A schema tells an agent what exists. It doesn’t say that "active" means status = 2 AND churned_at IS NULL, or what a "settlement" actually is.
Text-to-SQL failures are rarely syntax. They’re semantics: the agent picked a plausible column, joined a plausible table, and answered a different question than the one asked — because the business meaning behind the schema was never available to it.
BrainBox connects your SQL schema and the supporting documents — definitions, policies, metric docs — into one brain. When the agent writes a query, it first resolves the business terms: what "active" means, which table backs Customer, which rule governs the metric. Vocabulary in, guesswork out.
A text-to-SQL company uses BrainBox to give their structured-data agent the unstructured business context — definitions, rules, vocabulary — that the schema alone doesn’t carry.
Grounded answers, not guesses — from a living model of your business.