By Hamza Faisal
SAP on Tuesday unveiled a series of AI innovations at TechEd, framing them as a shift from artificial intelligence that learns a business to AI that actually runs it. Chief Technology Officer Philipp Herzig put it bluntly: this year, SAP is showing how it is moving from AI learning your business to running it.
Joule Work and Joule Studio get agentic upgrades
At the centre of the announcement are Joule Work and Joule Studio, running on the SAP Business AI Platform, which SAP describes as one place to build, contextualise and govern AI across SAP and third-party environments. Joule Work is where users state their intent and get work done across the SAP landscape, while Joule Studio is the development environment, reached through Joule Work, for building and extending enterprise agents.
New agentic thinking capabilities are designed to plan and orchestrate the steps needed to reliably complete complex, multi-step business workflows. Agents gain support for open protocols including Agent2Agent and the Model Context Protocol, and third-party agents will be able to securely invoke Joule Agents through a governed agent gateway. Developers also get new observability controls, a choice of more than 70 AI models, and n8n workflow design for early adopters.
Governance got equal billing. A new SAP Enterprise AI Gateway will centralise management of AI traffic, consolidating authentication, policy enforcement, routing, monitoring, usage tracking and auditing. A new AI Governance Assistant aims to help customers meet compliance requirements under the EU AI Act. SAP also repeated that customer data is not shared to train or improve large language models.
A tabular model trusted enough for orbit
The most eye-catching announcement was also the most technical. SAP unveiled SAP-RPT-2.0, the latest version of its tabular foundation model, which expands beyond classification and regression to support anomaly and outlier detection and richer analysis through embeddings. A companion model, SAP-RPT-TS, will help customers predict values across time-indexed business metrics for forecasting, using the SAP Integrated Business Planning solution without requiring custom machine learning.
Then came the space test. With HPE, SAP announced the deployment of SAP-RPT on the HPE Spaceborne Computer-2 aboard the International Space Station, which it described as one of the most significant edge-compute validations of any enterprise AI model to date. The model is planned to run onboard to triage anomalies, predict issues before failures escalate, reconstruct missing sensor values, and generate explainable mission-health reports, without sending data back to Earth. Herzig framed it as proof of trust: the same model that powers supply chains can also work in space.
Data grounding and a wider partner net
None of this works without data agents can reason over. The SAP Business Data Cloud Ontology Manager will provide an AI-native way to build ontologies across SAP and third-party data, complementing the SAP Knowledge Graph. SAP is also adding native support for Apache Iceberg and Apache Polaris through Dremio, and deepening its Databricks partnership by bringing Lakeflow, Agent Bricks, Unity Gateway and Lakebase into SAP Business Data Cloud as SAP Databricks.
Partner announcements included a collaboration with Intel bridging enterprise workflows and autonomous robots through physical AI, plus extensions with Docusign, Parloa and Redwood for running core processes end to end. For select customers, running custom agents on the SAP-managed Joule Studio runtime will remain free through December 2026. The real test will be whether governed agents can reliably run multi-step processes inside companies. SAP gave no availability date for every item, so some of this remains roadmap, but an orbital trial is an unusually public way to argue its agents are ready for serious work.



