The project is being developed with a free note-taking app, Obsidian, which offers many important functionalities for the task and is reasonably easy to deploy and edit by anyone interested, including old timers like me. I am using ChatGPT 4o to test an LLM’s response to the use-case, i. e. if the LLM improves its understanding of the specific business. The Decision-Making Architecture is made of: Ontology In computer science, Ontology is a standardized architecture for the representation of Knowledge, in the form of agreed domain semantics. The fundamental asset of ontologies is their relative independence of applications, i.e. an ontology consists of relatively generic knowledge that can be reused by different kinds of applications/tasks. An ontology is built by connecting triples of entity, relationship, and another entity. For example: Ontology: Risk Driver → Informs → Decision Some authors call this a “triplet of subject, verb and predicate”. Knowledge graphs are built the same way but are more specific (answering “which risk driver informs which decision?”) to the use-case: Knowledge Graph: Product Contamination (Risk Driver) → Informs → Product Safety Validation (Decision) An ontology can be designed by an institution or group to communicate knowledge in a specific domain that many organizations can replicate in computer applications. This is widely done in health sciences, biology and many other domains. For example, my custom GPT can retrieve triples from the Obsidian files and build a knowledge graph in seconds for practical applications. Applied Ontology | Knowledge Graphs If an organization adopts an ontology, it must develop an applied ontology (the “metadata layer”) and one or more knowledge graphs. This means that the organization adapts to the generalized framework, adding specific entities underneath the ontology, and will transfer that information to a specialized graph software (such as Neo4j, GraphDB, ArangoDB, Memgraph, Metaphacts and many others). Applied ontologies also add real-world details, such as causes of risks and relevant controls, enabling advanced querying and automation. The Enchiridion applied ontology provides a blueprint that any organization can tweak, regardless of industry, because it covers all common decision points and respective trade-offs. Graph databases  A graph database is built in one of the graph applications mentioned above, by adding real data underneath a knowledge graph. For example, the results of the risk assessments of product contamination and the test data from product sampling can be connected to the rest of the decision system.