Skip to main content

Discoverability & Cataloging

An AI Product that cannot be discovered may as well not exist.
Discoverability ensures that AI Products can be found, understood, and adopted.
Cataloging provides the structured context for discovery through metadata, search, and AI-assisted exploration.

Why Discoverability & Cataloging Matter​

  • Adoption → Consumers cannot use what they cannot find.
  • Governance → Discovery mechanisms enforce accountability and compliance.
  • Transparency → Catalog entries reveal purpose, ownership, and risk.
  • Trust → Products cataloged via trustworthy mechanisms are more credible.
  • Ecosystem Value → Shared catalogs enable cross-organization reuse and comparability.

Discoverability Requirements​

Every AI Product must be:

  1. Registered

    • Declared in a catalog, registry, or marketplace.
    • Assigned a persistent identifier (e.g., URN, DOI, or URI).

  2. Searchable

    • Indexed by capability, purpose, domain, and metadata tags.
    • Searchable via both keyword and AI-assisted semantic queries.

  3. Described

  4. Linked

    • Connected to upstream and downstream dependencies (see Lineage & Provenance).
    • Cross-referenced across catalogs (internal and external).

Cataloging Standards​

  • Metadata Serialization: JSON-LD, RDF, YAML.
  • Schema Alignment: Must align with BPS and AIPROD schema.
  • Catalog APIs: Support for OData, GraphQL, or REST for automated discovery.
  • AI-Assisted Search: Semantic embeddings and LLM-powered recommendations for discovery at scale.

Example​

AI Product: Enterprise Knowledge Assistant

  • Registered: urn:aiprod:enterprise-assistant:v1.0.0 in corporate AI catalog.
  • Searchable: Indexed under tags: agentic, generative, knowledge management.
  • Described: JSON-LD metadata includes purpose, inputs/outputs, governance markers.
  • Linked: Declares upstream dependency on corporate-knowledge-graph-dp:v3.0.

Summary​

  • Discoverability ensures AI Products are findable, describable, and linkable.
  • Cataloging requires registration, metadata, standards, and AI-assisted search.
  • Without discoverability, an AI Product risks being an undocumented AI asset.

Principle: If an AI Product cannot be discovered through a trustworthy cataloging mechanism, it cannot be trusted, governed, or reused.