Glossary · AI & ML
Grounded Answers
Grounded answers are AI responses that are backed by verifiable source data–in database tools, answers computed by running a real query against your actual database rather than generated from the model's memory.
An ungrounded AI answer is a guess dressed as a fact: ask a chatbot "what was our churn last month?" and it will invent a plausible number, because it has no access to your data. A grounded answer is different in kind–the AI writes a query, the database executes it, and the number in the response is the number the database returned. Grounding matters because business decisions ride on these numbers; "plausible" is not a standard you can take to a board meeting. The test for grounding is auditability: can you see the query that produced the answer and run it yourself?
In practice
How AI for Database applies it
Fig — every answer ships with the tables, rows and SQL behind it.
See it on your own database
Connect read-only in minutes. Free models included.