Question 110
At fictional Aspen Education Trust, a Snowflake data and AI engineering team is conducting an AI retrieval and source-citation assurance review with regional service policies and accountable stakeholders. The separately composed technical incident is: A commerce team selects the largest available language model for every task without comparing latency, price, supported regions, structured output reliability or acceptable fallback behavior. Assume only these documented facts and use safe, authorized production-change procedures; exact current Snowflake feature syntax and region or account availability require current review. Which repeatable assurance or control is most relevant to preventing this situation?
Explanation
Writer-proposed answer: B (UNVERIFIED)
Preliminary individual option explanations — NOT expert validated
Correct Answer
Periodically rebalance the approved model matrix using actual production metrics, release changes and documented business acceptance tests. Model suitability depends on the task, governance, performance and cost rather than parameter size or marketing claims alone.
Incorrect Answers
This response concerns Prompt injection in retrieved business content, which is different from the stated failure. A commerce team selects the largest available language model for every task without comparing latency, price, supported regions, structured output reliability or acceptable fallback behavior.
Custom model artifact governance is not the cause of the reported failure: A commerce team selects the largest available language model for every task without comparing latency, price, supported regions, structured output reliability or acceptable fallback behavior.
Grounded employee policy assistant architecture alone does not establish the needed evidence: A scored model comparison shows response quality, region fit, cost and latency within agreed limits, including observed behavior during a controlled fallback exercise.