Q
A customer-support agent uses a RAG workflow over an Azure AI Search index to answer product and warranty questions. Grounding documents are ingested from Microsoft SharePoint Online on a nightly schedule.
The document library contains duplicate files, and some policy documents are updated during the day. As a result, the agent sometimes returns outdated answers, and token usage per request has increased.
You need to ensure that responses use the most current and relevant documents, while reducing retrieval and inference costs.
What should you do?
Question Info
Select All That Apply
Tick every correct option, then press Check Answer.
Explanation
Objective:
1.1 Analyze requirements for AI-powered business solutions
What This Item Tests:
Review data for grounding, including accuracy, relevance, timeliness, cleanliness, and availability
Additional Reading:
Understand data quality and data preparation costs of AI agents
Test and evaluate AI workloads on Azure
Rationale:
Deduplicating and normalizing documents during ingestion ensures that only the most current and relevant content is indexed and retrieved. This reduces the risk of outdated responses and minimizes unnecessary context retrieval that drives up token usage. Fine-tuning does not address frequently changing source content, increasing the context window increases costs without fixing data quality issues, and prompt instructions cannot reliably compensate for stale or duplicate grounding data.
Share This Question
Challenge a friend or share with your study group.