This study evaluates the performances of an LSTM network for detecting and extracting the intent and content of com- mands for a financial chatbot. It presents two techniques, sequence to sequence learning and Multi-Task Learning, which might improve on the previous task.