SnowPro Specialty: Gen AI Question 266
Single answerUsage quotasA Snowflake administrator is enabling AI features for several internal teams. The finance team wants to prevent runaway costs from large-language-model usage while still allowing analysts to experiment with AI-powered SQL functions in a governed way. The administrator needs a control that limits how much AI consumption can occur for a specific group over time and can be monitored as part of account governance. Which approach best meets this requirement?
- A
Create and assign a usage quota for the appropriate AI consumption scope so the team is limited to a defined amount of usage over the quota period.
- B
Resize the virtual warehouse used by the finance team so AI requests automatically stop when the warehouse reaches its credit limit.
- C
Use a resource monitor on the finance team's warehouse because Cortex AI function usage is controlled only through warehouse credit thresholds.
- D
Grant fewer database privileges to the finance team because usage quotas are enforced through object-level access control rather than metering limits.
Show answer and explanation
Correct answer: A
Explanation
The key requirement is to allow AI usage but constrain it to a defined amount for governance and cost control. In Snowflake, usage quotas are the purpose-built mechanism for this kind of limit on supported AI consumption. By contrast, warehouse resizing and resource monitors are warehouse-compute controls, and object privileges are access controls; neither provides the same quota-based AI usage governance. Candidates should distinguish between access management, warehouse cost management, and AI usage management. Snowflake documentation and product guidance describe usage quotas as a governance feature for limiting eligible AI usage over a quota period, which is the best fit for this scenario.
- A. Correct.
Correct. Usage quotas are intended to cap eligible usage for supported Snowflake AI/GenAI consumption scenarios over a defined period. This is the most direct governance mechanism when the goal is to limit spend or consumption for a team while still permitting authorized use. It aligns with the requirement to control and monitor AI usage rather than block access entirely.
- B. Incorrect.
Incorrect. Virtual warehouse sizing affects compute resources for warehouse-executed workloads, but it is not the primary governance control for limiting Snowflake AI model usage through usage quotas. A smaller warehouse may reduce performance, but it does not provide the required quota-based limit for AI consumption over time.
- C. Incorrect.
Incorrect. Resource monitors are designed to track and control warehouse credit consumption. They are a common governance tool for compute cost management, which makes this distractor plausible. However, they do not serve as the main mechanism for quota-based control of supported AI usage in the way described in the scenario.
- D. Incorrect.
Incorrect. Privilege grants determine whether users can access objects or execute functions, but they do not set a metered consumption ceiling. Reducing privileges could prevent use entirely, which conflicts with the business requirement to allow experimentation under a controlled usage limit.