Twenty-plus years working with ontologies, semantics, and data models — and I keep hearing the same conflations in every AI conversation. Let me clear them up.
1. “We need a taxonomy AND an ontology.”
No. A taxonomy is a hierarchy — broader/narrower, parent/child. An ontology already expresses that, plus much more: relationships, constraints, properties, rules.
If you have an ontology, your taxonomy is a view derived from it, not a second artifact to build and maintain. Teams that maintain both separately end up with two sources of truth that drift apart.
Keep one model. Project the hierarchy when you need it.
2. Knowledge management ≠ context management.
Knowledge management is organization-scoped and durable: capturing, structuring, and governing what the enterprise knows. Context management is AI-specific and use-case-scoped: deciding what a model needs to know right now, for this task, and assembling it at runtime.
They draw on the same assets but answer different questions. Knowledge management asks “what do we know?” Context management asks “what does this agent need, in this moment, to not be wrong?”
3. Context layer ≠ context.
The context layer is the machinery: retrieval, assembly, injection, versioning, access control — the tech that gets the right information in front of the model. Context is the content itself: the modeled semantics, the entities, the relationships, the business meaning.
You can buy a context layer. You cannot buy your context. That part is modeling work, and there is no shortcut around it.
The Pattern
The pattern behind all three: mistaking the container for the content, and the tooling for the thinking.
The tooling is getting commoditized fast. The modeling is where the durable advantage lives.