The study on recommender systems always blinds to the profitability difference among customers. In our research, we have extended our previous work of segmenting customers based on their contributions by proposing a hybrid recommendation system which uses multi-agent system, collaborative filtering together to serve the customers with high profitability, while using simplified top-N to generate high efficient recommendations for the lower profitability customers. Also, agent-based recommendation strategy selection architecture has been proposed and examined in the hybrid recommendation system.