We’ve made phenomenal progress embedding AI across the suite. By the end of 2025, we will have 400 SAP Business AI use cases delivered in our solutions, including 40 Joule Agents, building on 2,100 Joule Skills. Our existing more than 300 use cases translate into 441 million EUR value add for a company with 10 billion EUR annual revenue.
This month at SAP TechEd Berlin, we announced a wave of SAP Business AI innovations all built on the same technology foundation that powers our embedded AI capabilities that we are now delivering to our customers and partners, allowing them to add even more value in the future.
We showed how the future of enterprise software is built on an AI-native architecture, powered by SAP app, data, and AI foundation. With this approach, we are enabling a platform shift across the tech stack in a non-disruptive fashion, empowering developers to work faster and smarter using the frameworks and tools of their choice.
SAP HANA Cloud and SAP Business Data Cloud: powering our AI-native future
SAP HANA Cloud is the database for SAP’s AI-native software architecture and the foundation of our broader data fabric strategy. At SAP TechEd, we announced new AI capabilities for SAP HANA Cloud that spur AI innovation.
For example, Model Context Protocol (MCP) support for SAP HANA Cloud is now generally available. This provides direct access to rich multi-model engines. Agents can be grounded in full enterprise data context: navigating relationships across customers and suppliers, understanding geographic dependencies through spatial data, and performing semantic searches through vector embeddings — all within a single in-memory engine.
We’re also expanding SAP HANA Cloud knowledge graph engine capabilities (Q1 2026) so customers can automatically generate knowledge graphs from SAP HANA Cloud metadata. What used to take weeks of manual modeling can now happen automatically in minutes. But that’s not all. We’re also enabling agentic memory in SAP HANA Cloud. With long-term memory, AI agents can memorize past inputs and decisions — learning and remembering just like humans — and become continuously smarter.
These advances show that SAP HANA Cloud is truly powering an AI-native future. Read more here.
Bringing together the power of SAP BDC and Snowflake
We are bringing the power of Snowflake together with SAP Business Data Cloud (SAP BDC), calling it SAP Snowflake. This partnership enables zero copy data sharing with Snowflake via SAP BDC Connect.
Enterprises already using Snowflake today can leverage SAP BDC Connect to integrate their existing instances of Snowflake with SAP BDC, giving them seamless, real-time access to combined, semantically rich SAP with non-SAP data in SAP BDC. SAP Snowflake will be made generally available in Q1 2026, and SAP BDC Connect for Snowflake in H1 2026. Find more information here.
SAP-RPT-1: a new category of AI models
One of our most exciting announcements at SAP TechEd was the launch of our first enterprise relational foundation model SAP-RPT-1, pronounced: “rapid one.”
Businesses run on structured data. But large language models (LLMs) struggle with a general understanding of table structures and associated semantics. This requires the use of machine learning, or “narrow AI,” for tasks like classification, regression, and more. But classical machine learning necessitates training a model on each task, which easily can lead to hundreds of separate models.
SAP-RPT-1 puts them all into one single, pre-trained model that understands relational business data and predicts business outcomes. Unlike language, image, or video models, SAP-RPT-1 accurately predicts business based on tabular data such as payment delays, supplier risks, upsell opportunities, customer churn risk, and more.
We believe that SAP-RPT-1 is a super capable foundation model today. It provides up to 2x better prediction quality compared to narrow models and 3.5x better prediction quality as compared to LLMs. Dive deeper in this SAP Community blog.
SAP-RPT-1 comes in three versions. SAP-RPT-1-small is for super-fast predictions and SAP-RPT-1-large is for highest accuracy. Both will be generally available in Q4 2025 in the generative AI hub in AI Foundation. SAP-RPT-1-OSS is the open-source version, available in Hugging Face and GitHub.
You can test SAP-RPT-1 today with your data or our use case data samples via no-code UI or via API in the new SAP-RPT-1 playground, an intuitive and interactive space to test for free and open to everyone Access the playground at rpt.cloud.sap and learn more about new models here.
We are continuously adding new capabilities to AI Foundation and models to the generative AI hub, empowering developers to experiment with orchestration tools and leading models to scale AI development and productization across SAP and non-SAP environments. For example, Perplexity is now generally available in the generative AI hub, so users can correlate business data with external data from the internet. Evaluation Services and Prompt Optimizer, in close collaboration with NotDiamond, are now also generally available in AI Foundation, freeing up users to adopt the most appropriate model for their use cases without the need for rewriting prompts. Read more here.
Digital sovereignty made in Germany, for Europe
Digital sovereignty is becoming increasingly important, reflecting the need for regional AI services that align with local regulations, standards, and values. As an example, Europe will benefit from its own strong, trustworthy infrastructure to support innovation, data protection, and ethical AI.
AI Foundation, including various models and all the services we offer, is already available on our own cloud infrastructure. As a next step, we are expanding our SAP Cloud Infrastructure offering in our SAP data center in Walldorf, Germany, to Deutsche Telekom through the Industrial AI Cloud project, providing secure, high-performance infrastructure for AI innovations across public institutions, defense, and society. SAP delivers SAP Cloud Infrastructure, SAP Business Technology Platform, and applications — including our AI Foundation with frontier AI from Mistral, Cohere, and others — on Telekom’s Munich data center. Both companies uphold the highest standards of data protection, security, and reliability.
This marks a milestone as more European companies join the Industrial AI Cloud project, advancing applied AI across Europe with trusted, business-embedded solutions that unlock the full potential of industry data. See the announcement here.
Enabling customers to build, extend, share, and orchestrate AI agents
To help manage Joule Agents and Joule skills, we have introduced the concept of AI Assistants — role-based AI teammates, accessed through Joule — like a financial assistant that brings together agents for cash collection, treasury, and more. We will provide AI Assistants in Joule for every core business role, offering our users an agentic experience like never before.
Out-of-the-box Joule Agents are powerful, but we know that every company has unique requirements. We believe AI should adapt to users’ systems, not the other way around, so we are enabling them to use Joule Studio to extend SAP’s pre-built agents with custom fields, tools, and reasoning logic while retaining all the deeply grounded integration capabilities SAP provides. Joule Studio also provides low-code tools to build custom agents that integrate with all other Joule Agents, Joule skills, and SAP BDC.
Using a low-code approach, users can build Joule Agents visually with natural language and drag-and-drop. But we also want to meet the needs of developers who want ultimate flexibility. Our pro-code approach gives developers the freedom to build agents using the agentic framework of their choice — for example, LangGraph, CrewAI, Google’s Agent Development Kit, and more. SAP Cloud SDK for AI now supports agentic development, ensuring these pro-code agents can be seamlessly integrated and giving developers the best of both worlds: deep integration and full flexibility.
No matter how you want to build agents, an important question is how to integrate them into the larger ecosystem beyond SAP. We’re making Joule Agents fully compatible with the agent-to-agent (A2A) protocol soon, so agents can discover and collaborate with each other.
A2A exposes rich semantics describing an agent’s capabilities, allowing both SAP and third-party agents to work together seamlessly. We are collaborating with partners — AWS, Google, Microsoft, ServiceNow, and more — to standardize this protocol for full interoperability. This capability will allow Joule to orchestrate tasks across multiple agents, both SAP and non-SAP, increasing automation and productivity across the enterprise. Read more here.
To manage and govern agents across the enterprise, SAP LeanIX agent hub is now generally available, providing centralized control of SAP and non-SAP agents. In addition, agent mining with SAP Signavio is available now for tracing agent actions, benchmarking against KPIs, and identifying bottlenecks or opportunities for agents to further improve business.
No SAP TechEd without ABAP news
The ABAP journey continues with SAP-ABAP-1, which will be available in the generative AI hub in Q4 2025. Trained on ABAP code, it is designed to build ABAP AI use cases, enabling developers to build smarter, custom AI solutions in modern ABAP code. Dive into the ABAP news in this SAP Community blog.
In addition, ABAP Cloud development is coming to Visual Studio (VS) Code. The new ABAP Cloud extension for VS Code delivers a streamlined, file-based development experience with built-in AI assistance. Powered by an ABAP language server, it will initially support SAP Fiori UI service development and expand to additional ABAP Cloud scenarios over time. This brings ABAP development into the same environment where developers already build with UI5 and CAP. General availability is planned for Q2 2026. Read more in this SAP Community blog.
What’s next: embodied AI and quantum
SAP TechEd is always an opportunity to look to the future. This year, that future includes not just humans, but also autonomous devices, including humanoid robots.
By integrating Joule Agents natively with robots, SAP is bringing business logic into the physical world, enabling a wide range of autonomous devices to operate with enterprise context. We highlighted our strategic partnerships with robotics companies and system integrators to serve customers like Sartorius, Bitzer, and Matur Fompak, demonstrating how our expanding physical AI ecosystem enables robots to understand business processes and execute complex tasks autonomously.
Early proof-of-concept deployments show Joule successfully integrated with SAP business applications and autonomous systems across asset performance, logistics, field services, and warehouse operations. While still in the pioneering stage, these implementations illustrate how SAP is extending Joule to serve both human users and autonomous devices, shaping the future of enterprise AI.
Read more about the partnerships and implementations here.
AI is a new compute paradigm that changes everything. But there is another compute paradigm on the horizon: quantum computing. It’s early days, but SAP is driving the future of enterprise computing with a vision to help businesses get ready for quantum computing.
SAP is not building quantum hardware; instead, we are focusing on creating quantum algorithms for business applications. These solutions are simple to deploy — on when needed, off when not — and are designed to be hardware-agnostic, collaborating with partners such as IBM to ensure seamless integration without re-platforming. This approach will enable organizations to unlock operational efficiency and drive better business results at enterprise scale.
I couldn’t be more excited about what’s next for our customers’ future as we bring SAP’s AI-native architecture to life.
Philipp Herzig is CTO of SAP.
