The shift to conversational analytics

What changes, and what does not

The pipeline upstream stays the same. Source systems are unchanged. The ingestion layer is unchanged. The warehouse, with its joins and aggregations, is unchanged. What changes is the surface on which a question gets asked and an answer gets read.

In the conversational model an executive types a question. An agent reads from the warehouse, picks the relevant tables, writes a query, renders a chart, and returns an answer. The chart did not exist before the question was asked. It will not exist after the answer is read. The artifact is generated per question and not retained.

The shift is happening for reasons the dashboard model earned. Dashboards proliferate. They go stale. They answer the questions someone asked last year rather than the question being asked now. A conversational interface that can produce a fresh chart in seconds eliminates a category of friction that the dashboard model accepted as the cost of doing business.

The three properties go away at once

The three properties that made the analyst's quality gate possible all change at once.

• Persistence is gone.The artifact is ephemeral. The chart exists for the length of the question. There is no shared page to bookmark, audit, or curate.
• Bounded questions are gone.The agent can join, filter, and slice in ways no dashboard would have been built to support. The space of possible questions is the space of valid SQL against the warehouse.
• Human cadence is gone.Answers arrive in seconds. A single conversation can produce ten answers in an hour. There is no walk-through-the-data slot between question and consumption.

None of this is a critique of conversational analytics. The shift offers something the dashboard model could not offer. The point is narrower: the conditions that made the last-mile quality gate possible do not survive the shift. Whatever the new model does better, it does not do this.