Question 140
At fictional Olivebridge Logistics, a Snowflake data and AI engineering team is conducting a trusted self-service analytics redesign under documented decision rights and minimum-privilege security requirements. The separately composed technical incident is: An order-management application expects a machine-readable JSON decision from an LLM, yet downstream code assumes every generated response matches its schema and crashes on malformed fields. 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: D (UNVERIFIED)
Preliminary individual option explanations — NOT expert validated
Correct Answer
Maintain schema-version tests, reject unknown critical fields and log structured-output validation failures without exposing sensitive prompts. Model-generated structure is untrusted input until checked against a typed contract and the application's business invariants.
Incorrect Answers
This response concerns RAG chunk boundaries break policy answers, which is different from the stated failure. An order-management application expects a machine-readable JSON decision from an LLM, yet downstream code assumes every generated response matches its schema and crashes on malformed fields.
Cortex Analyst ambiguity in dimension vocabulary is insufficient to demonstrate this result: Negative and boundary tests show schema-invalid generations are rejected without mutating orders, while approved structured responses preserve all required fields.
Verified query repository for business metric questions is not the cause of the reported failure: An order-management application expects a machine-readable JSON decision from an LLM, yet downstream code assumes every generated response matches its schema and crashes on malformed fields.