Prioritizing Tabular Models for Enterprise Data

SAP targets structured data in tables and databases—core to its accounting, HR, and procurement software—over general LLMs, acquiring Prior Labs to fast-track this. Prior Labs' TabPFN open-source models, downloaded over 3 million times, excel at predictions from tabular data. Post-acquisition, SAP invests €1 billion ($1.16B) over 4 years, keeping the lab independent in Freiburg for research speed while integrating via SAP AI Core, Business Data Cloud, and Joule agents. Founders Frank Hutter, Noah Hollmann, and Sauraj Gambhir get a 'healthy exit' with over half a billion dollars cash upfront (undisclosed total), after raising $9.3M pre-seed in Feb 2025 from Balderton. This builds on SAP's prior work like SAP-RPT-1 and investments in Anthropic, Cohere, Aleph Alpha, positioning Prior Labs as Europe's 'frontier AI lab for structured data.' For builders, TFMs offer production-ready alternatives to LLMs for enterprise databases, with SAP promising open-source continuity and product paths.

Controlling Agentic AI Access in Ecosystems

SAP blocks unauthorized AI agents like OpenClaw from its APIs per its policy, restricting to 'SAP-endorsed architectures' like its beta Joule Agents and Nvidia's NemoClaw (built on Nvidia Agent Toolkit, integrated with Joule). This defensive move counters agentic AI threats amid SAP's 2026 stock drop from 'SaaSpocalypse,' unlike Salesforce's open Headless 360 allowing OpenClaw. Builders targeting SAP customers must use endorsed tools to avoid blocks, emphasizing security and scale in enterprise R&D—SAP CFO notes maintaining 'economies of scale advantage' via rapid AI integration.