This article presents the research and development of a question-and-answer system based on the BERT model for the Kazakh language. The article presents a new combination of normative and statistical approaches to the analysis of questions used in the system of answering questions of a closed subject area for agglutinative languages ​​such as Kazakh. Question analysis consists of focusing and classifying questions. To focus, we have several experts based on the rules of the Kazakh language FAQ. BERT is undoubtedly a useful model in using machine learning for natural language processing. It provides fast and accurate set-up and can be used in many practical and subsequent applications.