Fundamentals · 24 July 2026
What is a knowledge graph? And what is an ontology for?
Things and their relationships, stored as a network. Why this structure makes the difference for AI in building operations.
By Aidocr
Key takeaways
- A knowledge graph stores things (nodes) and their relationships (edges), not texts.
- A question becomes a short walk through a network, instead of a text search.
- The ontology is the blueprint: it defines which things and relationships exist and which rules apply.
- Validity becomes a relationship with a date, so which version applies today is unambiguous.
A knowledge graph does not store texts, it stores things and their relationships. The things are called nodes: a building, a piece of equipment, a room, a contract, a company, a document. The relationships are called edges: located in, maintained by, belongs to, supersedes. Out of millions of such connections grows a network you can query like a map.
An example from operations
Ventilation unit L-204 is located in building B, maintained by company X, governed by contract W-2023-17, last serviced on 14 May, evidence on page 3 of the report. Five statements that today typically live in five different places: CAFM, PDF contract, email, spreadsheet, folder. In the graph they form one connected picture. The question of which contracts cover the ventilation in building B is then no longer a text search but a short walk through the network: from building B to its units, from there to the contracts.
And the ontology?
The ontology is the graph's blueprint. It defines which kinds of things exist, which relationships between them are allowed, and which rules apply. A maintenance always refers to a piece of equipment. A change order can supersede a line item in the bill of quantities, from a specific date onwards. Rules like these let the machine draw conclusions and spot contradictions instead of merely storing data.
For the physical side of the built world, established vocabularies exist, first and foremost IFC for building models. For the operational side, meaning contracts, maintenance, faults and responsibilities, there is no continuous standard. That vocabulary is built per company, from its own terms and processes.
Why this is more than tidiness
The difference shows in validity. If the change order supersedes the line item as an edge with a date, the question of today's valid price has one clear answer, instead of remaining a guessing game between two similar-sounding passages. And because every answer is a path through the network, that path can be shown: traceability is part of the structure, not an afterthought. How search and graph together produce AI answers is covered in the next article: What is GraphRAG?
Frequently asked questions about knowledge graphs
- What is a knowledge graph?
- A knowledge graph stores things as nodes, such as a building, a piece of equipment or a contract, and their relationships as edges, such as located in, maintained by, supersedes. Out of many connections grows a queryable network.
- What is the difference between a knowledge graph and a database?
- A classic database stores values in tables. A knowledge graph represents the relationships between things explicitly and walks along them, instead of assembling them via keys.
- What is an ontology for?
- The ontology is the graph's blueprint. It defines which kinds of things exist, which relationships are allowed and which rules apply, so the machine can draw conclusions and spot contradictions.
- Are there standards for building ontologies?
- For the physical side, yes, first and foremost IFC for building models. For the operational side, with contracts, maintenance and responsibilities, there is no continuous standard, so that vocabulary is built per company.