

Let's start with a simple example.

A business user needs to understand some insights that sit behind a sales dashboard. They are asking a question for which there are no pre-configured cubes, dimensions, or reports — a question that would traditionally require a developer to build a new report or extract.

Ontology — Business User Question

The magic is operated here by the combination of Azure Fabric Ontology and AI. Instead of waiting for a new report, the AI assistant can reason over the ontology to find the answer — in real time.

Step 2 — How Ontology Works with AI@CDS​

But how does the ontology know enough to answer the question?

In our approach, we complete the SAP S/4HANA data flow to your data platform with your own business rules coming directly from your backend system. We primarily use configuration elements from SPRO (SAP customising), but this can be extended with additional elements — such as custom ABAP code for domain-specific logic like sales order BADIs, user exits, and enhancement spots.

Ontology — SAP SPRO to Ontology Attributes

In this illustration, we are extracting SD SPRO tables and mapping their values to Ontology attributes — for example, mapping the item category configuration to an attribute like BillingRelevance. The ontology now understands not just the raw data, but the business meaning encoded in the SAP configuration.

Step 3 — The Functional Reasoning Behind​

Now that the ontology knows your SAP configuration, it can provide the required knowledge for the Azure AI assistant to reason functionally about your business processes.

Ontology — Functional Reasoning from SPRO

Example — determining whether a sales order item is free of charge:

  • Table TVAP stores item category configuration. If the PSTYV field (item category) has the value TANN, the ontology knows that this configuration is not billing-relevant — the item is therefore treated as Free of Charge.
  • Table TVEP stores schedule line key configuration, which determines whether a sales order item triggers a stock movement — for example, movement type 601 for goods issue, 638 for returns, or blank for non-inventory items.

The ontology maps these technical values to business-level properties — so the AI assistant can reason in terms of BillingRelevance and StockMovement, not raw table fields.

Step 4 — Delivering the Ad Hoc Insight​

With the ontology populated from SAP configuration, the AI assistant now has most of the knowledge required to answer the business user's question — on demand, without a new report.

Ontology — AI-Generated Insight Delivery

This illustration shows a simplified example of a Power BI DAX measure generated by Copilot, based on Ontology knowledge.

Important note: This is a simplification only. In the real world, the ontology will not use PSTYV = 'TANN' directly — instead it will use its own semantic property, such as BillingRelevance = 'NotBillable' or FreeOfCharge = true. The AI reasons at the ontology level, not at the raw database field level.

This is the power of the combined AI@CDS and Ontology approach: business users get answers to questions that could never be pre-built — and the system becomes smarter as more SAP configuration is mapped into the ontology.