Ontology

The rule book that defines what each type of data is, how it connects and what trade terms mean.

Definition

An ontology is the rule book for an AI agent's data. It defines what each type of thing is (a product, a SKU, a customer, a distributor), how they connect, and what trade terms mean. That's how an agent knows that "2 peti" of a product packed 20 to a case means 40 units, not two.

Key points

<ul><li>Think of it as the legend of a map. The map shows the data, and the legend explains what every symbol means.</li><li>Example: a running shoe has a SKU, a brand, a product family and facets like material, colour and polyester %. The ontology says those properties exist and how they connect.</li><li>Without an ontology, agents fill gaps by guessing, which leads to hallucinations.</li><li>Ontologies get reused within an industry. The first one in a vertical is built from scratch, and later ones mostly reuse it with some customisation. Once a business runs on an ontology, switching means rebuilding every workflow, agent and permission rule.</li></ul>