At SAP TechEd 2025 in Berlin, Philipp Herzig, chief technology officer of SAP, finished the opening keynote with a flourish in space. The supplier’s tabular data foundation model, RPT-2, is being deployed, with technology from HPE, aboard the International Space Station. RPT-1 [pronounced “rapid one”] was announced at TechEd last year. It functions like a large language model [LLM] that will predict the next word in a sentence, but for structured data, according to SAP. It is used to make predictions for business scenarios, such as delivery delays or payment risk.
“If we can run maintenance predictions in one of the most unforgiving environments, think what this means for our operations here on Earth,” said Herzig. “That’s AI [artificial intelligence] proven under pressure.”
In a blog post, Brenda Brown, the supplier’s chief marketing officer for SAP Business AI Platform and Joule, said: “SAP [has] announced the planned deployment of SAP-RPT, our tabular AI foundation model, on HPE Spaceborne Computer-2 aboard the International Space Station (ISS), one of the most significant edge compute validations of an enterprise AI model to date.”
She further said the model “makes predictions directly from structured data – even with some data missing – and without requiring the usual model training or fine-tuning. Running directly onboard the ISS on HPE Spaceborne Computer-2, it supports anomaly triage in real-time telemetry data, predicts issues before failures escalate, reconstructs missing sensor values based on historical data, and generates explainable mission-health reports for astronauts and ground teams.”
In a press statement, Herzig added: “The same trusted SAP-RPT model that powers global supply chains also can work in space.
“SAP’s deep-rooted, proven expertise has led to the International Space Station’s use of our tabular foundation model to predict potential mission-critical failures in orbit,” he said. “With SAP RPT and TabPFN from [SAP acquisition] Prior Labs, we bring the unique capabilities of tabular AI into the enterprise and across every frontier.”
The broad theme of TechEd is to put some flesh on the bones of the “autonomous enterprise” vision adumbrated at this year’s Sapphire conferences in Florida and Madrid.
Evolution
Florian Kunzke, SAP’s global director of artificial intelligence strategy, said in a pre-conference press briefing with international press: “We will see an evolution of our strategy so far, moving from the slides in our announcements at Sapphire into the productive environments of our customers, which is the hardest thing to do.”
The supplier is announcing enhancements to its Joule Studio and Joule Work technologies, and its Business AI Platform, including the tabular foundation model on board the space station. The supplier describes Joule Work as “a foundational capability within SAP’s AI suite that orchestrates complex business workflows, automates cross-functional tasks, and manages end-to-end operational processes autonomously across enterprise applications”.
SAP is also putting forward new tools to build AI agents and endeavouring to provide a “richer data and business context for those agents to reason with, and strengthened governance controls to run them responsibly across the enterprise”.
Herzig said, in the main statement for the event: “SAP TechEd has always been where builders shape what comes next,” he said, adding that this was the 30th event of its kind. “This year, we are showing how we are moving from AI learning your business to running it. Joule Work is where you put your intent and get work done, in one experience across the entire SAP landscape and beyond. That is the shift we are delivering at scale.
“Everything underneath runs on SAP Business AI Platform, delivering faster time to value, business understanding, higher accuracy, enterprise-readiness that is secure, governed and compliant, and a unique context layer built on over 50 years of process and industry expertise, encoded into every agent action.”
At last week’s SAP AI Tour event in Birmingham, Manos Raptopoulos, global president of customer success, said in the keynote session: “AI means technology is now back in the boardroom, but is it delivering business value? When LLMs can’t do Arithmetic, 90% accuracy is not good enough for businesses.”
He argued that grounding SAP’s AI technology Joule in a “context layer” made possible by customers’ use of the supplier’s enterprise resource planning over many years “reduces hallucinations by 89%”, adding that specific business knowledge, semantic richness and governance are what AI needs to deliver business value for SAP customers.
In an interview with Computer Weekly, Raptopoulos said that when it comes to getting business value from AI, “the first layer is the intelligence layer, which is your generative AI model, LLM, call it whatever. There is another layer, which is the context layer, and context here is twofold. One is the context of the system – when it comes at least to our own system, it’s the Knowedge Graph, as we call it, or the ontology, as some other people call it, but it’s also the context of the company, which we call company memory in terms of, “So how do we do things around here?”
“To get to what we call the autonomous enterprise, every customer is at a point A, and they have to go to point B,” he said. “There are certain customers that are more advanced on that journey, and every customer is somehow unique, but they are very much aware of that journey, of that transformation that they need to do.”
At TechEd, the supplier is announcing capabilities for Joule Work, said to help make AI agents be more effective at handling complex business processes via open protocols including Agent2Agent (A2A) and Model Context Protocol (MCP).
It talks about “agentic thinking capabilities that plan and orchestrate the steps assistants and agents need to reliably complete complex, multi-step business workflows across SAP landscapes”.
The supplier is also trumpeting master data management capabilities provided by recent acquisition Reltio and data lake capabilities proffered by another recent acquisition, Dremio. That is said to have native support for Apache Iceberg and Apache Polaris, plus a high-performance, federated query engine that will enable unified data access without moving or copying the data.
It is also announcing an expansion of its relationship with Databricks, whose CEO, Ali Ghodsi, had a message for the conference reaffirming its work with SAP.
