SAP SE today announced a set of artificial intelligence innovations at its annual TechEd conference for developers, enhancing its Joule assistant solutions and Business AI Platform, including tabular data foundation models.
New capabilities in Joule Work will help AI agents become more efficient and interoperate with open protocols. In this case, the Model Context Protocol and Agent 2 Agent allow Joule, the company’s agentic AI assistant, to gather context, make decisions and execute multistep business workflows across the enterprise, while also bringing in third-party data or services.
Way explained that SAP would like customers’ own agents to collaborate with its business agents across company systems. The proposition is that opening those interconnections also requires a controlled entry point where organizations can enforce and oversee activity. This is where a new orchestration layer will come in, enabling assistants and agents to invoke Joule Agents through a governed gateway.
In addition, a recent update will allow agents to trigger n8n workflows, an important connector through Joule Studio. N8n is a node-based workflow automation platform that connects different apps and uses deterministic automation; this means that AI agents can trigger it, but unlike generative AI, n8n uses event-based and trigger-based logic for decision-making. It cannot hallucinate; it is not random and will follow exact rules from start to finish.
That means n8n is suitable for highly regulated or sensitive work that depends on exact needs.
“It needs to do these five steps in this order exactly like this every single time,” Way explained, so in that case, n8n is the go-to.
SAP’s approach combines flexible, natural-language agentic triggers with a prescribed workflow. An agent may help determine when it is needed, while compliance-sensitive procedures can still require an exact sequence of steps.
SAP builds on cloud data to ground the autonomous enterprise
The company today announced the expansion of SAP Business Data Cloud Ontology Manager in SAP Data Cloud, or BDC, to provide an AI-native authoring experience for businesses by giving agents an understanding of how applications relate to one another. It does so by defining relationships between company facts, allowing agents to reason more accurately and make decisions across enterprises.
SAP’s acquisition of Reltio, recently announced at SAP Connect, brings on board multidomain data management to BDC. That helps unify customer, supplier and product records across SAP and third-party systems into trusted data products for applications and agents.
The new product is designed to complement the company’s existing Knowledge Graph, which provides rich business context from SAP data.
Additionally, the company said, it’s expanding its lakehouse capabilities with Dremio, another acquisition, by adding native support for Apache Iceberg and Apache Polaris and a high-performance, federated query engine that enables unified data access without moving or copying data.
Bringing AI models to tables
“Not every enterprise problem is a language problem,” Way said, explaining that although much of the AI approach has focused on natural language, much business data is stored in tables.
To handle this, SAP announced its latest tabular AI model, SAP-RPT-2.0, which expands beyond classification and regression to support anomaly and outlier detection, providing AI-powered analysis through embeddings. It will give multitable capabilities to reduce manual engineering and deliver faster, deeper insights from structured data.
The company also launched a new SAP-RPT-TS model, which will help customers predict values across time-indexed metrics for key forecasting. It uses SAP Business Planning without custom business machine learning, allowing less technical users to deploy the model on their systems.
In a partnership with Hewlett Packard Enterprise Co., SAP announced it’s deploying SAP-RPT on the International Space Station. The aim is to assist with predicting issues before failures escalate, reconstruct missing sensor values from historical data, and generate explainable mission health reports for astronauts and ground teams, all without sending data back to Earth before analysis.
