In the current innovation and entrepreneurship education in colleges and universities, the traditional teaching model lacks diversified learning resources, personalized teaching plans and efficient practical guidance, which makes it more difficult to cultivate students’ innovative thinking and entrepreneurial ability. To solve this problem, this study designs and applies an intelligent tutoring system based on artificial intelligence. First, the BERT (Bidirectional Encoder Representations from Transformers) model is used to build a personalized knowledge graph to accurately describe students’ learning background and interests. Then, the Proximal Policy Optimization (PPO) algorithm is used to optimize resource recommendations to achieve intelligent matching and path optimization of learning content. Then, the GPT-4 (Generative Pre-trained Transformer 4) dialogue generation model is combined for intelligent question answering and online case push to promote students’ innovative thinking and entrepreneurial decision-making ability. Finally, Apache Spark is used to track and evaluate user behavior and quantify the educational effect. As the conclusions said, the intelligent tutoring system has significant advantages in enhancing students’ innovation and entrepreneurship abilities, optimizing personalized learning paths, and improving resource recommendation accuracy. It can effectively support innovation and entrepreneurship education in universities and promote the development of students’ comprehensive abilities.