Question 490
At fictional Greyson Customer Network, a Snowflake data and AI engineering team is conducting an accountable generative AI operations change with legal, privacy and information-security review for any restricted inputs. The separately composed technical incident is: A finance team processes hundreds of low-quality scanned PDFs through the most expensive parsing path without identifying file classes that can use simpler validated extraction. 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
Track extraction quality and cost by document class, review failures and reassess routing rules after parser or pricing changes. Document AI optimization balances extraction quality, latency, privacy and price across heterogeneous input types.
Incorrect Answers
This response concerns Document OCR versus layout parsing choice, which is different from the stated failure. A finance team processes hundreds of low-quality scanned PDFs through the most expensive parsing path without identifying file classes that can use simpler validated extraction.
Idempotent document reindexing pipeline is not the cause of the reported failure: A finance team processes hundreds of low-quality scanned PDFs through the most expensive parsing path without identifying file classes that can use simpler validated extraction.
Document stage ingestion and least privilege alone does not establish the needed evidence: A controlled sample meets field-accuracy objectives while document-level processing costs fall and unreadable files enter a review queue.