Ask about pipeline, bookings, rep comparisons, and retention in the same synthetic scenario as the QBR.
Grounding: Claude translates each question into a structured tool call. The server validates its period, filters, grouping, threshold and ranking, then calculates and formats the answer from the same synthetic records as the QBR. No model-written numerical answer is displayed. There is no live CRM or warehouse connection.
Evidence: Each answer shows its applied query and executed server tool. Questions are independent of earlier chat turns. QBR rollups, rep shortfalls, retention ratios and cohort denominators are calculated in code. Interpretation can still be ambiguous; inspect the applied filters and compare figures with the QBR.
Limits: No legal-review dates, activity history, loss reasons or causal experiments are supplied. Unsupported measures and invalid tool inputs request clarification. Incomplete model responses are never shown as numerical answers. Requests are rate limited; API connectivity is required.