Decision-making and prediction using Case-Based Reasoning involve two sequential steps: retrieval, where similar cases are selected, and adaptation, where these retrieved cases are used to infer a solution to the target problem. Traditionally, both steps are performed by a single agent, and prior research has emphasized the importance of adaptation-guided retrieval, i.e. considering the adaptation method when conducting retrieval. This paper explores an alternative setting, in which retrieval and adaptation are carried out by two distinct agents. A particularly relevant scenario arises when retrieval is performed by an AI agent, while adaptation is handled by a human. Since adaptation is conduced externally, adaptation-guided retrieval is only feasible if there is a model of how adaptation is performed. Two scenarios are examined: (1) when the retrieval agent knows the target solution and seeks to guide the adaptation toward it, and (2) when the retrieval agent does not have access to the target solution. Our approach is evaluated with a series of experiments on both a symbolic and a numerical task, using models of varying complexity. The results highlight the importance of inferring a correct model of the adaptation.