The rapid proliferation of generative AI tools like ChatGPT, Bard, Llama, and Anthropic has led to a surge in excitement and confusion. These large language model (LLM)-based tools can be highly effective if they are trained on reliable data. However, the quality of the data used for training these models is crucial as it directly impacts their efficacy and accuracy. Bad or unreliable data can lead to incorrect predictions and outcomes, which could have serious consequences in fields like healthcare and autonomous driving systems. To ensure reliable data for AI, teams should apply data observability and shift-left the data reliability checks. Additionally, enterprises should focus on people, process, and technology when implementing an AI strategy.