The operation of process plants can be supported by assistance systems that rely on engineering models to answer various questions. However, suitable answers can only be computed when the questions posed by the operators using the assistance are compatible with those considered by the engineers developing the models. Such compatibility cannot be taken for granted, because operator cognition and action rarely play a role during plant engineering. How can we determine whether a model is useful for answering particular operator questions? A promising strategy is to apply compatible frameworks for describing the domain concepts captured by the models and the cognitive action representations of operators. To this end, we introduce the concept of abstraction hierarchies that are used both by engineers to model the technical system and by humans to identify their actions. Thus, they provide an opportunity for matching the two. We argue that such matching cannot be performed algorithmically. If humans are to be responsible for the matching, this presupposes that models are equipped with understandable descriptions of their capa-bilities. Accordingly, we argue that model descriptions are a cornerstone of successful model-based operator assistance. We discuss the challenges of this approach and identify directions for future work.