Advances in Artificial Intelligence (AI) have made its integration into the educational domain easy, with the improvement of Intelligent Tutoring Systems (ITS): adaptive computer systems capable of supporting (during the learning process) students in general, but particularly those with Specific Learning Disorders (SLD) such as dyslexia. The use of ITS has shown satisfactory results in many disciplines, but in the field of music it remains largely unexplored. In this paper, the concept of personalized adaptive learning is addressed in order to design and implement an AI-Tutor to support the individual study of the (dyslexic and non-dyslexic) student in the field of music composition. Taking advantage of Markov’s chains, a system was created that could note down (for each individual student) recurring errors in the harmonization of a musical bass line, and then suggest some reinforcement exercises to help overcome the difficulties: exercises generated by the system independently on the basis of the student’s mistakes, and not previously prepared by the teacher. The results showed that the AI-Tutor can be a useful tool to engage and motivate the student in the study of musical composition, thanks also to the personalized feedback proposed during the performance of the reinforcement exercises.