Interrogable business intelligence

A dashboard answers the question it was designed for. A good analyst answers the follow-up.

That is the useful wedge for AI inside BI. The work is making the business intelligence layer answer questions that were not known when the dashboard was designed.

This matters most when the data is operational. A lender needs more than the fact that PAR30 moved. They need to know whether the movement is concentrated by product, vintage, cohort, channel, or data freshness. The fixed dashboard can show the signal. The analyst layer helps decide what to check next.

What makes it real

Interrogable BI needs a governed data layer underneath it. The agent has to inspect schema, use canonical metric definitions, query live data, explain freshness, and say when the dataset shows movement without showing cause.

The trust story matters as much as the chat experience. If the system can invent a number, the product is broken. If the system can trace the number to a query, a snapshot date, a filter, and a metric definition, it starts to become useful.

The product shape

The first pass is conversational: ask a question, get an answer, inspect the evidence.

The stronger version turns useful questions into reusable work. A portfolio-review question can become a saved check. A weekly risk question can become a scheduled artifact. A chart can either stay live for ongoing monitoring or freeze because it supported a decision on a specific day.

The forward-looking version is Scenario Planning Is Decision Rehearsal.
A team asks what would happen if delinquency rolls forward, if collections cures
part of a bucket, or if a product is paused. That still belongs near BI because
the scenario starts from the live book. It just needs the assumptions stated
more carefully.

That is when "chat with a database" becomes an analyst workflow.

Notto example

Notto AI Agent is moving in this direction. It plans analyses before SQL, verifies SQL against intent, runs through read-only access, tracks execution traces, verifies final answers, and can create artifacts, reports, and recurring BI checks.

The lesson travels beyond Notto: dashboards are fixed interfaces, and business questions keep moving. The useful product is the layer that can move with them without losing the discipline of governed BI.