Creating AI-workloads seems easy, when you view it through the looking glass of the web interface/app of Your 'go-to' AI tool like ChatGPT, Copilot, Claude ... But what it means to implement design methodology and principles for AI workloads on Azure? We are going to do a walk trough of the newly established principles of the Well-Architected Framework, and how it is different for what we already know about Azure workloads. And then, we will talk about practical solutions for common mistakes in AI Workload Architecture and what to do about them.