SAP Unveils AI-Native Enterprise Architecture and Launches Relational Foundation Model at TechEd
At the recent SAP TechEd conference in Berlin, SAP unveiled a major strategic pivot toward an AI-native enterprise architecture, signaling a fundamental shift in how the company builds and delivers business software. The announcement centers on a wave of AI innovation designed to embed sophisticated capabilities across the entire SAP suite, accelerating enterprise automation and driving demonstrable business value. The core of this strategy is the new relational foundation model, SAP-RPT-1, alongside significant enhancements to its data foundation, including SAP HANA Cloud and the SAP Business Data Cloud (BDC).
This development is critical because it addresses the persistent challenge of integrating complex Machine Learning and generative Artificial Intelligence into the highly structured, relational data environments that underpin global business operations. SAP claims that by year-end 2025, it will have delivered 400 SAP Business AI use cases, including 40 Joule Agents, highlighting the rapid pace of integration. The company asserts that its existing AI use cases are already translating into significant value-add for large enterprises, underscoring that the future of enterprise technology is intrinsically linked to intelligent, autonomous data processing.
The Foundation of an AI-Native Future
SAP’s strategy is built on the premise that effective business AI requires a unified foundation that seamlessly integrates application, data, and AI capabilities. This foundation aims to provide developers with a non-disruptive platform for faster, smarter work using familiar frameworks.
SAP HANA Cloud and the Power of Context
SAP HANA Cloud is positioned as the definitive database for this new AI-native architecture. Its enhanced capabilities are crucial for grounding autonomous AI agents in rich, real-time enterprise context.
A key technical advancement is the general availability of the Model Context Protocol (MCP) support for SAP HANA Cloud. MCP provides direct access to the multi-model engines within HANA, allowing agents to understand complex relationships across disparate data types within a single in-memory engine. This includes:
Relational Data: Understanding transactions and master data structures.
Spatial Data: Analyzing geographic dependencies for logistics and supply chain optimization.
Vector Embeddings: Performing semantic searches necessary for generative AI applications.
Further solidifying its data foundation, SAP announced an expansion of its knowledge graph engine capabilities, slated for Q1 2026. This innovation will allow customers to automatically generate knowledge graphs from SAP HANA Cloud metadata, drastically reducing the time required for manual modeling from weeks to minutes. Furthermore, the introduction of agentic memory in SAP HANA Cloud provides AI agents with long-term memory, enabling them to learn from past inputs and decisions, making them continuously smarter and more effective in persistent business workflows.
Zero-Copy Data Strategy with Snowflake
Recognizing the reality of hybrid data landscapes, SAP announced a strategic partnership with Snowflake, dubbed SAP Snowflake. This collaboration aims to provide seamless, real-time access to combined, semantically rich SAP and non-SAP data via SAP Business Data Cloud (BDC) Connect.
The core technical feature here is zero-copy data sharing. Enterprises currently leveraging Snowflake can integrate their existing instances with SAP BDC, eliminating the need for costly and complex data movement. This integration, scheduled for general availability in Q1 2026 (for SAP Snowflake) and H1 2026 (for BDC Connect for Snowflake), is a significant step toward creating a unified, accessible data fabric essential for holistic enterprise Machine Learning and AI.
Introducing SAP-RPT-1: A New Category of AI Models
The highlight of the TechEd announcements was the launch of SAP-RPT-1 (“rapid one”), the company’s first enterprise relational foundation model. This model addresses a critical limitation in the current AI landscape: the struggle of general-purpose large language models (LLMs) with highly structured, relational business data.
While LLMs excel at processing text and code, enterprise operations run on tables and structured data. Traditionally, predicting business outcomes (like payment delays, supplier risks, or customer churn) required training hundreds of specialized, “narrow AI” models for each specific task.
SAP-RPT-1 centralizes this capability into one single, pre-trained model specifically designed to understand and predict business outcomes based on tabular data. SAP claims that the model delivers up to 2x better prediction quality compared to narrow models and 3.5x better prediction quality than generalized LLMs for these relational tasks.
SAP-RPT-1 comes in three versions to meet different needs:
SAP-RPT-1-small: Optimized for speed and super-fast predictions.
SAP-RPT-1-large: Focused on highest accuracy for critical decision-making.
SAP-RPT-1-OSS: An open-source version, available on platforms like Hugging Face and GitHub, fostering community innovation.
This release signifies a new category in the Machine Learning space, focusing on data science within the highly specific context of enterprise resource planning (ERP) and business operations.
Orchestrating the Agentic Enterprise
Beyond foundation models, SAP is heavily invested in the automation and orchestration of AI agents. The company is moving toward providing role-based AI Assistants, accessed through Joule, which bring together specialized agents (e.g., for cash collection, treasury, and more) to provide a unified, agentic experience for every core business role.
To foster broader innovation, SAP introduced Joule Studio, a low-code environment allowing users to extend pre-built SAP agents with custom fields and tools, or to build entirely new custom agents that integrate seamlessly with all existing Joule agents and the SAP BDC.
For professional developers, the SAP Cloud SDK for AI now supports agentic development, offering flexibility to use external frameworks like LangGraph and CrewAI while ensuring deep integration with SAP’s core systems.
Crucially, SAP is pushing for interoperability by making Joule Agents compatible with the proposed Agent-to-Agent (A2A) protocol. This standardized protocol will allow SAP and third-party agents (from partners like AWS, Google, Microsoft, and ServiceNow) to discover, describe, and collaborate with each other, dramatically increasing the potential for cross-system automation and productivity across the enterprise landscape. Centralized governance for this complex agentic environment is provided by SAP LeanIX agent hub for control, and SAP Signavio for tracing agent actions and identifying bottlenecks.
Digital Sovereignty and the European Context
SAP also addressed the increasing demand for regional AI services with a focus on digital sovereignty, particularly in Europe. The company is expanding its SAP Cloud Infrastructure offering through a partnership with Deutsche Telekom, as part of the Industrial AI Cloud project. This collaboration will provide secure, high-performance infrastructure in Germany, ensuring that AI Foundation services and frontier AI models (from partners like Mistral and Cohere) align with local European regulations, data protection standards, and ethical values. This move is critical for fostering trusted, applied AI across European public institutions and regulated industries.
Future Implications and Takeaway
SAP TechEd showcased a commitment to moving AI beyond isolated features into a core architectural principle. The launch of SAP-RPT-1 and the extensive work on data integration and agentic automation signal that the focus of AI innovation is shifting from generalized intelligence to deeply embedded, specialized business intelligence. The immediate challenge for enterprises is to leverage these specialized tools responsibly, utilizing governance systems like LeanIX and adhering to the developing A2A protocol to create secure, interconnected, and highly productive autonomous workflows. The future of the global enterprise system is becoming undeniably AI-native, driven by sophisticated Machine Learning that understands the structure and semantics of business data.
Source: https://news.sap.com/2025/11/business-ai-innovation-unveiled-at-sap-teched/
