AI in SAP
Where AI pays off in an SAP landscape, and where a deterministic check should stay in charge.
What it covers
- Where AI pays off, and where it does not
- A bounded capability layer around the model
- Joule and BTP AI
- Agents, and what an API policy has to say about them
Typical questions
- Which step in this process is worth a model, and which step must stay a rule?
- How do we keep Joule and BTP AI inside a boundary we can audit?
- What may an agent propose, and what needs a person or a deterministic check before it lands in the system?
How I approach it
The work starts with the return. A capability has to earn its place. AI stays in a bounded layer: it can propose, and people plus deterministic logic commit anything that posts, pays, or changes master data. The ERP keeps the work it already does well.
Example projects
AI tools and data products for the SAP world
Design, development, and operation of my own products: Analygits (opens in a new tab), a Chrome extension for SAP Analytics Cloud with its own SAC syntax MCP server; CoConsultant (opens in a new tab), an AI-powered design partner for clean-core extensions as a web app; and data from the Marktstammdatenregister as an OData feed for SAC.
The AI tools base their answers on documented sources rather than on the model's memory. Details are on the product pages.
Process payment advices automatically with SAP Document AI
Design and implementation of a global SAP Document AI pipeline for processing payment advices, integrated into S/4HANA Public Cloud with a full clean-core approach. Built as a side-by-side extension on BTP, so the digital core stays free of custom code.