You are designing a solution that will answer questions about human resources (HR) policies stored...

Comprehensive and Detailed 150 to 250 words of Explanation From Azure AI Apps and Agents Developer/Course Guide/topics:

Use Azure Language in Foundry Tools , specifically custom question answering . Microsoft explicitly identifies custom question answering as appropriate when an organization has static information , including content imported from PDFs, and when different users asking the same question should receive the same answer . Microsoft Learn

The HR policy PDFs can be imported into a custom question-answering project. The service extracts or allows administrators to author question-and-answer pairs, alternate question forms, metadata, and multi-turn relationships. After the project is published, a user question is matched against the curated knowledge base and the service returns the best predefined answer and a confidence score.

Azure OpenAI is inherently generative and probabilistic. Lowering temperature can reduce variability, but it does not provide the same deterministic curated-answer behavior. Azure Document Intelligence can extract text and structure from PDF files but does not itself supply the conversational Q & A layer. Azure Machine Learning would require substantially more custom development.

Therefore, B is the current Microsoft-documented choice. The supplied document shows A, but Microsoft Learn specifically states that custom question answering should be used when the same question must return the same answer. Microsoft Learn

====================