As a former product leader at Google AI, building models that "just work" was crucial due to the highly regulated industries they served, where poor or biased predictions came with steep consequences. However, even with high values on vanity metrics, issues with data quality persisted, and it became clear that data analysis and fixing were top problems for many ML leaders. Despite the use of tools like sheets and scripts, which are still state-of-the-art, ML data intelligence is a rapidly maturing space that enables teams to holistically understand and improve data health, removing biases and production mishaps proactively. The five pillars of ML data intelligence include identifying noise/errors quickly, providing a framework for inspecting, analyzing, and fixing data, and enabling automated data health tests and tweaking. This field is critical as it shifts the focus from ML monitoring to ML data intelligence, empowering data scientists with tools to quickly inspect, fix, and track the data they work with, ultimately leading to huge gains in model performance.