SnowPro Specialty: Gen AI Question 62
Single answerCortex AnalystA retail company is building a natural-language analytics assistant for business users. The team wants to use Cortex Analyst so users can ask questions such as "What were online sales in the Northeast last quarter?" The data model includes fact tables for sales and returns, several dimensions, and business-specific metric definitions. During testing, the team notices that similar questions sometimes produce inconsistent SQL because analysts use different terms such as "revenue," "sales," and "net sales." Which action should the team take to improve the reliability and business accuracy of Cortex Analyst responses?
- A
Create and maintain a semantic model that defines trusted metrics, relationships, dimensions, and business synonyms for terms used in natural-language questions.
- B
Increase the size of the virtual warehouse used for the underlying queries so Cortex Analyst can infer business meaning more consistently.
- C
Replace Cortex Analyst with Cortex Search so the system can retrieve matching rows directly instead of generating SQL from a semantic model.
- D
Store a larger volume of historical sales data in Snowflake so Cortex Analyst has more examples to learn metric definitions automatically.
Show answer and explanation
Correct answer: A
Explanation
The best answer is to create and refine the semantic model used by Cortex Analyst. In real implementations, the main source of inconsistent natural-language analytics results is ambiguity in business terminology, metric logic, and table relationships. Cortex Analyst is intended to operate with a semantic layer that defines measures, dimensions, relationships, and synonyms so generated SQL aligns with trusted business definitions. Best practice is to encode approved KPI definitions and alternate user phrasing directly in that model rather than trying to solve the issue with more compute or more raw data. Snowflake documentation for Cortex Analyst emphasizes the role of semantic models in grounding natural-language questions to structured data and producing reliable SQL for analytics use cases.
- A. Correct.
Correct. Cortex Analyst is designed to translate natural-language questions into SQL using a semantic model that captures business context, including metrics, dimensions, joins/relationships, and synonyms. If users ask for "revenue," "sales," or "net sales," the right way to improve consistency is to encode those definitions and terms in the semantic model so the generated SQL reflects the organization's approved business logic.
- B. Incorrect.
Incorrect. Warehouse size can improve execution performance for queries after SQL has been generated, but it does not improve Cortex Analyst's understanding of business terminology or resolve ambiguity in metric definitions. This option reflects the common misconception that compute tuning fixes semantic interpretation problems.
- C. Incorrect.
Incorrect. Cortex Search is intended for retrieval over indexed text or data for relevant results, not for governed natural-language-to-SQL analytics grounded in a business semantic layer. Replacing Cortex Analyst with Cortex Search would not address the need for consistent metric definitions and analytical SQL generation.
- D. Incorrect.
Incorrect. Cortex Analyst does not automatically derive authoritative business metric definitions just from storing more historical data. More rows may help analytical coverage, but they do not resolve ambiguity between terms like "sales" and "net sales." The business logic must be defined explicitly in the semantic model.