The ability to adapt old plans to new situations is essential for a planning system. A perceived problem with the adaptation methods in traditional memory-based planning approaches has been the need for a strong domain model in addition to a library of past plans. In this paper, we argue that the process of adaptation does not need any domain knowledge outside of that which a generative planner has available. We present an approach to plan reuse in which information relevant to the adaptation process is left in the form of annotations on generated plans. Adaptation of a past plan to a new problem situation is focused by a process of annotation verification, which locates applicability failures and suggests appropriate refitting strategies. The generative planner is then called upon to carry out these refits. We give examples from automated process planning (in computer-aided manufacturing) to show how annotation verification helps in adaptation. These examples also help in showing how adaptive planning can be an efficient and viable alternative to the generative and semi-automated variant planning techniques currently used in process planning.