Generative AI has moved beyond simple calculations and is now capable of creating and imagining things, but evaluating these systems poses new challenges. Assessing Generative AI requires a different approach than traditional software evaluation methods, as it involves evaluating the quality of generated content rather than just its functionality. This demands the use of effective AI evaluation methods that delve deeper into what the AI is producing, such as metrics for how well the prompts are used and how to choose the right model and vector stores. The process also requires a deep understanding of the specific AI application being worked with, as well as the ability to set up clear guidelines and robust logging systems to track every step the agent takes. Additionally, human oversight is crucial in evaluating AI systems, particularly in high-stakes areas such as healthcare or finance, where accuracy matters. To tackle AI hallucinations, which occur when models generate responses that don't match the input data or real-world context, a layered approach is needed, including defining clear contexts, integrating quality data, and maintaining human oversight. Ultimately, evaluating Generative AI requires a tailored approach that takes into account the unique challenges posed by these systems.