I have interesting question to add to yo

I have interesting question to add to your bag of tricks from my latest interview for SSAS skills. How you design cubes to support ad-hoc reporting? It was related to discussion on chaging requirements and often hidden questions that users do not articulate ahead of time. I have practiced the concept of not to design generic cubes as well as generic dimensional models. There is no such thing cause the cube answers specific busines questions to help to define business strategies. However, the users want flexibility in analytics and asking more and more for open-ended sort of design inclusive of dosens of dimensions with hierarchical complexity over 2 bln rows facts or more even if more then half of the aggs will not be used but may present potential interest. The manager's support this objective by having the development cost reduction on sort of monster generic "al questions in one cube" deal. What do you think, do we know now with 2008 or 2012 versions the best practice to optimize the design of such monsters?