As the amount of information provided to the customer becomes larger, unnecessary information can lead to difficulty in finding the information customer wanted. This also may lead to a customer being dissatisfied with the service. However, the existing systems that apply a similar concept fail to address the individual user?s demands. In addition, in considering similarities between users, each item?s relative importance (weighted value) to each customer are not being taken into account. This leads to problems of scarcity if the common preference items are small and also the problems of expansion, where system slowdown occurs as number of users increase. These inefficiency problems will be dealt with, in this paper, an adaptive e-commerce agent system is proposed to cater for individual user?s taste for products. This system includes a monitoring agent that monitors user?s intentions, a similarity referencing agent that learns user?s activities to reference a group with a similar pattern, and an interest extraction agent that creates and updates individual user?s activity database whenever change in activity is detected.