SnowPro Specialty: Gen AI Question 103
Single answerSUMMARIZEA support analytics team stores customer case notes in a Snowflake table. They want to generate a short, readable summary of each case directly in SQL so analysts can quickly review long text entries. The team does not need structured extraction or sentiment analysis, only a concise natural-language summary of each note. Which approach is the most appropriate?
- A
Use the SUMMARIZE function on the case note text column in a SELECT statement.
- B
Use CLASSIFY_TEXT on the case note text column because classification models produce condensed summaries.
- C
Use EXTRACT_ANSWER with a generic prompt so Snowflake can infer the most important points from each case note.
- D
Create vector embeddings for each case note and query the embedding values directly to retrieve a summary.
Show answer and explanation
Correct answer: A
Explanation
The best choice is to use SUMMARIZE because the business need is straightforward summarization of unstructured text into concise natural language. In Snowflake Cortex AISQL, SUMMARIZE is the function intended for this use case. By contrast, CLASSIFY_TEXT is for labeling text into predefined categories, EXTRACT_ANSWER is for pulling specific answers from source text, and embeddings support semantic retrieval rather than direct summarization. This reflects a key exam skill: selecting the Cortex function that matches the business task instead of using a more generic or indirect GenAI approach.
- A. Correct.
Correct. SUMMARIZE is designed to generate concise natural-language summaries from input text. In this scenario, the requirement is specifically to produce short readable summaries of long support case notes directly in SQL, which aligns with the purpose of SUMMARIZE.
- B. Incorrect.
Incorrect. CLASSIFY_TEXT is intended for assigning input text to one of a set of categories, not for generating a free-form summary. A candidate might choose this option if they confuse text understanding tasks with text generation tasks.
- C. Incorrect.
Incorrect. EXTRACT_ANSWER is used to extract answers from text based on a question or prompt-like input, not to create a general summary of an entire document. It is better suited for question-answering or targeted information retrieval than summarization.
- D. Incorrect.
Incorrect. Embeddings are useful for semantic search, similarity, clustering, and retrieval workflows, but embedding vectors themselves do not provide a human-readable summary. A candidate might choose this if they associate all GenAI text processing tasks with embeddings, which is a common misconception.