How we build a company's ontology

What it is about

An agent that processes an order, holds back a shipment or prepares a credit note needs something most companies have never written down: their operating model. This report describes how we build it, layer by layer: the entities and relationships of the business, the deterministic functions that compute over them, and the typed actions that are the only things authorised to change anything, with permissions, a signature where a rule demands one, and a window to undo. It separates an ontology from a semantic layer, which is a dictionary of metrics that reads and does not write, and sets out the limits the system will not let anyone break. It is design and method: it contains no measured client results, and the report says why in its limits section. It also describes the rollout order: what the company gets at each step, and why it can stop halfway and keep its own model.

What is inside

1. 1What holds this report up
2. 2What everyone else sells when it sounds similar
3. 3The six layers
4. 4The order of the stack is the order of construction
5. 5What lives in the bottom three layers
6. 6What lives in the top three layers
7. 7We model backwards from the agents
8. 8The nevers and the always
9. 9Twelve rules the engines enforce
10. 10What a verb carries declared
11. 11The journey stops at the signature
12. 12The reading side
13. 13Two cadences and a new crossing
14. 14The connection, and the rules of the copy
15. 15Two levels and a visible breakdown
16. 16What runs on top
17. 17Three classes and a ladder
18. 18When a semantic layer is enough
19. 19The boundary moves; the checklist survives
20. 20How this starts inside a company
21. 21What this report does not prove
22. 22References

Subjects

enterprise ontology · semantic layer · data twin · typed actions · AI agents · agent governance