We developed a method to predict the success probability of innovation in unknown cases by analyzing known business cases using machine learning. This study updates the original method by employing generative AI to automatically convert cases into multidimensional feature vectors, nearly automating the process of analyzing extensive case documents and finding similar cases. By designing an appropriate list of F questions based on innovation factors, case documents are automatically transformed into F-dimensional feature vectors. This approach enables the automatic data conversion of numerous existing business cases, significantly accelerating machine learning analyses such as clustering, classification, similarity search, and outlier detection. We report on the details of this automatic generation technique and the substantial improvement in analysis speed achieved.