This presentation outlines techniques for developing advanced AI systems, focusing on common challenges and best practices. We cover the AI Test Driven Product Development Lifecycle (experiment - evaluate - deploy - monitor), emphasizing experimentation with prompting techniques, testing various architectures, and maintaining a model agnostic approach. The presentation also explores the emerging paradigm of Agentic Workflows, detailing components such as planning, execution, refinement, and interface design. Key takeaways include identifying low-risk use cases, iterating on models and prompting techniques, implementing robust evaluation methods, and maintaining quality through continuous monitoring and edge case identification.