Introducing the SAP AI Reference Architecture

The SAP AI Reference Architecture provides Enterprise Architects with a shared language for navigating a reality they encounter every day: AI is now woven throughout the SAP landscape, and organizations increasingly expect Enterprise Architects to articulate how these capabilities fit together. With applications, data, and AI becoming increasingly interconnected, EAs need a straightforward and consistent approach to demonstrate where AI is positioned, its relationship with existing systems, and its impact on delivering tangible business outcomes.

The reference architecture provides the needed clarity. It organizes the landscape into four understandable layers, each representing a part of the journey from a user’s intent to an AI-supported business action. For Enterprise Architects, this structure provides a practical way to explain AI capabilities to business and technical stakeholders, making complex AI interactions easier to visualize and discuss.

1. User Experience layer – Where users meet AI

   This is where users begin their interaction. Joule and embedded AI patterns provide a single, intuitive starting point for users to initiate a conversation with SAP, using natural language. Rather than navigating multiple applications, users simply describe what they wish to accomplish in their own words to receive guided, context-aware workflows that span SAP S/4HANA, SAP SuccessFactors, SAP Customer Experience, and more.

2. Process layer – AI inside end-to-end business flows

   Core SAP processes, such as order-to-cash, hire-to-retire, plan-to-fulfill, and others, remain the backbone of execution. Embedded AI capabilities and Joule Agents operate within these workflows to automate steps, surface insights, and recommend next actions. Crucially, these AI features remain anchored in SAP’s standard, governed business processes.

3. Foundation layer – Trusted data, context, and model access for AI

   Intelligence receives grounding in this layer. SAP HANA Cloud (including its vector engine), SAP Business Data Cloud, and SAP Knowledge Graph provide the business data, relationships, and semantics required by AI. The Generative AI Hub offers secure, governed access to the SAP Family of Models—including SAP’s own foundation models, such as the SAP Domain Models (SAP-native models trained on SAP code, metadata, processes, and best practices) and the predictive SAP-RPT-1—alongside leading third-party large language models, with deployment options depending on the chosen setup. Integration and identity services ensure models operate within the boundaries of authorized, high-quality, auditable data. These grounding, data, and model-access capabilities are unified under the Contextualize & Reason pillar of the SAP Business AI Platform.

4. Platform layer – The technology base

   SAP’s platform capabilities—previously delivered as SAP BTP—are unified within the SAP Business AI Platform. The Build pillar (Joule Studio and SAP Integration Suite) is where extensions, custom agents, and integrations are created and connected, while the Govern pillar (SAP AI Agent Hub, the SAP-managed agentic runtime, and identity, access, and security) runs and oversees them. Capabilities EAs already know, such as security services, lifecycle management, and monitoring, continue to operate AI workloads alongside traditional applications.

   Together, these layers form a flywheel at the core of SAP’s strategy.

   Applications generate data. Data then fuels AI. Then, AI drives and returns improved actions to the applications your organization already uses. Over time, this continuous cycle strengthens business processes, improves decision-making, and increases the value organizations derive from their SAP landscape.

The technologies that deliver these layers have converged into a single SAP Business AI Platform. The four conceptual layers above remain a useful way to explain how AI flows from a user’s intent to a business action, and they map directly onto the platform: the User Experience layer is delivered through the Engage layer (Joule, Joule Work, Joule Assistants, and Joule Agents); the Foundation layer corresponds to the Contextualize & Reason pillar (SAP Knowledge Graph, SAP Business Data Cloud, and the SAP Family of Models via the Generative AI Hub); and the Platform layer is delivered by the Build pillar (Joule Studio and SAP Integration Suite) and the Govern pillar (SAP AI Agent Hub, the SAP-managed agentic runtime, and identity, access, and security). For Enterprise Architects, this means the same end-to-end story now rests on one unified, governed foundation rather than a set of separate services.

This architecture enables EAs to explain not only what AI can do but also how it works within SAP, making AI conversations clearer, more structured, and better aligned with real enterprise landscapes.

Understanding the SAP AI Reference Architecture gives you a clear, shared language to explain how AI fits into the SAP landscape. By breaking AI-enabled scenarios into user experience, process, foundation, and platform layers, you can clearly show where AI operates, how it connects to data and applications, and how it drives real business outcomes. You are now equipped to explain SAP AI in a structured, end-to-end way that resonates with enterprise landscapes.