SnowflakeExpert level

SnowPro Specialty: Gen AI exam dumps: 287 free SnowPro Specialty: Gen AI practice questions

Free SnowPro Specialty: Gen AI practice questions for the SnowPro® Specialty: Gen AI exam, with the correct answer and a full explanation for every option. Read the first 10 below, browse all 287 by number, or take a timed practice exam.

Question bank last updated April 2026

Start timed practice exam Browse all 287 questions

Free SnowPro Specialty: Gen AI practice questions

Questions 1 to 10 of 287

Pick an answer before you open the explanation. Each question also has its own page with a permalink.

SnowPro Specialty: Gen AI Question 1

Single answer1.1 Define Snowflake's Gen AI principles, features, and best practices.

A financial services company wants to deploy an internal Gen AI assistant in Snowflake to help analysts summarize policy documents and answer questions about approved internal knowledge sources. The security team requires that sensitive data remain governed in Snowflake, that access controls continue to apply to the source data, and that the solution follow Snowflake-recommended Gen AI best practices for accuracy and safety. Which approach should the architect recommend?

A
Export the policy documents to an external LLM provider, generate embeddings and prompts outside Snowflake, and return the responses back into Snowflake for reporting.
B
Use Snowflake Cortex capabilities within Snowflake, keep the documents and prompts governed by Snowflake roles and policies, and implement retrieval grounded on trusted enterprise data to reduce hallucinations.
C
Fine-tune a public model on all policy documents by copying unrestricted data extracts to a developer-managed environment, because model customization is the primary Snowflake best practice for accuracy.
D
Allow analysts to prompt a general-purpose model directly with unrestricted access to all database schemas, because broader model context is more important than enforcing existing Snowflake governance controls.

Show answer and explanation

Correct answer: B

Explanation

The best answer is Option 2 because it applies Snowflake's Gen AI principles in a realistic enterprise setting: keep data protected in Snowflake, preserve governance and access controls, and improve model output by grounding responses in approved internal data. Snowflake emphasizes secure AI on governed data, using existing Snowflake security controls and minimizing unnecessary data movement. For enterprise assistants, a retrieval-based pattern over trusted content is a widely recommended best practice to improve factuality and reduce hallucinations. This is generally preferable to exporting data to external platforms or granting overly broad access. Candidates should recognize that Snowflake Gen AI solutions are designed to combine AI capabilities with Snowflake's data governance, security, and enterprise data platform strengths.

A. Incorrect.
Incorrect. This approach moves sensitive content outside Snowflake and weakens the governance posture the company explicitly wants to preserve. A core Snowflake Gen AI principle is to bring AI to governed data where possible, rather than exporting data unnecessarily to external systems. This option also makes it harder to consistently enforce Snowflake access controls, masking, and other policies on the data used for prompting and retrieval.
B. Correct.
Correct. This approach aligns with Snowflake Gen AI principles and best practices: keep data in Snowflake, apply existing governance and role-based access controls, and use grounded retrieval on trusted enterprise content to improve response quality and reduce hallucinations. For enterprise use cases, grounding LLM responses with relevant internal context is a key best practice, especially when accuracy and compliance matter.
C. Incorrect.
Incorrect. This reflects a common misconception that fine-tuning is the default or primary best practice for enterprise Gen AI accuracy. In many business scenarios, retrieval-based grounding on current enterprise data is more appropriate than copying data to another environment for customization. Copying unrestricted extracts to a developer-managed environment also conflicts with the requirement to keep sensitive data governed in Snowflake.
D. Incorrect.
Incorrect. This violates least-privilege and governance best practices. Snowflake's Gen AI approach emphasizes secure, governed access to enterprise data, not broad unrestricted exposure. Giving a model access to all schemas increases the risk of overexposure of sensitive information and does not address answer quality as effectively as retrieval over approved sources.

SnowPro Specialty: Gen AI Question 2

Single answer1.1 Define Snowflake's Gen AI principles, features, and best practices.

A financial services company wants to build an internal assistant in Snowflake that summarizes support tickets and answers employee questions using company knowledge. The security team requires that sensitive enterprise data remain governed inside Snowflake, and the platform team wants to minimize custom infrastructure while following Snowflake Gen AI best practices. Which approach BEST meets these requirements?

A
Export ticket data daily to an external vector database and third-party LLM service, because Snowflake Gen AI features are intended only for unstructured files rather than governed enterprise tables.
B
Use Snowflake Cortex AI functions and related Gen AI capabilities directly on data in Snowflake, applying role-based access controls and existing governance policies so prompts and retrieved context follow the same data protections as other workloads.
C
Fine-tune a custom model outside Snowflake first, then reload the generated embeddings and responses into Snowflake, because governance can only be applied after AI outputs are written back to tables.
D
Move all support ticket data into a separate Snowflake account dedicated to AI, because Snowflake best practice is to isolate Gen AI workloads from the source data used by business analysts.

Show answer and explanation

Correct answer: B

Explanation

The best answer is to use Snowflake's native Gen AI capabilities, such as Cortex AI functions, directly where the enterprise data already lives. A key Snowflake Gen AI principle is minimizing data movement and applying AI within the platform's existing security, governance, and access-control framework. This helps organizations use structured and unstructured data for AI while preserving centralized governance and reducing operational overhead. In practice, best practices include keeping sensitive data in Snowflake when possible, using existing RBAC and governance features, and preferring native platform capabilities over unnecessary external pipelines when requirements can be met internally. These principles are consistent with Snowflake documentation on Cortex AI, data governance, and secure AI application design.

A. Incorrect.
Incorrect. This option conflicts with a core Snowflake Gen AI principle: bring AI to the data rather than moving governed enterprise data out unnecessarily. Exporting sensitive data to external services increases operational complexity and governance risk. Snowflake Gen AI capabilities are designed to work with structured and unstructured data in Snowflake, not just files.
B. Correct.
Correct. This is the best fit for the scenario and aligns with Snowflake Gen AI principles and best practices: use built-in Cortex AI capabilities on governed data where it already resides, minimize data movement, and rely on Snowflake security and governance controls such as roles and policies. This reduces infrastructure overhead while supporting secure, practical enterprise AI adoption.
C. Incorrect.
Incorrect. This reflects a common misconception that custom external model work is required before Snowflake can be used effectively for Gen AI. While organizations may choose external model workflows in some cases, it is not the best answer here because the stated goal is to keep data governed inside Snowflake and minimize custom infrastructure. Governance in Snowflake applies to the data and access model before and during AI workflows, not only after outputs are stored.
D. Incorrect.
Incorrect. Creating a separate account solely for AI is not a general Snowflake Gen AI best practice. It can add complexity and may fragment governance rather than simplify it. The requirement is to keep sensitive data governed and reduce custom infrastructure, which is better addressed by using native Snowflake capabilities with existing controls instead of introducing unnecessary account separation.

SnowPro Specialty: Gen AI Question 3

Single answerSnowflake Cortex

A retail company stores millions of product reviews in a Snowflake table. The analytics team wants to classify each review as POSITIVE, NEGATIVE, or NEUTRAL directly in Snowflake and then join the results to existing sales dashboards. They want the fastest path to production with minimal infrastructure to manage and no model training. Which approach is the MOST appropriate using Snowflake Cortex?

A
Use a Cortex LLM function such as SNOWFLAKE.CORTEX.COMPLETE with a prompt that asks the model to return one of the three sentiment labels, and persist the results in a table for downstream SQL joins.
B
Export the reviews to an external GPU environment, fine-tune a custom sentiment model, and write predictions back to Snowflake because Cortex requires model training for classification tasks.
C
Create a Snowpark Container Services application first, because Cortex functions cannot be called directly from SQL for text processing workloads.
D
Use Dynamic Tables only, because they provide built-in sentiment classification without requiring any Cortex or model inference functions.

Show answer and explanation

Correct answer: A

Explanation

This scenario is about selecting the most practical Snowflake-native approach for sentiment classification when the team wants minimal operational overhead and no custom training. Snowflake Cortex is designed for exactly this type of use case: applying AI models directly within Snowflake using SQL, Python, or REST-based patterns depending on the feature. For prompt-based text classification, a Cortex LLM function can be used to instruct the model to output one of a constrained set of labels such as POSITIVE, NEGATIVE, or NEUTRAL. The output can then be persisted and joined to other business data using normal Snowflake SQL workflows. The key best-practice idea is to keep the workload close to the data and avoid unnecessary movement to external infrastructure unless there is a clear need for custom model hosting or specialized tuning. This aligns with Snowflake guidance around Cortex as a managed AI capability for inference inside the Snowflake platform.

A. Correct.
Correct. Snowflake Cortex provides SQL-accessible AI functions, including LLM inference through functions such as SNOWFLAKE.CORTEX.COMPLETE, which can be used to classify text with prompt-based instructions. For a team seeking minimal infrastructure and no model training, using a Cortex function directly in Snowflake is the most appropriate approach. The results can be stored in a table or view and joined with other Snowflake data using standard SQL.
B. Incorrect.
Incorrect. This adds unnecessary operational complexity and contradicts the requirement for the fastest path to production with minimal infrastructure and no model training. Cortex is specifically designed to let users invoke foundation models from within Snowflake without managing external GPU environments or training custom models for common text tasks like classification.
C. Incorrect.
Incorrect. Snowpark Container Services is intended for deploying and running custom containerized applications and models, but it is not required to use Snowflake Cortex. A common misconception is that all AI workloads in Snowflake require custom application hosting. In reality, Cortex functions are designed to be invoked directly from SQL or programmatic interfaces for many inference use cases.
D. Incorrect.
Incorrect. Dynamic Tables help automate data transformation pipelines, but they do not provide built-in sentiment classification by themselves. They can be used to orchestrate refreshes of derived results, including outputs from Cortex functions, but they are not a replacement for model inference.

SnowPro Specialty: Gen AI Question 4

Single answerSnowflake Cortex

A product support team stores customer cases in a Snowflake table with columns CASE_ID, PRODUCT_LINE, CASE_TEXT, and CREATED_AT. They want analysts to run SQL queries that automatically generate a short summary of each case and classify sentiment as positive, neutral, or negative. The solution must stay inside Snowflake, minimize operational overhead, and avoid building or managing external model endpoints. Which approach best meets these requirements?

A
Use Snowflake Cortex functions such as SUMMARIZE on CASE_TEXT and SENTIMENT on CASE_TEXT directly in SQL queries or views.
B
Export CASE_TEXT to an external LLM service through a custom application, store the responses back in Snowflake, and query the enriched table.
C
Train a custom Snowpark ML model for summarization and sentiment analysis, then deploy it as a user-managed service outside Snowflake.
D
Create a dynamic table that automatically infers summaries and sentiment labels without calling any AI function, because dynamic tables include built-in text understanding.

Show answer and explanation

Correct answer: A

Explanation

The best answer is to use Snowflake Cortex functions directly in SQL. Cortex is designed to let teams apply AI capabilities such as summarization and sentiment analysis within Snowflake without exporting data to external services or standing up separate inference infrastructure. This aligns with common Snowflake best practices: keep data in-platform when possible, reduce operational complexity, and expose reusable logic through SQL queries or views for analysts. External LLM integrations and custom model deployments may be valid in other scenarios, but they do not best satisfy the requirements here. Dynamic tables can help operationalize recurring transformations, but they are orchestration objects, not AI features by themselves; they would need to call Cortex functions to produce the desired outputs.

A. Correct.
Correct. Snowflake Cortex provides built-in AI functions that can be invoked directly from SQL for common generative AI and text analysis tasks. Using functions such as SUMMARIZE and SENTIMENT keeps processing inside Snowflake, reduces integration complexity, and avoids managing external endpoints. This matches the requirement for low operational overhead and analyst-friendly SQL access.
B. Incorrect.
Incorrect. Although this can work technically, it violates the stated goal of minimizing operational overhead and avoiding external model endpoint management. It also adds data movement, integration, security, and latency considerations that Snowflake Cortex is designed to reduce for in-platform AI use cases.
C. Incorrect.
Incorrect. A custom Snowpark ML approach may be appropriate for specialized predictive models, but it is unnecessarily complex for standard summarization and sentiment tasks already supported through Cortex functions. It also conflicts with the requirement to avoid building or managing external serving infrastructure.
D. Incorrect.
Incorrect. Dynamic tables help automate data transformation pipelines, but they do not provide built-in generative AI or text understanding by themselves. To derive summaries or sentiment, you would still need to call the relevant Cortex functions within the transformation logic.

SnowPro Specialty: Gen AI Question 5

Single answerLLMs

A retail company is building a customer-support assistant in Snowflake that uses a large language model (LLM) to answer questions from product manuals, warranty terms, and return-policy documents. During testing, the team finds that answers sometimes sound fluent but include policy details that are not present in the source documents. The team wants to reduce these hallucinations while still allowing the model to generate natural-language responses. Which approach is the MOST appropriate?

A
Implement retrieval-augmented generation (RAG) so the application retrieves relevant document chunks at query time and grounds the LLM response in that context
B
Increase the model temperature so the LLM explores more possible completions before answering policy questions
C
Replace the LLM with a larger general-purpose model and remove document context, because larger models memorize more enterprise policy details
D
Fine-tune the LLM only on historical support chats, without including the official policy and manual content in the inference workflow

Show answer and explanation

Correct answer: A

Explanation

The key issue is hallucination in a document-based question-answering workflow. Best practice for this type of LLM application is to ground generation using retrieval from trusted enterprise content, commonly referred to as retrieval-augmented generation (RAG). This allows the model to generate fluent answers while basing them on relevant passages from manuals, warranty terms, and policy documents. In practice, teams also often combine RAG with prompt instructions, citations, and lower-temperature settings for higher factual reliability. By contrast, simply increasing model size or temperature does not address grounding, and fine-tuning without retrieval does not reliably keep answers aligned with the latest official content. This aligns with common generative AI architecture guidance for enterprise LLM applications, including Snowflake-oriented patterns that pair vector search or semantic retrieval with LLM inference to improve answer accuracy.

A. Correct.
Correct. Retrieval-augmented generation (RAG) is a standard pattern for reducing hallucinations in enterprise question-answering scenarios. Instead of relying only on the model's parametric knowledge, the application retrieves the most relevant passages from trusted source documents and provides them as context to the LLM at inference time. This grounds the response in authoritative data such as manuals and policy documents, which is especially important when answers must reflect current business rules.
B. Incorrect.
Incorrect. Higher temperature increases randomness and creativity in token selection. That can be useful for brainstorming or marketing copy, but it generally makes factual consistency worse, not better. For support and policy answers, teams usually prefer lower temperature and stronger grounding in source content.
C. Incorrect.
Incorrect. A larger model may improve general reasoning or language quality, but it does not guarantee accurate knowledge of a specific company's current warranty or return policies. Removing document context would typically increase the risk of unsupported answers because the model would have to rely on pretraining knowledge rather than the enterprise's source of truth.
D. Incorrect.
Incorrect. Fine-tuning on historical support chats may teach tone and common response patterns, but if those chats contain mistakes, outdated policies, or incomplete information, the model can reproduce those issues. Fine-tuning alone also does not ensure that answers reflect the latest official documents at runtime. In this scenario, grounding responses in current source material is the more appropriate primary solution.

SnowPro Specialty: Gen AI Question 6

Single answerLLMs

A retail company is building a customer-support assistant in Snowflake that uses a large language model (LLM) to answer questions from product manuals, warranty documents, and return policies. During testing, the team notices that the model sometimes produces fluent but incorrect answers when the requested information is not present in the source documents. They want to reduce these hallucinations without fine-tuning the base model and while keeping the implementation maintainable. Which approach is the BEST fit for this requirement?

A
Implement retrieval-augmented generation (RAG) so the application retrieves relevant document chunks at query time and includes them in the prompt, while instructing the model to answer only from the provided context
B
Increase the temperature setting so the model explores a wider range of possible answers and is less likely to repeat incorrect patterns
C
Replace the LLM with a larger parameter model and rely on its pretrained knowledge instead of grounding responses in company documents
D
Fine-tune the model on a small sample of support tickets so it memorizes the correct answers for future prompts

Show answer and explanation

Correct answer: A

Explanation

The best answer is to implement retrieval-augmented generation (RAG). In enterprise LLM applications, hallucinations often occur because the model is asked to answer questions that require current, proprietary, or highly specific information that is not reliably embedded in the model's pretrained parameters. RAG addresses this by retrieving relevant data from a trusted knowledge base at runtime and passing it as context to the model. This pattern is widely considered a best practice for question answering over business documents because it improves factual grounding, supports changing source content without retraining, and keeps the system more maintainable than repeated fine-tuning cycles. In Snowflake-based GenAI architectures, this aligns with using data stored in Snowflake as the source of truth and building LLM workflows that retrieve and condition on enterprise data rather than relying solely on the base model's prior knowledge.

A. Correct.
Correct. RAG is a standard pattern for reducing hallucinations in enterprise question-answering use cases. By retrieving the most relevant passages from trusted documents at inference time and supplying that context to the model, the application grounds the response in authoritative data. Adding prompt instructions such as answering only from supplied context and indicating when the answer is unavailable further improves reliability. This approach also avoids the operational cost and maintenance burden of model fine-tuning when the main issue is missing or changing factual context.
B. Incorrect.
Incorrect. Higher temperature generally increases randomness and creativity, which is usually the opposite of what you want when trying to reduce hallucinations in factual support workflows. For grounded question answering, lower temperature is often preferred because it encourages more deterministic outputs. The misconception here is assuming that more variation helps correctness; in practice, it often makes factual consistency worse.
C. Incorrect.
Incorrect. A larger model may improve overall reasoning or language quality, but it does not solve the core problem: the model is answering without reliable grounding in the company's current documents. Pretrained knowledge can be outdated, incomplete, or irrelevant to proprietary policies. This option reflects the common misconception that model size alone is the best way to improve factual accuracy in enterprise-specific tasks.
D. Incorrect.
Incorrect. Fine-tuning can help shape style or task behavior in some cases, but it is not usually the best first solution when the issue is access to up-to-date source content. A small support-ticket dataset may not cover all policy and product-document variations, and fine-tuning does not inherently provide live access to changing documents. This also adds lifecycle complexity compared with a retrieval-based design.

SnowPro Specialty: Gen AI Question 7

Single answerCortex Search

A retail company wants to build an internal support assistant that answers employee questions using thousands of product manuals stored in a Snowflake table. The team needs low-latency retrieval over the manual text and wants search results to include structured metadata such as product_line, region, and last_updated so the application can filter or rank results before sending them to an LLM. Which approach should the team use with Cortex Search to best meet these requirements?

A
Create a Cortex Search service on the manuals table, define the manual text as the searchable content, and include product_line, region, and last_updated as attributes returned with search results
B
Create a dynamic table that generates embeddings for each manual chunk, then query the dynamic table directly with SQL similarity calculations because Cortex Search cannot return metadata fields
C
Use Cortex Complete directly on the manuals table and rely on prompt instructions to search all records at inference time, because Cortex Search is only for exact keyword lookups
D
Store the manuals in an external stage and use Snowflake file metadata for filtering, because Cortex Search cannot index data stored in Snowflake tables

Show answer and explanation

Correct answer: A

Explanation

The best answer is to create a Cortex Search service over the manuals stored in Snowflake and configure it so the text content is searchable while business metadata is returned as attributes. This aligns with the typical retrieval-augmented generation pattern: use Cortex Search for fast retrieval, then pass the most relevant passages plus metadata into an LLM such as Cortex Complete. A key practical benefit of Cortex Search is that it can return structured attributes with each result, enabling application-side filtering, personalization, or ranking. By contrast, using an LLM alone for retrieval is not a best practice, and building a custom similarity search pipeline is unnecessary when Cortex Search already provides managed retrieval for Snowflake data. See Snowflake documentation for Cortex Search service concepts, including searchable text, returned attributes, and its use in RAG applications.

A. Correct.
Correct. Cortex Search is designed for low-latency retrieval over Snowflake data and supports returning attributes alongside matching results. In this scenario, the manual text should be indexed as the searchable content, while fields such as product_line, region, and last_updated can be exposed as attributes for downstream filtering, ranking, or prompt construction in a RAG workflow.
B. Incorrect.
Incorrect. While teams can build custom embedding pipelines, this option is wrong because it claims Cortex Search cannot return metadata fields. Cortex Search services are specifically intended to return relevant results together with selected attributes. Choosing this reflects a common misconception that search output is limited to only the matched text.
C. Incorrect.
Incorrect. Cortex Complete is for text generation, not document retrieval. Prompting an LLM to search a large table at inference time is not an efficient or reliable replacement for a retrieval layer. This option also incorrectly characterizes Cortex Search as only supporting exact keyword lookups; Cortex Search is intended for semantic and lexical retrieval use cases in enterprise search and RAG scenarios.
D. Incorrect.
Incorrect. Cortex Search operates on data in Snowflake tables and views, so there is no requirement to move manuals to an external stage. File metadata from staged files is also not a substitute for indexing document content and returning business-specific attributes from a search service.

SnowPro Specialty: Gen AI Question 8

Single answerCortex Search

A retail company is building an internal support assistant in Snowflake. The assistant must answer employee questions using the latest product manuals stored in a Snowflake table, and each answer should cite the exact manual section used. The team wants to avoid building and maintaining its own vector pipeline. Which approach best meets these requirements with Cortex Search?

A
Create a Cortex Search service on the table containing the manual text and include columns such as document title and section as attributes to return with search results for grounding and citations.
B
Use AI_COMPLETE directly on the manuals table and rely on the model to infer the most relevant passages without a search index.
C
Export the manuals from Snowflake to an external vector database, generate embeddings outside Snowflake, and return only the final answer to the assistant.
D
Create a dynamic table that stores one row per manual and query it with SQL LIKE predicates before sending matching rows to the model.

Show answer and explanation

Correct answer: A

Explanation

The best answer is to use Cortex Search as the retrieval layer over the manuals stored in Snowflake. Cortex Search is intended for RAG-style applications where relevant chunks or rows must be retrieved from Snowflake data and returned with metadata that can be used for grounding and citations. This is especially useful when the organization wants to keep data in Snowflake and avoid managing a separate embedding and indexing stack. In practice, teams define a Cortex Search service over the source data and include useful attributes, such as document identifiers or section labels, so the application can display source references with the generated answer. This follows Snowflake best practices for building grounded GenAI assistants: use retrieval to supply relevant context, then pass that context to generation rather than relying on the model alone.

A. Correct.
Correct. Cortex Search is designed for retrieval over Snowflake data and handles the underlying search index so teams do not need to build and operate a separate vector pipeline. By defining the service over the manuals data and returning attributes such as document title, section name, or URL, the application can ground responses and present citations alongside retrieved passages. This aligns directly with the scenario's requirements for fresh content, retrieval, and source attribution.
B. Incorrect.
Incorrect. AI_COMPLETE can generate answers, but by itself it is not a retrieval system over enterprise documents. Relying only on the model without retrieval increases the risk of hallucinations and does not provide a robust way to fetch the latest relevant manual sections or return precise citations. A common misconception is assuming generation alone replaces retrieval.
C. Incorrect.
Incorrect. While an external vector database could be used in some architectures, the scenario explicitly wants to avoid building and maintaining a custom vector pipeline. Moving data out of Snowflake and managing external embeddings adds operational complexity and is unnecessary when Cortex Search provides managed retrieval directly on Snowflake data.
D. Incorrect.
Incorrect. SQL LIKE can support simple keyword matching, but it is not the best fit for a support assistant that needs high-quality semantic retrieval across manuals. It also places the burden on the team to engineer brittle text-matching logic. Another common misconception is treating traditional pattern matching as equivalent to modern retrieval for RAG use cases.

SnowPro Specialty: Gen AI Question 9

Single answerCortex Analyst

A retail company wants business users to ask natural-language questions such as "What were online sales in Germany last quarter by product category?" using Cortex Analyst. The analytics team has a semantic model over sales data, but users report that Analyst sometimes picks the wrong metric when the word "sales" is used because the model contains both GROSS_SALES and NET_SALES. The team wants to improve answer accuracy without changing the underlying warehouse tables or requiring users to learn SQL. Which action is the best way to address this issue?

A
Update the semantic model to define clearer business-friendly metric names, descriptions, and synonyms so Cortex Analyst can better map user language to the intended measure
B
Create a new virtual warehouse dedicated to Cortex Analyst queries so the model has more compute available when interpreting the prompt
C
Convert the semantic model into a search service so Cortex Analyst can retrieve matching rows before generating SQL
D
Replace the semantic model with a dynamic table that materializes both GROSS_SALES and NET_SALES into one column named SALES

Show answer and explanation

Correct answer: A

Explanation

Cortex Analyst is intended for natural-language analytics over structured data and depends on a well-designed semantic model to map user questions into the correct SQL. When multiple metrics are semantically close, such as GROSS_SALES and NET_SALES, the most effective remediation is to refine the semantic model by improving metric names, business descriptions, and synonyms so common user language is unambiguous. This aligns with Snowflake best practices for governed semantic layers: make business terms explicit, model metrics and dimensions clearly, and use user-friendly terminology that matches how consumers ask questions. Compute scaling may help latency, and transformation features like dynamic tables may help data preparation, but neither addresses the root cause of incorrect metric selection in Cortex Analyst.

A. Correct.
Correct. Cortex Analyst relies on the semantic model to understand business concepts and map natural-language requests to the correct metrics, dimensions, and filters. If users say "sales" but the model has multiple candidate measures, the best practice is to improve the semantic layer with clearer metric definitions, descriptions, and synonyms/aliases that reflect how business users actually speak. This directly addresses ambiguity while preserving governed access to the underlying data.
B. Incorrect.
Incorrect. A larger or dedicated virtual warehouse can improve query execution performance, but it does not resolve semantic ambiguity in natural-language interpretation. The problem described is that Analyst is selecting the wrong metric, which is a modeling and business-language issue rather than a compute-capacity issue.
C. Incorrect.
Incorrect. Cortex Analyst is designed to work from a semantic model that helps generate accurate SQL over structured data. A search service is not a substitute for the semantic definitions needed to disambiguate business terms like GROSS_SALES versus NET_SALES. This option reflects a common misconception that retrieval infrastructure solves all LLM accuracy problems, even when the main issue is semantic modeling for analytics.
D. Incorrect.
Incorrect. Combining distinct business metrics into a single column would reduce clarity and governance, and it would likely make reporting less accurate rather than more accurate. Dynamic tables can help with transformation or precomputation workflows, but they are not the correct tool for resolving meaning in user questions. The better solution is to preserve separate governed metrics and improve how they are represented in the semantic model.

SnowPro Specialty: Gen AI Question 10

Single answerCortex Analyst

A retail analytics team is building a natural-language analytics experience for business users. They want users to ask questions such as "What were online sales in the Northeast last quarter by product category?" and have Cortex Analyst generate accurate SQL against governed Snowflake data. The team has a semantic model file, but early testing shows that Analyst sometimes chooses the wrong revenue field because the model contains both GROSS_SALES and NET_SALES with similar descriptions. Which action is the best way to improve the reliability of Cortex Analyst's generated SQL without giving users direct SQL access?

A
Refine the semantic model by giving clearer business definitions, metrics, and field descriptions so Cortex Analyst can map user terms to the intended data objects.
B
Grant business users broader privileges on the underlying tables so Cortex Analyst can inspect more raw data at runtime and infer the correct measure.
C
Replace Cortex Analyst with Cortex Search because search indexes are designed to resolve ambiguous metric definitions in structured analytics queries.
D
Move the semantic model logic into a Python stored procedure so Analyst does not need metadata about metrics and dimensions.

Show answer and explanation

Correct answer: A

Explanation

The best answer is to improve the semantic model. Cortex Analyst is built for natural-language querying over structured enterprise data using a semantic layer that defines metrics, dimensions, relationships, and business meaning. When generated SQL is unreliable, especially because of ambiguous or overlapping fields, the highest-value correction is to make the semantic model more explicit and business-friendly. This aligns with Snowflake best practices for governed analytics: define clear metric semantics, reduce ambiguity, and let Analyst generate SQL against approved data objects instead of expanding raw access. In Snowflake documentation for Cortex Analyst, the semantic model is the core artifact that helps the service map natural language to the correct SQL representation. Therefore, refining that model is the most effective and secure remediation.

A. Correct.
Correct. Cortex Analyst relies on the semantic model to understand business concepts, metrics, dimensions, relationships, and synonyms when translating natural language into SQL. If two fields such as GROSS_SALES and NET_SALES are ambiguously described, the most effective fix is to improve the semantic model with precise names, descriptions, and business definitions. This is the practical governance-first approach for improving SQL generation accuracy.
B. Incorrect.
Incorrect. Cortex Analyst does not become more accurate simply because end users have broader direct access to base tables. Expanding privileges can weaken governance and does not address the root cause: ambiguity in the semantic model. Best practice is to keep access controlled and improve the semantic layer rather than expose raw tables unnecessarily.
C. Incorrect.
Incorrect. Cortex Search is intended for retrieval over indexed content, not for generating governed analytical SQL from a semantic model. It is not a replacement for Cortex Analyst in BI-style natural language analytics scenarios. A candidate might choose this option by confusing retrieval use cases with text-to-SQL analytics use cases.
D. Incorrect.
Incorrect. Moving logic to a Python stored procedure does not solve Analyst's need for well-defined semantic metadata. Cortex Analyst is designed to use a semantic model to interpret user intent and generate SQL. Procedural logic may be useful elsewhere, but it is not the primary mechanism for disambiguating business metrics in Analyst.

Timed practice exam

Take a SnowPro Specialty: Gen AI practice test under exam conditions

55 questions in 85 minutes, drawn from this bank, with a score report and a per-question review when you finish.

Start timed exam

What the SnowPro Specialty: Gen AI exam covers

The objectives this question bank covers most, by number of questions.

Cortex Analyst
10 questions
Cortex Search
8 questions
Snowflake Cortex
6 questions
Cortex Agents (Public Preview)
6 questions
REST APIs
6 questions
Cortex Fine-tuning
4 questions
Required privileges
4 questions
CORTEX_MODELS_ALLOWLIST parameter
4 questions

All 287 SnowPro Specialty: Gen AI practice questions

Every question has a page with the answer and explanation. Numbers are stable, so you can bookmark or share them.

SnowPro Specialty: Gen AI exam dumps FAQ

Are these SnowPro Specialty: Gen AI dumps real exam questions?

No. These are original practice questions written to the SnowPro® Specialty: Gen AI exam objectives, not questions copied from a live exam. Memorising leaked questions violates Snowflake's candidate agreement and stops working the moment the question pool rotates. Use this bank to check your understanding of each domain and to find the topics you still need to study.

How many SnowPro Specialty: Gen AI practice questions are there?

287 questions, each with the correct answer, an explanation of the answer, and a note on why every other option is wrong. The first 10 are on this page and every question has its own page linked below.

Are the SnowPro Specialty: Gen AI exam dumps free?

Yes. Every question, answer and explanation on this page and the linked question pages is free to read without an account. A free HydraNode account adds timed practice exams, scoring and progress tracking across attempts.

How do I take a timed SnowPro Specialty: Gen AI practice test?

Sign in and start the SnowPro® Specialty: Gen AI exam on HydraNode. A session gives you 55 questions drawn from this bank in 85 minutes, then a score report with a per-question review.

What topics does the SnowPro Specialty: Gen AI exam cover?

The questions in this bank are grouped under: Cortex Analyst; Cortex Search; Snowflake Cortex; Cortex Agents (Public Preview); REST APIs; Cortex Fine-tuning; Required privileges; CORTEX_MODELS_ALLOWLIST parameter.