Ad hoc business questions answered in real time instead of 2 days, management meeting prep down 80%

Business analysis at the group ran on a dozen or so fixed reports. When an executive wanted to see, say, the gross margin trend for one category in one region, they waited two days for the data team. Business units defined metrics differently, so the same KPI showed different numbers in different meetings.

The group was moving to data-driven management. It wanted metric definitions unified first, and then executives able to question the data directly.

Governance first, then intelligence: the FDE team spent 6 weeks with finance and the data team building a business semantic layer, then launched the business Q&A Agent.

• The semantic layer defines the calculation, source and access rights for 120 core business metrics and is the single trusted definition
• The Q&A Agent turns natural-language questions into queries against the semantic layer and returns charts, figures and the metric definition
• Anomaly explanation: when a metric moves, likely causes and drill-down paths are suggested automatically
• Management meeting packs are pre-generated by the Agent from templates; analysts only review and add judgment

The semantic layer sits on top of the existing data warehouse. The model only generates queries against the semantic layer and never touches the underlying tables. Row-level permissions are pushed down with the user's identity, so executives and business units see numbers from the same source.