This study aims to address the e-inclusion problem that relates to the inclusion of as many individuals as possible to enjoy benefits of information and communication technology. Study contributes to research of e-inclusion in context of blended e-learning. In this paper we propose architecture designed to predict e-inclusion degree of student based on machine learning and intelligent agent approach. We identified two main processes in the e-inclusion prediction system. The first process consists of agent learning activities. Intelligent agents learn the most appropriate algorithm to predict e-inclusion degree of student based on linear regression or cluster analysis. The second process includes activities to predict e-inclusion degree of student. This process covers analysis of e-inclusion risks and communication between student and instructor also. Proposed e-inclusion model consists of goal diagram, use cases diagrams and main algorithms of the system.