AI-103: Developing AI Apps and Agents on Azure

By Microsoft · Question Mode

QuestionQ37

VerifiedImplement information extraction solutions

You have a Microsoft Foundry project that ingests scanned PDF invoices stored in Azure Blob Storage. Each invoice contains printed line items and uses a table-based layout.

The extracted results are stored as structured JSON and used as grounding data for an agent in a Retrieval Augmented Generation (RAG) solution.

You need to create a single analyzer that meets the following requirements:

  • Extracts the invoice number, invoice date, vendor name, and total amount across varying templates
  • Returns confidence scores so that results with confidence below 0.80 can be routed for supervisor review

What should you use?

Explanation

A custom Content Understanding analyzer lets you define the exact schema of fields to extract — invoice number, invoice date, vendor name, and total amount — regardless of the invoice template, and enabling estimateFieldSourceAndConfidence (or estimateSourceAndConfidence per field) causes the analyzer to return a confidence score for each extracted field. Those scores can then be compared against the 0.80 threshold to route low-confidence results for supervisor review. The prebuilt-layout analyzer only extracts generic layout elements (text, tables) without the named business fields, prebuilt-documentSearch plus Azure AI Search's search.score reflects retrieval relevance rather than field-extraction confidence, and a Foundry agent with groundedness guardrails evaluates response grounding, not per-field extraction confidence.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!