Appendix
FAQ, glossary, and sources
Consolidated for reference. Chapter-specific FAQs remain at the end of each chapter.
FAQs
Glossary
Agentic AI
AI systems that perceive their environment, reason over changing inputs, and take multi-step actions to achieve goals, rather than responding to a single prompt.
AQL (Arango Query Language)
A single query language that spans graph, document, vector, key-value, and search data in one execution path.
AutoGraph
Automatically builds and maintains your knowledge graph from enterprise data. Discovers entities, maps relationships, and determines the right graph structure for retrieval — no manual ontology design required.
AutoRAG
Determines the optimal ingestion approach and graph structure for your data so that context is retrieval-ready before any query runs. The quality of retrieval is decided at ingestion time, not at query time.
Contextual data layer
A persistent, multimodel architectural tier that unifies how enterprise data is modeled, connected, governed, and retrieved for AI.
Deep Search
At query time, understands the intent behind each question, breaks it into subqueries, and automatically applies the right retriever for each one — GraphRAG, HybridRAG, or VectorRAG — then aggregates the results into a single governed response.
Frankenstack
An enterprise AI architecture assembled from multiple specialized systems (vector store, graph platform, search index, orchestrator) stitched together at query time.
GraphRAG
Retrieval-augmented generation that uses a knowledge graph to answer multi-hop relationship questions, going beyond vector similarity.
HybridRAG
Retrieval combining semantic similarity (vector) with structured relationships (graph) in one pass.
Knowledge graph
A model that represents entities and their relationships as connected nodes and edges. One component of a contextual data layer.
Model Context Protocol (MCP)
The industry standard for how AI agents discover and invoke tools and data sources through a consistent, vendor-neutral interface. Backed by Anthropic, OpenAI, Google, and Microsoft, and now governed under the Linux Foundation.
Multimodel platform
A data platform that stores and queries multiple data representations (graph, document, vector, key-value, search) natively.
Provenance
The record of where a piece of data originated, how it was transformed, and how it was used — so AI outputs can be traced back to source.
RAG (retrieval-augmented generation)
A pattern for grounding LLM outputs in external data by retrieving relevant content at query time.
Semantic layer
An architectural tier that codifies business meaning and relationships so AI systems interpret data consistently. One dimension of a contextual data layer.
Temporal state
The dimension of context that distinguishes what is true right now from what was true historically.
Sources
- Tavakoli, A., Goodman, B., Soller, H., Rowshankish, K., et al. “Building the foundations for agentic AI at scale.” McKinsey, April 2026. https://www.mckinsey.com/capabilities/mckinsey-technology/our-insights/building-the-foundations-for-agentic-ai-at-scale
- Sheng, E., Zhu, R., O’Rourke, B., Pedzinski, D., Zhang, K. “The Three Layers of an Agentic AI Platform.” Bain & Company, April 2026. https://www.bain.com/insights/the-three-layers-of-an-agentic-ai-platform/
- Sheng, E., Zhu, R., et al. “Governance, Trust, and the Data Foundation.” Bain & Company, April 2026. https://www.bain.com/insights/governance-trust-and-the-data-foundation/
- “The Forrester Data, AI, and Analytics Architecture Model.” Forrester, October 2025.
- Marwaha, R. “Arango Introduces Contextual Data Platform 4.0.” Arango Blog, March 16, 2026. https://arango.ai/blog/arango-introduces-contextual-data-platform-4-0/
- “PSI Reduces Clinical Trial Site Identification From Weeks to Minutes Using AI Powered by a Contextual Data Layer.” Arango Customer Story, 2026. https://arango.ai/case-studies/psi/
- “HPE Aruba Networking: From six siloed databases to one trusted data platform.” Arango Customer Story, 2026. https://arango.ai/case-studies/hpe-aruba-networking/