No Data, No Problem: How to Kickstart an AI-driven Product

The article discusses the challenges faced by product managers in training AI/ML models due to insufficient or poor-quality data. It highlights three viable solutions to overcome this obstacle: starting internal data collection, sourcing data internally or externally, and generating synthetic data. Additionally, it emphasizes the importance of data integration for centralized storage and accessibility by multiple data science teams. The article concludes with a brief overview of the subsequent steps in training an AI model, including selecting an appropriate algorithm, evaluating its performance, and refining it iteratively until it meets the product goals.