SAP announced a series of AI updates for developers on Tuesday, October 27, 2026, at its annual SAP TechEd conference in Berlin, expanding Joule Studio, Joule Work and the SAP Business AI Platform with new tools to build, govern and run autonomous enterprise agents. The centrepiece is a shift SAP describes as moving from AI that learns a business to AI that runs it: Joule Work becomes the place where an employee states an intent and gets work done across the SAP landscape, while new tabular AI models and governance controls keep that work accurate and inside company rules. The announcement also includes a striking validation: a SAP model running aboard the International Space Station.
Joule Work Gets Agentic Thinking
Joule Work is SAP’s unified interface for AI at work, where employees ask questions, form tasks and hand work to agents inside the applications that already hold their company’s data. The new capabilities add agentic thinking: the ability to plan and orchestrate the steps assistants and agents need to complete complex, multistep business workflows across SAP systems, rather than answering one question at a time. Joule Work agents can now interoperate using open protocols, including Model Context Protocol and Agent2Agent, so work can flow between systems without being rebuilt for each one.
Interoperability is paired with a governed gateway. Third-party agents can securely invoke Joule Agents through that gateway, with enforced access control, so agents from different providers can collaborate without bypassing the permissions and audit trails a company already has. That matters because the biggest barrier to enterprise agents is not capability but control: a finance or human-resources workflow cannot be handed to an agent that ignores who is allowed to approve what. SAP is also adding a new Enterprise AI Gateway to centralise management of AI traffic, consolidating authentication, policy enforcement, routing, monitoring, usage tracking and auditing.
Joule Studio, the development environment inside Joule Work, gets new observability controls, giving developers visibility into agent behaviour, model usage and spend. Users can build agents with no-code and pro-code tools, choosing from more than 70 AI models, and early-adopter customers can design and run n8n workflows inside Joule Studio, bringing deterministic automation to tasks where a fixed set of rules must be followed exactly.
Governance And Data Grounding
SAP is also expanding its governance stack, with a new Joule Assistant and SAP AI Agent Hub, and a new AI Governance Assistant designed to help customers meet EU AI Act compliance requirements. The SAP Knowledge Graph will now incorporate data from additional SAP and non-SAP sources unique to each customer, with answers grounded in their own information. Other enterprise vendors are tackling agent control from different angles, but SAP’s argument is that its agents are safer because they start from the system of record.
Tabular AI, Validated In Space
The most concrete new model is SAP-RPT-2.0, SAP’s latest tabular foundation model. Unlike the large language models that dominate AI headlines, tabular models work on structured business data: rows, columns and tables, the kind of data that fills spreadsheets, databases and enterprise systems. SAP-RPT-2.0 expands beyond classification and regression to support anomaly and outlier detection, and richer analysis through embeddings, with native multi-table capabilities that reduce manual engineering. A companion model, SAP-RPT-TS, will predict values across time-indexed business metrics for forecasting, using SAP Integrated Business Planning, without requiring custom machine learning.
Then comes the space test. In partnership with Hewlett Packard Enterprise, SAP announced the deployment of SAP-RPT on the HPE Spaceborne Computer-2 aboard the International Space Station, which SAP calls one of the most significant edge-compute validations of any enterprise AI model to date. The model is planned to run directly aboard the station to support anomaly triage, predict issues before failures escalate, and generate explainable mission-health reports, all without sending data back to Earth. SAP chief technology officer Philipp Herzig said the same trusted model that powers global supply chains can now work in space.
Why It Matters
SAP’s announcement matters because it shows where enterprise AI is heading: not toward a single brilliant chatbot, but toward governed agents that do real work inside real systems, with an audit trail. The companies that succeed here will be the ones that can prove an agent followed the rules, used the right data, and can explain what it did afterwards. SAP’s bet is that its decades of running business systems give it an advantage no frontier lab can copy quickly: the context, the permissions and the data model are already there. The space deployment adds a public test of that trust.
Background
SAP is one of the world’s largest enterprise software companies, and its systems run finance, procurement, human resources, supply chain and customer experience across nearly every industry. Joule has grown into a full agent platform, with Joule Work as the interface, Joule Studio as the builder, and SAP Business AI Platform as the foundation underneath. Developer tools built for agents are proliferating across the industry, and SAP is positioning Joule Studio as the enterprise counterpart: a place where agents are built once, governed centrally, and deployed inside the systems where work already happens. Agent platforms for specific business functions are also attracting heavy investment, but SAP’s scale gives it a different starting point: it must convince existing customers to let Joule do more of the work it already does.
Source: SAP TechEd announcement
0 Comments