SAP has moved to expand its development, data, and control tools all at once, targeting the “agentic AI” market in which artificial intelligence goes beyond learning about work to planning and executing tasks across multiple systems on its own.
SAP SE (NYSE: SAP) said on the 28th that it unveiled new features and partnership plans for Joule Studio, Joule Work, and the SAP Business AI Platform at its annual developer event, SAP TechEd 2026, held in Berlin, Germany, on the 27th (local time). The core of the announcement is providing the foundation for companies to build AI agents based on their own data and put them to work on real tasks while controlling permissions and costs.
The most striking change is the agent-building environment. Within the SAP Business AI Platform, business users and developers can build agents with Joule Studio’s no-code and pro-code tools, and the number of selectable AI models has surpassed 70. Customers in the early adopter program will also be able to design and run n8n workflows within Joule Studio, with plans to broaden access in the future.
Joule Work, which breaks complex tasks into multiple steps, plans them, and coordinates the order of execution, is also being enhanced. It uses open technologies such as Agent2Agent (A2A), a communication protocol between agents, and the Model Context Protocol (MCP) to connect different systems and business domains. A gateway with access controls sits on the path where external agents call Joule agents, ensuring that company-defined permissions and operating policies are maintained.
The newly introduced SAP Enterprise AI Gateway combines authentication, policy enforcement, request routing, monitoring, usage tracking, and auditing into one. It serves to manage the use of AI services within an enterprise from a single place. An AI governance assistant is designed to support AI system classification and compliance management, helping companies respond to the European Union’s AI Act.
Key AnnouncementsDetailsJoule StudioMore than 70 selectable AI models, n8n workflow design and execution (early adopters), added observability of agent behavior, model usage, and spendingJoule WorkMulti-step task planning and coordination based on A2A and MCP, access-control gateway applied to external agent callsSAP Enterprise AI GatewayIntegrated authentication, policy enforcement, routing, monitoring, usage tracking, and auditingSAP Business Data CloudIntroduction of ontology manager, additional connections of customer-specific SAP and external data sources to the knowledge graphSAP-RPT-2.0Expanded scope to include anomaly detection and embedding-based analysis in addition to classification and regression
Some of the cost burden is also being eased. Joule Studio has added management features for checking agent behavior, model usage, and spending. SAP will waive, through December of this year, the cost of deploying and running custom agents on the SAP-managed runtime for some customers, and will provide design tools for building and testing agents and n8n workflows free of charge within fair-use limits.
The data foundation agents use for decision-making is also broadening. The ontology manager being introduced to SAP Business Data Cloud (BDC) supports the task of defining relationships between SAP and external data. It is a tool that helps AI understand how information scattered across departments and systems is connected. SAP’s knowledge graph will additionally connect customer-specific SAP and external data sources, enabling users to query internal data in natural language and receive answers grounded in it. SAP emphasized that ownership of customer data rests with the customer and that it does not share that data externally for training or improving large language models.
Data platform collaboration is expanding as well. Using Dremio technology, SAP is strengthening support for Apache Iceberg and Apache Polaris and adding a query engine that queries multiple repositories without moving or copying data. With Databricks, SAP will broaden the SAP Databricks capabilities offered in SAP BDC. It plans to introduce Lakeflow, Agent Bricks, Unity Gateway, and Lakebase, and to pursue integration between SAP application data foundations and Joule and Genie. SAP Reltio, recently unveiled at SAP Connect, will also be introduced to BDC to consolidate and clean up customer, supplier, and product information.
AI models specialized for structured data analysis were also unveiled. SAP-RPT-2.0 will expand its support to include anomaly detection and embedding-based analysis in addition to classification and regression analysis, with the goal of reducing manual work in data preparation through features that handle multiple tables directly. SAP-RPT-TS for time-series forecasting enables users to forecast key business metrics in SAP’s integrated business planning solution without building separate custom machine learning models.
SAP also announced a plan with HPE to deploy SAP-RPT on the HPE Spaceborne Computer-2 aboard the International Space Station (ISS). The approach performs analysis inside the space station rather than transmitting data to the ground, and is expected to be used for classifying anomalies in sensor data, predicting the likelihood of failures, restoring missing sensor values, and generating mission status reports for astronauts and ground control teams.
Philipp Herzig, SAP’s chief technology officer, said, “This year SAP is showing how AI is evolving beyond the stage of learning about customers’ businesses to the stage of actually executing.” He added, “Joule Work is where, when you enter an intent, the work gets done within a single experience spanning all of SAP and beyond.” He also said, “The same SAP-RPT model that powers global supply chains can also work in space.”
The announcement aligns with a trend in the enterprise software industry in which the axis of competition is shifting from adopting generative AI to autonomous, execution-capable agents. For SAP, the strategy is to leverage its vast existing customer base and more than 50 years of accumulated process and industry knowledge to embrace external AI models and agents while retaining control over enterprise permissions, regulation, and costs. The expansion of integration with open protocols and external platforms such as Databricks and Dremio is seen as targeting enterprise customers’ desire not to be locked into a particular model or cloud.
Market observers believe that, with model usage fees and computing costs from running agents emerging as a new burden for enterprise AI adoption, usage tracking and governance features will determine the actual pace of adoption. Some note that the key question is how much cost savings SAP’s temporary free policy will translate into after the shift to paid usage.