Reference file

AI Knowledge Stack — Quick Reference

ai-knowledge-stack-quick-reference.md

AI Knowledge Stack — Quick Reference

Last updated: April 2026. Pricing and ratings current as of this date.

What the knowledge layer does

Gives AI agents and reps access to institutional knowledge (call transcripts, playbooks, CRM data, documented processes) rather than generic LLM output. Four components: ingestion → chunking + indexing → retrieval → delivery.


US Stack

Buy path

Use case Tool Price
Enterprise knowledge search Glean (best-in-class) $50+/user/mo, 100-seat min
Revenue team KM Guru $25/seat/mo
Team knowledge base Notion AI Agents $12-27/user/mo
Agent platform Dust.tt Custom

Build path (custom RAG)

Component Tool Price
Orchestration LlamaIndex (retrieval) or LangChain (agents) Free
Vector DB Pinecone $25-500/mo
Embeddings OpenAI text-embedding-3-small $0.02/1M tokens
Reranking Cohere Rerank $50-200/mo
Total: $900-2,000/mo + 2-4 week build

EU Stack (GDPR-first)

Buy path

Use case Tool Data residency
AI platform + knowledge Langdock EU-hosted, GDPR-native
M365 shops Microsoft Copilot + SharePoint EU DC available
Team KB with DPA Slite or Notion + EU DPA US-hosted, DPA covered

Build path (EU sovereign)

Component Tool Data residency
Orchestration LlamaIndex Self-hosted EU
Vector DB Qdrant Cloud EU EU cloud
Embeddings Mistral (EU) or local all-MiniLM EU-native or on-premise
LLM Mistral Large or Claude via Langdock EU
Total: €800-1,500/mo + 2-4 week build

Decision tree

GDPR personal data involved?
├── No → US stack fine
├── Yes + unregulated → US tools with EU DPA (or Langdock)
└── Yes + regulated OR works council → EU sovereign stack

Stage guide

ARR US EU
€1-5M Notion AI Notion + EU DPA
€5-15M Guru or Notion AI Agents Langdock
€15-50M Glean or custom RAG (Pinecone) Custom RAG (Qdrant EU)
€50-100M Glean + custom RAG Copilot EU + custom RAG
€100M+ Glean Enterprise Aleph Alpha or Copilot EU

Top vendor ratings (G2 / Capterra)

Vendor Rating Reviews Note
Glean 4.8/5 130+ Gartner eMQ Emerging Leader
Guru 4.8/5 624 (Capterra) #1 KM satisfaction
Dust.tt 4.9/5 19 Small sample, very positive
Pinecone 4.6/5 39 #1 vector DB on G2
Weaviate 4.8/5 30 Best knowledge graphs
Qdrant ~12 reviews EU sovereign option
ChromaDB Limited Prototype only, not production
Mem.ai 1/5 2 Avoid

Critical insight

Chunking quality > embedding model choice. Semantic chunking: 0.79-0.82 faithfulness. Naive chunking: 0.47-0.51. A 60% improvement. Design chunking first, pick tools second.


Full vault docs

  • Full AI knowledge stack guide (US/EU reference)
  • 14-platform landscape analysis (2025-2026)
  • G2/Capterra review data (April 2026)