In this paper we present MailOfMine, an approach and a software tool aimed at automatically building a set of workflow models, which represent the artful processes laying behind the knowledge workers' activities, on top of a collection of email messages. The advantages are numerous: the unspecified agile processes that are autonomously used become formalized. Since such models are not defined a priori by experts but rather inferred from real-life scenarios that actually took place, they are guaranteed to respect the true executions (often Business Process Management tools are used to show the discrepancy between the supposed and the concrete workflows). Moreover, such models can be shared, compared, preserved, so that the best practices might be put in evidence from the community of knowledge workers, to the whole business benefit. Finally, an analysis over such processes can be done, so that bottlenecks and delays in actual executions can be found out. In MailOfMine, workflow models are described according to a declarative approach, with a specific visual notation. After presenting the architecture of the system, this paper reports some performance tests and the results of the application of the tool on a real case study.