What is Natural-Language Analytics?
Asking data questions in plain language and getting governed, correct answers back, the model behind Snowflake's CoCo and CoWork.
Natural-Language Analytics Defined
Natural-language analytics lets a business user type "how did in-stock rate trend by region this quarter?" and get an answer, with no SQL and no report builder. On Snowflake, this is what powers CoCo, which answers the question, and CoWork, which acts on it, both part of the broader conversational BI category. The promise is reach; the risk is a wrong number that looks right.
What Keeps Natural-Language Answers Correct?
A governed semantic layer. Turning a sentence into a query requires knowing what each business term means; left to raw tables, the model guesses and produces confident, wrong answers. CoCo and CoWork get theirs from Horizon Context and Semantic Views, governed definitions that lift accuracy on complex enterprise queries from 47% to 83%, the difference between analytics you can act on and text-to-SQL you have to double-check. Extended by AtScale, the same definitions reach every surface CoCo and CoWork don’t: Power BI, Excel, and outside AI agents.
In One Sentence
Natural-language analytics lets anyone ask questions in plain English, and it’s what powers CoCo and CoWork on Snowflake. But a universal semantic layer is what keeps the answers correct everywhere else.
Frequently Asked Questions
Natural-language analytics is the ask-in-plain-English interface; conversational analytics is the broader generate-the-answer experience, of which CoCo and CoWork are Snowflake’s version.
CoCo answers questions. CoWork takes action across multi-step tasks. Both draw on the same governed definitions in Horizon Context.