AI-103: Information Extraction
Information Extraction is one of the skill areas tested on the Developing AI Apps & Agents on Azure (AI-103) exam. Below are free AI-103 practice questions with worked answers and a concept diagram — each with a plain-language explanation. Practice the first 10 questions of AI-103 free, no signup.
The concept, in one picture
2 free AI-103 Information Extraction questions
Scenario: An accounts team reviews supplier invoices by hand to confirm that the figures match the signed supplier agreements. The invoices come in many formats with embedded tables, logos, and inconsistent layouts, and the reviewing solution must interpret both the visual arrangement and the written content of each invoice. You need to recommend a solution that removes the manual review burden. What should you include in the recommendation?
- ✓ Azure Content Understanding in Foundry Tools
- chat completions
- Azure Document Intelligence in Foundry Tools
- Image Analysis
Azure Content Understanding targets multimodal document workloads. It ingests unstructured and multimodal content, including documents and images, and produces structured output for AI apps, using analyzers to pull text, layout, tables, fields, and relationships from varied document types, combining visual and textual understanding for downstream verification. Chat completions alone will not reliably extract structured invoice fields; Document Intelligence covers OCR, layout, and tables but Content Understanding is the stronger end-to-end Foundry option here; Image Analysis works at the image-feature level and cannot handle invoice fields and tables.
You run a Microsoft Foundry project that pulls scanned PDF invoices from a blob container. The invoices use table layouts with printed line items, and their extracted data is saved as structured JSON to ground a RAG agent. You must build one analyzer that: - Pulls the invoice number, invoice date, vendor name, and total amount even though templates differ. - Emits per-field confidence so any result under 0.80 can be sent to a reviewer. What should you use?
- the Azure Content Understanding in Foundry Tools prebuilt-layout analyzer
- a Foundry agent with groundedness guardrails enabled to extract the invoice fields and confidence scores
- ✓ a custom Azure Content Understanding in Foundry Tools analyzer whose schema defines the required fields as extracted fields and returns confidence scores for routing
- the Azure Content Understanding in Foundry Tools prebuilt-documentSearch analyzer with search.score from Azure AI Search results used for routing
The task needs structured field extraction across differing invoice templates, not just OCR or layout. Azure Content Understanding analyzers combine extraction, AI analysis, and structured output, and you can build a custom analyzer for specific fields. A custom schema defining invoice number, invoice date, vendor name, and total amount returns structured JSON for RAG grounding, and each field carries a 0-to-1 confidence score, so a 0.80 threshold becomes the routing rule. The prebuilt-layout analyzer keeps layout but defines no invoice fields; groundedness guardrails judge generated answers rather than extraction; and Azure AI Search search.score reflects retrieval relevance, not extraction confidence.
Practice AI-103 free
The first 10 questions of every exam are free. No signup, no email wall.
Start practicing →Get a free AI-103 study plan by email
A short plan to work through AI-103 by skill area, plus a note when we add new questions. Optional — the practice above stays free. No spam, unsubscribe anytime.