This work describes and discusses the implementation of AHP-Tutor, an Intelligent Tutoring System (ITS) that provides training in various methods of derivation of priorities from an AHP comparison matrix. The purpose of our system is to offer exercise problems. The corresponding theoretical knowledge has to be acquired by other means, for example in the classroom, from books or through conventional computer-assisted learning. Based on bibliographical research, five methods were selected for implementation: - the method of the maximal right eigenvalue; - the method of the maximal left eigenvalue; - the method of the mean of the normalized values; - the geometric mean of the rows; - the geometric mean of the columns. A comparative study applying these methods to 500 inconsistent matrices did not reveal any major difference in the classification of alternatives. Based on these findings, a curriculum consisting of three types of exercises was developed. Each exercise uses one of the following three teaching styles: strict guidance, solver, or free discovery. The choice of style was fixed according to the teaching objectives. AHP-Tutor was developed with Visual Prolog 5.2 and runs under MS Windows. The main emphasis is on the expert module, which is the backbone of an intelligent tutoring system. Its explanation component shows the student the steps leading to the solution. It reuses the reasoning of the system and adapts it to the student. This component is domain-independent.