Prompt Engineering neu denken: Wie strukturierte Daten KI verbessern | INSIDE AI #27

Effective AI usage begins with structure. In this episode of INSIDE AI, Tommy Falkowski explains how structured prompts, data formats such as JSON or YAML, and automated processes lead to better results. With practical examples and tools for AI integration in companies. 📺 Subscribe to our YouTube channel: @FraunhoferIEM 🔗 LinkedIn:   / fraunhofer-iem   📸 Instagram:   / fraunhofer.iem   📩 Newsletter: https://www.iem.fraunhofer.de/newsletter Overview 0:00 – Introduction: Thinking in data with AI 1:19 – Understanding language models as black boxes 2:16 – From prompt engineering to RAG 3:12 – Structuring context and inputs correctly 5:18 – Defining structured outputs 6:25 – Comparability through structured results 8:50 – What does “thinking in data” mean in concrete terms? 10:09 – Introduction to JSON, YAML & TOML 11:19 – Example tool for visualizing data structures 13:12 – Use of structured data in business processes 15:16 – Advantages of data validation through schemas 17:06 – Automated extraction & transformation 18:45 – Integration into ERP and CRM systems 19:53 – Structured output mode at OpenAI, Anthropic, and others 21:08 – Prompt optimization with DS-Pi 23:00 – Grammar constraints and deterministic output 25:19 – Conclusion: Thinking about and automating processes in a structured way