Natural Language Query

A natural language query is a question asked of company data in ordinary words, which the system translates into the query that answers it.

The translation is the easy half. Turning "how many enterprise accounts renewed last quarter" into valid SQL is something current models do well. Knowing which table holds accounts, what this company counts as enterprise, when its quarter ends and which of three date columns means renewal is the half that decides whether the answer is right.

This is why natural language querying is a context problem wearing a syntax problem’s clothes. A system with the schema but not the definitions produces fluent, confident, wrong numbers — the worst kind of failure, because nothing about the answer signals that it should be checked. Showing the query it ran is the minimum defence.

Frequently Asked Questions

For the long tail of one-off questions, largely. For the definitions those questions rest on, no: someone still has to establish what the company means by its own metrics.

Ask the system to show the query it ran and the source it ran against. An answer you cannot trace is a claim rather than a result.

Only what their permissions already allow. A permissions-aware system answers from the data the person could have queried themselves.