📊 SAP TechEd 2026 — Session ST108 Presentation
Summary
This SAP TechEd 2026 session (ST108) by Niklas Siemer (SAP) and Sebastian Schmidt (Digital Science/metaphacts) presents how to build a custom knowledge layer using SAP HANA Cloud’s multi-model capabilities.
Key Topics
1. Three-Level Knowledge Hierarchy
- Level 1 — SAP Knowledge Graph: Enterprise-wide, prebuilt semantic backbone with 50+ years of process knowledge (7.3M+ entities)
- Level 2 — SAP Business Data Cloud: Governed data products, semantics, and AI domain models on SAP HANA Cloud
- Level 3 — Custom Knowledge Layer: Customer-built domain graphs, vector stores, and ML models extending the SAP knowledge core
2. Event-Driven Agents
AI agents that sense, contextualize, and respond in real-time using SAP HANA Cloud + SAP Integration Suite advanced event mesh. Use cases include fraud detection, dynamic pricing, supply chain, and predictive maintenance. Reduces polling and token consumption.
3. Custom Knowledge Graphs on SAP HANA Cloud
- RDF-native knowledge graph engine — SPARQL + SQL, ontology persistence
- Multi-model RAG — combines graph traversal, vector similarity search, and SQL filters in a single unified context
- Cross-engine retrieval — graph + vector + relational in one query with 14ms grounding latency
4. metaphactory by metaphacts Partnership
- Visual no-code ontology modeling — domain experts build knowledge graphs without SPARQL
- AI-assisted modeling — LLM copilot translates business questions into W3C ontology proposals
- Zero ETL — direct connection to SAP HANA Cloud, in-engine SQL pushdown
- SHACL validation — automatic enterprise governance guardrails
- 5-minute setup — connect metaphactory to existing SAP HANA Cloud instances
5. Zero-Hallucination Agent Grounding
Demo showed a maintenance scenario: operator asks about Pump-402 alert → agent traverses knowledge graph → joins live telemetry → returns verified response with full audit trail. 100% in-HANA pushdown, 14ms grounding latency.
Key Takeaway
SAP HANA Cloud is positioning itself as the unified multi-model engine (relational + vector + graph) for enterprise AI grounding. The metaphactory partnership enables business users to build knowledge graphs visually, while the event-driven agent pattern (sense-contextualize-respond) represents the next evolution of enterprise AI beyond simple RAG.
Speakers: Niklas Siemer (SAP HANA Cloud Product Manager) and Sebastian Schmidt (EVP, Digital Science)
