Processing documents is a critical and crucial aspect in an enterprise environment. The management of documents involves several people and many times becomes a long and wasting-time process. Many systems of document workflow have been proposed but usually they are too rigid and complex. Therefore we have developed a document workflow engine based on the email paradigm. When a user wants to make an order, a request of authorization and, in general, any kind of procedure that involve a document, starts her/his request by filling in a form and sending it by attaching it to an email. To this purpose the user has to use our web application that appears as a normal webmail client. Our solution overcomes the actual limitation in the use of document workflow software, especially for what concern the user experience; with our system there is no need, for users, to learn the functioning of a new framework. In addition, users with different roles have different customized view of the document. According with the roles of the users, we trained the system to suggest to the user, at each step, a possible receiver of the email. Currently this feature is based on the fact that the system knows in advance the flow associated with different type of documents. As improvement, we will perform a statistical analysis of interactions between senders and receivers. This analysis will be used to improve the suggestion mechanism: the system will learn the most frequent interactions for each user, depending on the history of previous flow and the document involved. Exploiting these information, the suggestion mechanism will advise to the user the possible receiver of the document.