In digital age, the application of information technology to life has become popular in almost every field in life and the fashion field in particular, and is also one of the things that researchers are paying attention to it. In this work, we present an information system approach that integrates machine learning algorithms to handle customer queries by applying AI techniques, deep learning and content-based image processing. It can analyze customer questions about clothing, predict customer intentions and predict conversation scenarios to provide appropriate advice. For natural language processing module, we conducted a survey on a data set collected from the Internet, mainly e-commerce sites about apparel in Vietnam. In image processing, we experimented on the DeepFashion data set with 52,712 images of various types of clothing using the FashionNet algorithm based on CNN architecture, the following results: Category classification: 82.8%; Landmarks detection: 69.1%; Attribute classification: 97.6%. The research showed 85% response accuracy and 72% entity and value recognition accuracy.