Information — Information Architecture for Enterprise AI Adoption
What should AI read, through which path should it retrieve it, and who is accountable for its quality?
About This Section
This site started from MCP and Skills, then expanded into Agents, Strategy, and Workflows. Meanwhile, in real-world enterprise LLM adoption, keywords such as RAG, DB/API, Knowledge Graph, and GraphRAG are treated at the same level as MCP and Skills.
This section is the home for those topics — viewed not as implementation mechanisms but as information architecture: which information to keep as documents, which to keep as structured data, through which path AI should retrieve it, and who is accountable for its quality and freshness.
Audience: Developers and architects designing AI adoption for enterprise systems. No Claude-ecosystem-specific knowledge is assumed.
NOTE
While the existing MCP / Skills sections are shelves of "implementation mechanisms", this section is a shelf of vendor-neutral design decisions. The two are connected as posts on the Architecture Map.
Relation to the scope of this site
For why this site scopes "AI agents" to LLM-driven agents, see The Scope of This Site (FAQ). This section covers the information side connected to the reasoning core: RAG and data curation are responses to the LLM's structural constraints (Knowledge Boundary / Hallucination), while information governance (ownership, access rights, quality) is a precondition for the connection that holds even before any LLM is involved.
Structure
| Page | Central Question | Status |
|---|---|---|
| Architecture Map | Where does each keyword sit in the overall architecture? | ✅ Published |
| RAG / GraphRAG | How to make document knowledge searchable (Architecture Map §2; no dedicated page) | ✅ Covered on the map |
| Information Governance | How do the five questions — owner, form, path, permission, quality — depend on each other? | 🚧 Planned |
| Adoption Failure Modes | How to make Evals, Prompt Injection defenses, and Human-in-the-Loop preconditions? | 🚧 Planned |
TIP
Start with the Architecture Map — it shows in a single view how this section relates to the five layers, each layer's pages, and strategy / workflows. The definition of the five layers themselves is II.1 Five layers.
Related Documents
- III.4 Memory — Conceptual foundation of Memory and Knowledge Graph
- MCP / Semantic Layer — Design discipline for structured data access
- MCP / Security — Security of connection paths
Last updated: August 2026