Company-to-query

Before the query comes the company.

Text-to-SQL made data feel conversational. But the hard part is rarely the syntax. It is knowing which metric definition, join, exclusion, source, and caveat the company trusts.

01

Business questions are underspecified

A request like “which accounts are ready to expand?” carries hidden assumptions about plan, workspace type, launch windows, event meaning, and audience. A SQL model alone cannot infer which interpretation is safe.

02

The best analysts resolve context first

They check accepted definitions, prior SQL, product changes, source owners, and caveats before writing the query. That judgment is the product surface Uncypher makes reusable.

03

The answer should carry its proof

A number is useful only when the team can inspect the definitions, joins, SQL, sources, and approved assumptions behind it. The proof should travel with the answer, not live in a separate thread.

04

Reusable context is different from chat history

Chat history remembers what was said. Company context remembers what was reviewed, who approved it, which query pattern worked, and which caveat must be attached next time.

05

Data teams still need the control plane

Self-serve only works if data teams can decide what becomes trusted, inspect conflicts, and stop one-off answers from becoming invisible production logic.

06

The category is company-to-query

The next layer after text-to-SQL is translating the company’s operating knowledge into the query path: metrics, entities, joins, filters, product rules, and decisions.