Undoubtedly, suitable information retrieval from internet and other large and very large scale business data sources is one of the most important problems in efficient use of information sources. Nowadays, web is the largest data source of documents and other forms of information and a suitable ground for evaluating the different Information retrieval techniques. The more the web is expanded, more the need for powerful search tools become evident. At the present time, there are lots of services for web search, but none of them are helpful as expected and actually in the most cases the results are dissatisfactory. Unspecialized searching tools are one of the most important reasons for this problem. So, orientation toward more specialized tools with learning capabilities is the natural solution for this problem. In this study on the same direction we try to design, implement and examine a complete architecture for a customized software intelligent agent which is able to retrieve commercial information from multiple sources based on user interests. The designed architecture is able to use information fusion methods and adaptive self-correction mechanisms to achieve this goal.