With the growth and commercialization of the Internet, many busy people find themselves receiving more e-mail than ever before. Thus, e-mail filtering has become an important problem. Many mailers provide filtering based on the mail header: the sender, the domain of the sender, and the subject. In the first phase of the AlterEgo project, we are developing an e-mail filtering agent that uses the content of the whole message, plus a model of the world and a model of the user, including their important goals. Initially, Case-Based Reasoning seemed llke a promising master approach; messages that are similar to messages that were deemed important in the past should be important now. However, since the user’s world may often change significantly and sometimes suddenly, the relevant similarity to a past message may be at a very abstract level. This will result in an expensive search for similarity. Hence, in our initial prototype, we have developed an approach that is driven by model-based reasoning, and is aided by case-based reasoning.