In this article we present our approach of intelligent tutoring systems which is based on adaptive workflows. The main idea is to use workflows for learning and assessment process. In fact, our goal is to have a system that could combine the existence of an intelligent tutor to carry out the tasks of teachers in the teaching process and a human tutor who takes part in the monitoring, control and follow-up of the execution of self-learning processes models in order to check out the learner path and to resolve blocking and demotivating situations of the learners. Our model use parameters provided by the adopted learner model especially, student knowledge and preferences, and teaching domain model to propose the optimal pedagogical workflow for each learner to reach his didactical objective. This pedagogical process will be adapted during its execution in the workflow engine by using some parameters of student interaction with the system. We aim also to make the evaluation process more fun and non-stressful by introducing pedagogical games in assessment process in order to motivate students to learn and make the learning process a mean to succeed the game.