Best practices frameworks such as ITIL provide a generic description of best practice processes that are intended to be followed by people. These processes are refined into more concrete steps before they are actionable. The refinement often is specific to the organization where the process is adopted, as well as people who are enacting the process. Modeling best practice processes is challenging. On one hand, these processes need a high-level, abstract representation. Current process modeling languages are too rigid for modeling them. On the other hand, automation of the enactment of these processes among people requires formal models. In this paper, we propose a framework for modeling best practice processes at three levels: user-level, formal process model level and machine representation level to support the collaborative and ad-hoc refinement of process models as well as the automation of their enactments. We also propose an approach to learn from the past enactments of processes to enable reuse of organizational domain knowledge.