Enterprise Architecture (EA) modeling languages are increasingly used for various enterprise wide analyses. In most cases one needs to adapt EA languages to an appropriate level of detail. However such an adaptation is not straightforward. Language engineers currently deal with analysis driven language adaptation in an ad-hoc manner, adapting languages from scratch. This introduces various problems, such as a tendency to add uninteresting and/or unnecessary details to languages, while important enterprise details are not documented. Moreover, adding detail increases the complexity of languages, which in turn inhibits a language's communication capabilities. Yet experience from practice shows that architects often are communicators, next to analysts. As a result, one needs to find a balance between a model's communication and analysis capabilities. In this position paper we argue for an approach for assisting language engineers in adapting, in a controlled manner, EA languages for model-driven enterprise analyses. Furthermore, we present the key ingredients of such an approach, and use these as a starting point for a research outlook.