6-step modeling approach

#	Step	Key question	Key actions
1	Scope	What is the scope?	Define the business questions the ontology must answer: they guide the whole modeling
1b	Reuse?	Does an ontology in GraphForge already partly cover the need?	Yes → start from that ontology and extend it. No → start from scratch
2	Identify	What are the key business concepts?	Extract the nouns from the business processes. Name them in English, explicitly. Avoid opaque acronyms alone (PREA → PRE_ASSEMBLY, or document it in the description)
3	Connect	How are they linked?	Write a subject-verb-object sentence. Explicit verb required (is_composed_of rather than has): the AI agent uses the relation name to understand the link
4	Document	Can the AI agent understand without explanation?	Fill in as much information as possible on nodes and relations. It is the AI agent’s fuel: an undocumented ontology is an ontology that cannot be queried
5	Instantiate	Is there real data in the Knowledge Graph?	Add real data. Without data, the AI agent has nothing to query. A minimal dataset is enough to test
6	Validate	Does the AI agent answer correctly?	Ask the AI agent business questions. If the answer is wrong or incomplete → enrich the information on nodes and relations

♻ An ontology is never frozen. The work is iterative: first model the core of the domain, map some data, validate your business questions with the AI agent, then extend gradually. A simple and correct ontology is better than an exhaustive and incorrect one. Go through steps 1b to 6 again at each significant change in the domain or the needs.