In this paper I propose a model for situated information processing based on a set of layers, each related to a different processing phase and to a different context type. The model is illustrated with reference to a system for processing semantically rich documents, typical of business and juridical domains. Starting from the conceptual level of document definition, through operations of selection, specialization, instantiation and presentation document are adapted to the situation to which they apply and to the local context in which they are issued.