Organizations face overwhelming volumes of repetitive conversations from customers, employees, tenants and patients. This increases support workload, slows response times and makes scaling costly. Traditional chatbots typically fail — not because the AI isn’t smart enough, but because they lack context understanding, workflow automation and deep integration with operational systems like SAP.

In this session I share how we build AI voice and messaging agents that materially reduce support and service workload.
I’ll show why building effective agents is like an iceberg: the visible part — the LLM, speech-to-text and text-to-speech — is only 10%. The real work lies beneath the surface: orchestration, system integration, context management, knowledge retrieval, testing and evaluation, security, compliance and continuous optimization.

Drawing from real-world implementations across various sectors, I’ll walk you through the full agent lifecycle — from design to production — and explain what it truly takes to deploy AI agents that deliver measurable results.

SAP HANA has supported machine learning for years via native predictive libraries and specialized clients, establishing a robust foundation for intelligent data applications. Now, SAP HANA Cloud extends this with tools for today’s AI needs — such as vector support, knowledge graphs, low-level NLP, and LangChain integrations.

This session dives into these features in a straightforward, engaging, and entertaining way, no hype attached.

This session shows how Hugging Face models can be used inside SAP AI Foundation to build real, production-ready AI use cases. You will learn how to bring your own models into the SAP ecosystem, deploy them with SAP AI Core, manage them via AI Launchpad, and connect them to SAP applications.

The focus is on practical patterns, not theory. Topics include model selection, containerization, training and serving workflows, and governance aspects like versioning and monitoring. The session also shares lessons learned from real projects, including common pitfalls, performance considerations, and cost awareness. You will see how open-source transformer models can be combined with SAP services to accelerate innovation while staying enterprise-ready. This talk is aimed at developers, data scientists, and architects who want to use modern NLP and foundation models in SAP landscapes. Basic knowledge of SAP BTP and machine learning helps, but the session stays accessible and hands-on.

By the end, you will understand how to bridge Hugging Face and SAP AI Foundation in a clean, scalable way and how to start your own BYOM journey in SAP.

With exciting updates to building blocks, a streamlined entry page, new AI runtime time functionality and powerful generative AI in SAP Fiori tools, developers can build and use applications faster and more flexible than ever. Get a first look at how we integrate MCP to orchestrate AI capabilities across SAP Fiori tools and SAP Fiori elements for faster app development.

Think your AI assistant is buggy? Think again…

In this session, I’ll explore why most “AI failures” are actually prompt failures, and how to fix them. I will show practical techniques for getting consistent, high-quality results from AI when tackling real (SAP) development challenges: applying TDD and SDD principles with AI assistance, refactoring legacy code without breaking everything, detecting malicious patterns before they hit production, etc. Through live demos, I’ll share how to write prompts and instructions that work like well-crafted requirements: specific, testable, and unambiguous.

Walk away with a systematic approach to AI-assisted development that turns frustrating trial-and-error into predictable, production-ready results. No more “the AI didn’t understand me.” Just better specifications, better code, and better outcomes.

SAP Joule requires S/4HANA private cloud, but most companies still run on-premise or ECC. This session shows how MCP servers on SAP BTP securely connect any SAP system to modern AI platforms via Cloud Connector with principal propagation. A live demo highlights querying SAP data through an AI agent, covering architecture, security, and implementation steps.

Part 1
At last year’s ABAPConf, we presented N8N as an alternative tool for AI use cases within the SAP ecosystem. We used it to access the SAP system directly and read all the development objects (classes, tables, structures, etc.) to automatically create an technical documentation.

In this session, we will show another use case in which we translate the texts of the development objects into any target language. This can save a lot of time because translating using SAP GUI or Eclipse often requires countless clicks.