Current RFID tag readers do not do much more than reading the tags and transmitting that information to a central server for processing. This places a heavy burden on the central server to keep track of all the incoming data from the reader when it can be spending its time more gainfully on other tasks. We propose building some intelligence into RFID tag readers so that they can process some of the data themselves and reduce the load on the central server. We discuss the technical details of building this intelligence in the four key functions of tag readers: (A) Automatic initialization of the tag readers, (B) Automatic identification of misplaced items, (C) Automatic generation of shelf replenishment alerts, and (D) Automatic generation of reorder alerts. In addition, we discuss possible network designs for interconnecting the tag readers and the central transaction server at the retail store. We provide a database model for storing transactions generated by RFID tag reads. We discuss how the current enterprise architectures can be modified to accommodate RFID transactions.