This paper introduces the methods to build intelligent multimodal assistants for medical purposes. In the experimental implementation of the bot, the authors use logistic regression and simple word embeddings models. But the experimental implementation shows the ability to combine empathic conversation with AI (without a specific goal) and functionality (satisfying the conceptual intent of the user) to maintain the patient's motivation to use such applications for a quick and useful response from the clinic. In the future, it is planned to develop a universal chatbot for health monitoring and routine communication with patients based on insights received from users.