Tactical networks are highly dynamic environments characterized by constrained resources, limited bandwidth, and intermittent connectivity. The limits on communication cause significant delays in the delivery of information to edge users. This paper focuses on an approach to improve the timeliness of access to information via prediction and pre-staging. The approach also incorporates a learning mechanism to dynamically adapt the information prediction algorithm. This capability has been integrated into the DisService peer-to-peer information dissemination system, which opportunistically exploits any available connectivity to address the challenging environment. The extended system, called DisServicePro (for Proactive) predicts the information needs of edge users using their mission description, including the routes that users may take as part of the mission. DisServicePro extends the capabilities of DisService by efficiently and proactively disseminating information to the edge nodes by means of replication and forwarding policies. The proactive behavior is the result of the integration of policies and a distributed learning algorithm that takes into account the history of previously requested information, along with the characteristics of the target nodes and the mission. As new information becomes available, DisServicePro matches it against the mission profile and pushes relevant information to the edge nodes. Information that is selected to be pushed is sorted based on the predicted time to use as well as the confidence value of the prediction.