- SAP's open AI strategy fosters interoperability with leading AI frameworks.
- An AI agent built with Langraph and AIHub autonomously tests SAP developer tutorials.
- Live demonstration reveals the agent's ability to provide critical, real-time feedback.
At SAP TechEd 2025, a groundbreaking demonstration showcased how an AI agent, leveraging Langraph and AIHub on SAP BTP, can autonomously navigate and test developer tutorials, highlighting SAP's commitment to an open and flexible AI ecosystem.
The session opened by reiterating SAP's strategic embrace of an open developer ecosystem, emphasizing compatibility with various AI agent frameworks. This flexibility, a cornerstone of SAP's AI strategy, was vividly brought to life through a practical challenge: automating the quality testing of tutorials on developers.sap.com. The speaker detailed the essential tools an AI agent would require for such a task, including browser access via Playwright MCP server, terminal access for code manipulation, and the AI Foundation grounding service for SAP-specific information lookups.
The demonstration focused on an AI agent designed to test a tutorial on creating an orchestration workflow in AI Launchpad. The agent's architecture was explained, highlighting the use of the SAP Cloud SDK for AI and a large language model (LLM) as its 'brain.' Crucially, the prompt registry of AIHub was utilized to store the agent's complex prompts, keeping them separate from the main codebase. This setup allowed the LLM to interpret tutorial instructions, create a step-by-step plan, and decide which tools to employ for execution.
The live demo was a testament to the agent's capabilities. As the agent ran, streaming its 'thoughts' in real-time, it autonomously navigated the tutorial, executing each step. A pivotal moment occurred when the agent encountered a new 'grounding module' not covered in the original tutorial. Instead of failing, the agent, initially confused, skipped the module and later provided insightful feedback, recommending that the tutorial be updated to address this new feature.
This ability to not only execute but also critically evaluate and provide actionable feedback underscores the transformative potential of AI agents in developer workflows. The speaker concluded by noting the immediate value for colleagues like Riley, who can now rely on AI agents to pre-scan tutorials, significantly reducing manual effort and enhancing the quality of SAP's developer content. This live, unscripted discovery solidified the message: AI agents on SAP BTP are not just automating tasks; they are intelligently augmenting human capabilities.
“The grounding module confusion... it says it the tutorial should mention whether to disable it or whatever to do with it. Anyway, it's really good feedback. That's exactly what I would have given as well.”
- Nora von Thenen, Developer Advocate, SAP
