SnowPro Specialty: Gen AI Question 204
Single answerSpecification filesA data engineering team is preparing a Snowflake Cortex Analyst semantic model for a retail analytics chatbot. They want the model to answer questions consistently across business units, use business-friendly terminology, and avoid exposing raw table and column names such as SKU_ID or NET_SLS_AMT to end users. The team is defining a specification file for the semantic model before deploying it. Which action in the specification file best addresses this requirement?
- A
Define logical business entities, metrics, and friendly field names in the specification so the model maps user questions to curated semantic concepts instead of raw physical schema names.
- B
Store all warehouse sizing and auto-suspend settings in the specification file so Cortex Analyst can optimize how it interprets business terms.
- C
List every possible natural-language question users might ask in the specification file so the model does not need semantic mappings.
- D
Embed role-based access grants directly in the specification file so the model automatically hides unauthorized columns without relying on Snowflake security controls.
Show answer and explanation
Correct answer: A
Explanation
The best answer is to define a curated semantic layer in the specification file using business-friendly concepts that map to the physical schema. For Cortex Analyst and related semantic-modeling patterns in Snowflake, specification files are intended to describe the meaning of data so natural-language requests can be interpreted in a governed, reusable way. This includes defining logical entities, dimensions, and metrics with names that make sense to business users, rather than forcing users to know source-system column names.
The distractors reflect common misconceptions. Operational settings such as warehouse configuration are managed outside the specification file. Enumerating all possible user questions is not a practical or durable substitute for semantic modeling. And security remains the responsibility of Snowflake's native access-control framework; semantic specifications complement governance but do not replace authorization controls.
This aligns with Snowflake best practices for semantic modeling in Gen AI analytics use cases: separate physical schema design, compute configuration, and security administration from the semantic definitions that help AI systems interpret business intent accurately.
- A. Correct.
Correct. The purpose of the specification file for a semantic model is to describe the business-facing semantic layer: curated entities, dimensions, measures/metrics, and user-friendly terminology that map to underlying data structures. This helps Cortex Analyst interpret natural-language questions using business concepts rather than exposing raw physical object names. In practice, this is how teams improve consistency, usability, and governance for conversational analytics.
- B. Incorrect.
Incorrect. Warehouse sizing, compute management, and auto-suspend behavior are Snowflake operational settings, not semantic model specification concerns. They affect performance and cost, but they do not define business terminology or how natural-language questions map to tables and columns.
- C. Incorrect.
Incorrect. Providing example questions can be useful in some AI workflows, but a semantic specification is not meant to replace structured semantic definitions with an exhaustive list of prompts. Relying on enumerating every possible question is not scalable and does not solve the core requirement of creating governed, reusable business mappings.
- D. Incorrect.
Incorrect. Access control should be enforced through Snowflake security mechanisms such as roles, grants, and object-level protections, not embedded as a substitute inside a semantic specification file. The specification can describe semantics, but it does not replace Snowflake's native authorization model.