SnowPro Specialty: Gen AI Question 45
Select 21.2 Outline Gen AI capabilities in Snowflake.A retail company wants to add a Gen AI assistant for customer support analytics without moving data out of Snowflake. Analysts need to summarize case notes, classify customer intent, and ask natural-language questions over product documentation stored in Snowflake stages and tables. The solution must rely on Snowflake-native Gen AI capabilities rather than custom model hosting. Which Snowflake capabilities should the architect recommend? (Select TWO.)
- A
Use Snowflake Cortex AI functions to perform tasks such as summarization and classification directly in SQL.
- B
Use Cortex Search to provide low-latency semantic retrieval over the company’s documentation for question-answering experiences.
- C
Deploy a custom GPU cluster inside a Snowflake virtual warehouse to host and fine-tune open source LLMs natively.
- D
Use external functions as Snowflake’s built-in method for vector embedding generation and semantic search.
- E
Store the source documents in Snowflake tables or stages and use Snowflake-native AI capabilities against that governed data.
Show answer and explanation
Correct answers: A, B
Explanation
The best answer is to combine Snowflake Cortex AI functions with Cortex Search. Cortex AI covers inference tasks such as summarization and classification directly within Snowflake, helping teams build Gen AI workflows close to governed enterprise data. Cortex Search addresses semantic retrieval and powers question-answering experiences over unstructured or semi-structured document corpora stored in Snowflake. Together, these capabilities support a typical retrieval-augmented generation pattern without requiring data movement or custom model hosting. Snowflake documentation and product guidance emphasize using Cortex AI for serverless LLM-powered functions and Cortex Search for low-latency semantic retrieval over Snowflake data. By contrast, customer-managed GPU model hosting inside virtual warehouses is not the intended Snowflake-native pattern for this use case, and external functions are integration tools rather than the primary built-in Gen AI capability asked for here.
- A. Correct.
Correct. Snowflake Cortex AI provides serverless AI/LLM capabilities that can be invoked from SQL, enabling common Gen AI tasks such as summarization, sentiment or intent-style classification, extraction, translation, and related inference workflows without exporting data from Snowflake. This matches the requirement to analyze support case notes using Snowflake-native capabilities.
- B. Correct.
Correct. Cortex Search is designed for semantic retrieval over enterprise data in Snowflake, supporting retrieval-augmented applications such as natural-language question answering over documents. This is the appropriate Snowflake-native capability for enabling users to query product documentation stored in Snowflake.
- C. Incorrect.
Incorrect. Snowflake virtual warehouses do not become custom GPU model-hosting clusters for customers to natively deploy and fine-tune arbitrary open source LLMs inside Snowflake in this manner. This option reflects a common misconception that warehouses can be repurposed as general model-serving infrastructure.
- D. Incorrect.
Incorrect. External functions are not Snowflake’s built-in semantic search mechanism. While external functions can integrate Snowflake with outside services, the question explicitly asks for Snowflake-native Gen AI capabilities. Cortex Search is the native retrieval capability, and Cortex AI provides native model inference functions.
- E. Incorrect.
Incorrect. Storing documents in tables or stages is a valid Snowflake data pattern, but this option by itself is too broad and does not identify the actual Gen AI capabilities needed to satisfy the requirements. It describes where data can reside, not the specific feature set the architect should recommend.