[RECORDED] There Is No AI Without IA: Information Architecture as the Semantic Foundation

AI retrieves based on linguistic similarity, not semantic understanding. This session examines how vocabulary, taxonomy, ontology, and metadata work together to create the semantic foundation that separates enterprise AI that retrieves reliably from one that guesses.

1. Why Raw Content Fails AI Retrieval: Documents written for humans rely on context, judgment, and experience that AI systems do not have. Without explicit structure and precise meaning, AI retrieves by similarity, and in enterprise environments, similar is not the same as correct.
2. Vocabulary Normalization and Consistent Retrieval: When different teams use different words for the same concept, AI treats each term as a separate concept and returns different results. Vocabulary normalization establishes a preferred term and maps all variants to it, ensuring consistent retrieval regardless of how the question is phrased.
3. Taxonomy as the Boundary Layer: Taxonomy defines what belongs together and what must stay separate, giving AI reliable retrieval boundaries. Without it, AI has no way to distinguish related content and retrieval errors follow.
4. Ontology Enables Reasoning Across Concepts: Taxonomy tells AI what things are called. Ontology tells AI how concepts relate. With those relationships explicitly mapped, AI can follow chains of meaning and return the full context a user needs, not isolated fragments.
5. Metadata and Applicability Logic: Metadata is what allows AI to retrieve by meaning rather than similarity: the right version, the right jurisdiction, the right role. Applicability logic must be designed into the system before content is ingested, not added after retrieval breaks down.

Speakers