This paper describes the background to a commercial software service called Threads that was developed to allow the sharing of various message types amongst a community of users – typically all working for the same organisation. Recognising the difficulty that many users have in finding even their own e-mail, a key design goal was to make the sharing and searching of large quantity of disparate messages easier than any existing personal mail client. The project has concentrated on developing a user interface that is easy to use rather than easy to program and, in so doing, has highlighted some interesting user perspectives. This paper discusses the typical user's needs and some of the features Threads uses to meet them. We then introduce a novel intelligent method for organising such messages automatically, according to statistics relating to less common words (called “keywords”) which they contain. Initial experiments using this method are described, using both e-mail data from the parent company, and on the publicly available ENRON dataset of e-mails and phone calls. These preliminary results are interesting, but suggest the method as currently implemented is only suitable for semi-automatic classification, but not yet for fully automated allocations to projects. The paper concludes with a glimpse into the future prospects for message sharing and planned future developments to the Threads system.