A planner at an aviation-services organization used Claude Code to build a direct-booking dashboard, selecting roughly 10% of nearly 100 tables. Elapsed work periods were one week for analysis, three days for design, two for implementation and two weeks for validation alongside regular duties. Joins that ignored one-to-many relationships produced incorrect totals, so the planner specified primary keys and aggregation units and compared results with independent queries. Work used a read-only environment, while internal deployment and permissions required responsible teams. AI assisted implementation, not accountability for data meaning and access control.