Start with the decision, then choose the interface
A dashboard is useful for a recurring operational view: department performance, pipeline health or budget against actuals. An ad-hoc question needs a different interaction. Both depend on the same thing: data people can trust and definitions they understand.
A conversational interface does not fix conflicting definitions of revenue or duplicated customers. Establish the business model before adding a new way to query it.
What is a semantic layer?
A semantic layer describes business concepts—customers, orders, measures and dimensions—over the underlying data. It helps different tools ask questions using consistent definitions instead of recreating calculations independently.
AgentData provides a governed business model and query interfaces for this purpose. It can sit alongside dashboards and scheduled reports, with natural-language access where that helps a user explore.
An illustrative question across systems
A customer-success team might ask which accounts have an approaching renewal, an unpaid invoice and an unresolved support issue. Answering requires consistent account identifiers across CRM, finance and support, agreed time windows and appropriate access permissions.
First reconcile the records and definitions. Then decide whether this belongs in a weekly report, a dashboard or an interactive question. This is an illustrative workflow, not a reported customer deployment.
Be precise about what an AI client receives
Query planning can operate on metadata and a question while execution happens against your data sources. Query results are a separate boundary. An external assistant connected to a query interface may receive authorised results, depending on its permissions and configuration.
Review the entire flow—including the consuming application—before deciding whether it meets your data-handling requirements. A locally executed query does not by itself mean every downstream result remains local.
Keep the reporting tools that earn their place
Use dashboards for recurring monitoring, scheduled reports for controlled distribution and conversational tools for exploration. A shared data model helps those interfaces agree. Explore reporting and analytics delivery or learn about AgentData.
Bring us the data problem.
We can help scope a practical first step, from an unreliable report to a system integration or data model.
Discuss your data project