From user intent to governed AI action — an interactive single-page learning guide to the SAP Business AI reference architecture for AI architects, built as one self-contained
index.htmlwith zero dependencies.
SAP Architecture Center · SAP Business AI · Source study conversation
No build step, no server, no dependencies (diagrams are inlined):
start index.html # Windows
open index.html # macOSOr serve it (e.g. for GitHub Pages):
python -m http.server 8000
# → http://localhost:8000| Section | What it covers |
|---|---|
| Knowledge Map | The six-domain learning path: strategic foundations → reference architecture → agentic AI & Joule → data & semantics → hybrid & platform → enablement & governance |
| Four Layers | The reference-architecture diagram plus a clickable deep dive per layer (user experience, process, foundation, platform) and the RA0005 build-lane components (CAP, GenAI Hub, orchestration pipeline, HANA Cloud engines) |
| Request Lifecycle | Press ▶ and trace "Show me all overdue invoices for customer ACME" through all four layers — five governance checkpoints light up as it runs |
| Stack vs Factory | The SAP Business AI stack vs. an enterprise AI factory, layer by layer, with the three interoperability seams (A2A, partner models, Delta Sharing) and the three-lane build decision |
| Edge Integration Cell | A tabbed before/after: cloud-only integration vs. EIC's cloud-designed, sovereignly-executed split, plus the five strategic use cases |
| Readiness | The Joule integration preparation sequence — identity first, and the one diagnostic question that predicts project difficulty |
Extras: light/dark theme toggle (remembers your choice), embedded architecture diagrams, animated lifecycle trace, hover-free tabbed views, fully responsive.
- The architecture is a communication tool first — four layers giving business and technical stakeholders a shared language for placing any AI capability.
- AI augments the governed process, never a parallel path — embedded AI and Joule Agents stay anchored in SAP's standard end-to-end business flows.
- Grounding and model access share one perimeter — HANA Cloud Vector Engine, Business Data Cloud, and Knowledge Graph sit beside the Generative AI Hub, so retrieval and invocation are governed together.
- Identity precedes retrieval — IAS/IPS gate every request before data moves; access control is a precondition, not an output filter.
- Control plane vs. execution plane — the repeated SAP pattern: governance centralizes (Agent Hub, Integration Suite design time), while data and execution stay sovereign (Delta Sharing, Edge Integration Cell).
- Three lanes, not two — extend Joule, build on RA0005 inside SAP governance, or build in the external AI factory and connect via the A2A / Delta Sharing seams.
Content and diagrams are derived from an SAP Business AI study conversation with Claude (Anthropic), shared by the author, which synthesizes SAP enablement materials — including SAP Architecture Center reference architectures (RA0005 Generative AI on SAP BTP, RA0024 Integrating and Extending Joule) and SAP Learning content. SAP, S/4HANA, SAP BTP, Joule, SAP HANA, and Fiori are trademarks of SAP SE. This unofficial guide is provided as-is for educational use and is not affiliated with or endorsed by SAP SE or Anthropic.
This is an unofficial educational explainer for personal learning. SAP, S/4HANA, Joule, Fiori, and related marks are trademarks of SAP SE. This work is not affiliated with, endorsed by, or sponsored by SAP SE (or Anthropic where study materials are cited). It is not professional advice. See DISCLAIMER.md.