Every AI project starts with the same architecture conversation. Which steps should an agent take on, and which should stay with rules, workflows or people? Which systems does the solution need to read from and act on, and under whose identity? What can be built with what the customer already has, and where is the North Star? Getting that conversation right early saves weeks later. It is also exactly where SAP's new AI Native Architecture Advisor helps.
In this blog post, I explain what the Advisor is and how it works, then walk through four scenarios where architects and developers can put it to work. I tested one of them in detail: a dispute-to-credit process that spans SAP S/4HANA on-premise, Salesforce, a logistics partner's portal and SAP BTP. I close with tips for better results and the limits to keep in mind.
In short
- The AI Native Architecture Advisor is an open-source skill that turns a business problem, a process model or an existing design into an initial AI-native solution design, grounded in SAP's own architecture guidance.
- It runs in Claude Code and OpenAI Codex, where it writes a solution brief, a reasoning log and a visual HTML report, and in Joule Work Desktop, where it delivers the brief as Joule cards.
- Use it to shape new AI use cases, review existing solutions, reimagine multi-stack processes and design reusable partner agents. Treat every result as a draft for an architect to review.
