Chinese students’ oral practice after class often falls into two dilemmas. First, there are few real contexts available. Second, the tools used at present lack interactivity, and the feedback is relatively simple. In view of this, this study uses communicative competence theory and output hypothesis as theoretical support, and uses theoretical analysis and system design to give a practice mode that integrates an AI situational coach and a real language partner community. In this mode, the AI coach can provide approximate, real, and repeatable training scene dialogues and give analytical feedback from a large number of perspectives, so as to help learners remember language knowledge. The real community can give real and various social interactions, so as to further improve the learners’ pragmatic and strategic abilities. From the analysis results, these two parts can cooperate with each other through the data cycle, and after they are combined into one, they can comprehensively strengthen the learners’ oral communication ability.