The document discusses using deep learning decision making with robotic process automation (RPA). It proposes configuring known features and an unknown business logic as inputs for a deep learning model in the RPA environment. The model would be trained on a dataset generated by crawling sample data 1,000 times without preprocessing. The trained model could then be used by an RPA bot to make automated decisions by requesting data, getting a response from the model, and taking action accordingly. Code requirements and files for implementing this deep learning decision making RPA are also listed.