When data is 'unlabeled' (meaning the outcomes or 'answers' are not provided), supervised methods like Random Forest (Option D) or XGBoost (Option A) cannot be used. 'Unsupervised learning' is specifically designed to discover 'underlying patterns,' clusters, or latent structures in data without human guidance. For an auditor, unsupervised techniques (like clustering) are invaluable for exploratory analysis, such as grouping similar control failures or identifying unusual transactional behaviors that have not yet been categorized as fraudulent or legitimate.