Old Knowledge, New AI: SAP's Strategy for the Autonomous Enterprise

Contents

SAP is using its recently launched customer event Sapphire to present the answer to an AI-driven software world, eagerly awaited especially by the financial sector. Little surprisingly, the target vision is the Autonomous Enterprise, where AI-driven assistants work hand-in-hand with ordinary staff and completely "take over" business processes.

The newly introduced Autonomous Suite automates the corresponding end-to-end processes. As SAP management likes to emphasize frequently, the work and execution here are based on the contextual knowledge of business operations gathered over fifty years since SAP's founding.

Knowledge Graph as the Semantic Heart

The technical foundation for the new suite is the also newly introduced Business AI Platform. This is less of a real new development and more primarily the consolidation of the Business Technology Platform (BTP), Business Data Cloud (BDC), and Business AI (Joule, Joule Agents, AI Foundation) into a regulated environment.

The semantic heart is formed by the Knowledge Graph, which SAP has also been working on for a long time. This provides agents with a structured overview of all business units, processes, and relationships within the entire SAP landscape of a user company. SAP has now created a series of domain models that have been trained using program code, data, metadata, business processes, architectural knowledge, and documentation. In combination with context graphs, these are intended to feed the inherent logic and control logic of the SAP application world to the new agents and assistants.

Let the AI Agents Handle It

At Sapphire, SAP is also introducing Joule Work, a new dynamic workspace that supports intent-driven work. Users are meant to simply articulate in natural language what they want to achieve and delegate the rest to the AI or the Joule assistant. The assistants recognize user intentions based on context and coordinate the corresponding agents throughout the company. The Work environment is designed to be interoperable, serving as a central interface between SAP and non-SAP environments.

Through Joule Work, users can also access the Studio environment. Here, they can create their own agents, which can also access third-party tools or agents via the Model Context Protocol (MCP) and the Agent2Agent Protocol (A2A).

For the implementation of agent-based functions, SAP does not rely solely on in-house developments but continues to use tools from relevant providers; for example, Anthropic's Claude will support Joule agents in the areas of human resources, procurement, and supply chain. With AWS, data integrations will be possible in the future without having to shuffle data back and forth between Athena and BDC.

Gaps Closed Through Acquisitions

Nvidia's OpenShell, in turn, serves as a trusted, secure runtime environment for Joule Studio to create, deploy, and manage any agents within the SAP Business AI Platform. Furthermore, acquisitions are closing several functional gaps in the offering. The Reltio acquisition, for example, is intended to consolidate data from different sources based on graphs into a consistent view of master data ("Golden Record"). The planned acquisition of Dremio accelerates the path of BDC, SAP, and non-SAP data into an Apache Iceberg-native Lakehouse.

The simultaneously announced acquisition of Prior Labs, a pioneer in Tabular Foundation Models (TFMs), supports the development of tabular foundation models for business tasks. Since enterprise applications typically organize data and numbers in tables, large language models literally lack fundamental understanding in this area. With RPT (Relational Pretrained Transformer), SAP itself has been offering its own product in this area for several months, which predicts things like payment delays, upselling opportunities, etc., based on tabular ABAP data. In the new 1.5 version, this now receives a chat assistant to further explore tabular data and the underlying results using natural language questions.

All activities, large and small, contribute to SAP maintaining control over the accumulated business and process expertise through the Business AI Platform, even in an AI-driven software world. The newly introduced Autonomous Suite is said to already comprise more than 50 domain-specific Joule assistants in the areas of finance, supply chain, procurement, human resources, and customer loyalty. These automate the corresponding end-to-end processes by coordinating over 200 specialized agents.

In addition, SAP has introduced eight autonomous industry solutions that handle industry-specific processes from start to finish, integrating their specific logic and regulatory requirements. SAP's policy of making AI offerings exclusively available to users of cloud-based enterprise programs is also being somewhat softened. They will also be accessible in the future to users with traditional on-premises platforms – provided they have started implementing or migrating their applications to the cloud and have concluded a corresponding (Rise) plan with SAP.

(vbr)