Yes. Semantic layer designs ~80–85% of common business questions accurately. The remaining 15–20% are edge cases that include complex analysis, unusual combinations, one-off requests. GenBI should route these intelligently: either (1) escalate to analysts (“I don’t know how to answer that; please contact analytics team”), or (2) enable power users (advanced query builder or direct SQL access for trusted analysts). The semantic layer improves over time: as you see questions fail, you refine the layer to handle them. We monitor question failures monthly and recommend semantic layer updates (Month 3, Month 6, Month 9 reviews).