AI Knowledge Stack — Vendor & Pricing Matrix (US / EU)
On-demand reference for the revops-tech-stack skill.
Vendor pricing data collected April 2026. Refresh annually.
The fuller vendor research behind the AI Knowledge Stack: US and EU stack options (buy vs. build), compliance decision tree, stage-appropriate dual recommendations, the key technical insight on chunking, and the vendor review summary. The capability-we're-solving-for framing and pointers live in SKILL.md. (This is distinct from ai-knowledge-stack-reference.md, the condensed reference.)
US Stack — Speed-First, Feature-Rich
Managed platform path (buy)
| Component | Recommended | Alternative | Price |
|---|---|---|---|
| All-in-one knowledge layer | Glean | Guru (revenue-specific) | Glean: $50+/user/mo (100-seat min). Guru: $25/seat/mo |
| Team knowledge base | Notion AI Agents | Slite | Notion: $12-27/user/mo. Slite: $8-15/user/mo |
| Agent platform | Dust.tt | — | Custom pricing |
When to pick this path: Time-to-value matters more than cost or control. Team is non-technical. Budget is $100K+/year for the knowledge layer. Already in Notion or similar ecosystem.
G2 ratings: Glean 4.8/5 (130+ reviews, Gartner Emerging Leader). Guru 4.8/5 (Capterra, 624 reviews). Notion 4.6/5 (10,149 reviews, G2 Leader).
Custom RAG path (build)
| Component | Recommended | Alternative | Price |
|---|---|---|---|
| Orchestration | LlamaIndex (retrieval-optimised) | LangChain (agent-optimised) | Free (open source) |
| Vector DB | Pinecone (managed) | Weaviate Cloud | Pinecone: $25-500/mo. Weaviate: $25-50/mo |
| Embeddings | OpenAI text-embedding-3-small | — | $0.02/1M tokens |
| Reranking | Cohere Rerank | LLM-based | $50-200/mo |
| LLM | Claude or GPT-4 | — | Per-token |
Total cost: $900-2,000/month + 2-4 weeks initial build + 2-4 hours/week maintenance.
When to pick this path: Need proprietary retrieval logic. Engineering capacity available. Want to optimise chunking strategy for specific content. Data sensitivity requires full control.
EU Stack — Compliance-First, Sovereign
The EU stack addresses GDPR, data residency, and works council requirements. This is critical for Neon's Dutch/EU client base.
Managed platform path (buy)
| Component | Recommended | Alternative | Price | Data residency |
|---|---|---|---|---|
| AI platform + knowledge folders | Langdock | — | €20/user/mo + usage | EU-hosted, GDPR-native |
| Enterprise knowledge (M365 shops) | Microsoft Copilot + SharePoint | Google Vertex AI Search | Included in M365 | EU data centre available |
| Team knowledge base | Slite or Notion (with EU DPA) | Guru (with EU DPA) | $8-27/user/mo | US-hosted with DPA |
The honest gap: There is no EU-native equivalent of Glean. Langdock comes closest for the AI layer but its semantic search is weaker than Glean's. For regulated industries (healthcare, finance, government), use the sovereign path below.
Custom RAG path — EU sovereign
| Component | Recommended | Alternative | Price | Data residency |
|---|---|---|---|---|
| Orchestration | LlamaIndex | LangChain | Free | Self-hosted (EU) |
| Vector DB | Qdrant Cloud EU | Weaviate Cloud EU | Qdrant: €27-102/mo. Weaviate: €25-50/mo | EU cloud |
| Embeddings | Mistral embeddings (EU) | Local model (all-MiniLM) | Mistral: API pricing. Local: free | Mistral: EU. Local: on-premise |
| Reranking | Jina Reranker (open source) | LLM-based | Free (self-hosted) | Self-hosted |
| LLM | Mistral Large (EU) | Claude via Langdock (EU wrapper) | Per-token | EU-native |
Total cost: €800-1,500/month + 2-4 weeks initial build + 2-4 hours/week maintenance.
When to pick this path: Regulated industry. Data cannot leave EU borders. Legal/compliance team has specific data sovereignty requirements. Government or public sector contracts.
Compliance Decision Tree
Does client data include personal data under GDPR?
├── No → US stack is fine. Standard DPA with vendors.
├── Yes → Is the client in a regulated industry?
│ ├── No → US tools with EU DPA acceptable for most use cases.
│ │ Langdock as AI layer adds compliance comfort.
│ └── Yes → Full EU sovereign stack required.
│ Custom RAG with Qdrant EU + Mistral + self-hosted.
└── Special case: Works council involved?
└── Yes → Sovereign stack. Works councils in DE/NL/FR often require
on-premise or EU-only data processing. Build this into the
change management plan.
Stage-Appropriate Recommendations (Dual US/EU)
| Stage | US recommendation | EU recommendation |
|---|---|---|
| Seed/Build (€1-5M) | Notion + built-in AI | Notion with EU DPA, or manual |
| Build/Scale (€5-15M) | Guru or Notion AI Agents | Langdock + Notion (EU DPA) |
| Scale (€15-50M) | Custom RAG (LlamaIndex + Pinecone) or Glean | Custom RAG (LlamaIndex + Qdrant EU) or Langdock |
| Expand (€50-100M) | Glean + custom RAG for proprietary data | Microsoft Copilot (EU DC) + custom RAG (Qdrant EU) |
| Enterprise (€100M+) | Glean Enterprise | Aleph Alpha PhariaAI or Microsoft Copilot (EU DC) |
Key Technical Insight
Chunking quality constrains retrieval accuracy more than embedding model choice.
Semantic chunking achieves faithfulness scores of 0.79-0.82 versus 0.47-0.51 for naive chunking — a 60% improvement. A well-designed custom RAG with good chunking on a cheap embedding model will outperform an expensive managed platform with basic chunking. Design the chunking strategy first. Pick tools second.
Vendor Summary (G2 / Capterra / Gartner)
| Vendor | G2 | Capterra | Gartner | Notes |
|---|---|---|---|---|
| Glean | 4.8/5 (130+) | — | eMQ Emerging Leader 2025 | Best enterprise search |
| Guru | 4.6+/5 | 4.8/5 (624) | 4.7/5 Peer Insights (131) | #1 satisfaction in KM |
| Notion | 4.6/5 (10,149) | Listed | G2 Leader (Knowledge Base) | Massive review base |
| Dust.tt | 4.9/5 (19) | — | — | Small sample, very positive |
| Langdock | Limited data | — | — | 37 customer references |
| Pinecone | 4.6/5 (39) | — | — | #1 vector DB on G2 |
| Weaviate | 4.8/5 (30) | — | — | Best for knowledge graphs |
| Qdrant | ~12 reviews | — | — | Speed + EU sovereign option |
| ChromaDB | Limited | — | — | Prototype/local only, memory leaks in production |
| Mem.ai | 1/5 (2) | — | — | Red flag: severe user issues |
Gartner note: No unified Magic Quadrant for knowledge management. Vendors appear across Insight Engines, KM Software, and Gen AI eMQ categories. Forrester Wave KM Q4 2024 names Atlassian (Confluence) as Leader.
Vault References
For the full research behind these recommendations:
Frameworks/AI-Use-Cases/ai-knowledge-stack-us-eu-reference.md— Dual US/EU stack recommendation by stageSources/Research/AI-Knowledge-Layer-Landscape-2025-2026.md— Full vendor research (14 platforms, pricing, features, data residency)Sources/Research/AI-Knowledge-Layer-G2-Capterra-Reviews-2026-04-02.md— Independent review platform data- Offering positioning reference (internal)
references/ai-knowledge-stack-reference.md— Condensed reference for skill use