Exam-style question formats

The AI-103 exam is not only four-option multiple choice. It also uses interactive items — the exam policy page calls these interactive components. These sets drill the three most common shapes so the interaction itself doesn't cost you time on exam day.

All-or-nothing

Like multiple-response items, every format on this page scores all-or-nothing — one wrong step, blank, or pair and the whole item is marked incorrect. Slow down and check the full answer before submitting.

Build-list · put the steps in order

Drag a row or use the ▲ ▼ buttons. The whole sequence must be right — there is no partial credit.

Ordering — AI-103 Question 1 / 8 · Score 0/0

RAG pipeline build

You are building a retrieval-augmented generation pipeline over a document library so an agent can answer grounded questions. Put the stages in the order you implement them.

Put the steps in the correct order — drag a row, or use the ▲ ▼ buttons.

1. 1Crack and chunk the source documents, applying OCR where the content is scanned⠿
2. 2Generate embeddings for each chunk and write them to the search index alongside the text⠿
3. 3Configure hybrid retrieval combining keyword and vector search, with semantic ranking⠿
4. 4Pass the top-ranked chunks to the model as grounding context in the prompt⠿
5. 5Evaluate the generated answers for groundedness and relevance⠿

Hotspot · complete each blank

A short scenario with several dropdowns. Every blank must be correct for the item to score.

Hotspot — AI-103 Question 1 / 8 · Score 0/0

Search retrieval modes

An agent's retrieval layer must satisfy three separate requirements. Choose the correct retrieval capability for each.

Choose the correct value for each blank. All must be right to score.

Match a query to passages that mean the same thing but share no keywords— select —Keyword searchVector searchFilter expressionExact-match search

Combine keyword and vector results into a single ranked list— select —Hybrid search with reciprocal rank fusionTwo sequential queriesFaceted navigationScoring profile only

Re-rank the merged top results with a deep model to improve precision— select —Semantic rankerScoring profileSynonym mapAnalyzer

Matching · pair each item with its target

Match every item on the left to the right option. All pairs must be correct.

Matching — AI-103 Question 1 / 8 · Score 0/0

Foundry service selection

Match each requirement to the Azure capability that addresses it.

Match each item on the left to the correct option. All must be right to score.

Index and retrieve grounding passages for a RAG application— select —Azure AI SearchAzure Content UnderstandingAzure AI TranslatorAzure AI Speech

Turn mixed documents, images, and video into structured fields for downstream reasoning— select —Azure AI SearchAzure Content UnderstandingAzure AI TranslatorAzure AI Speech

Translate product documentation into twelve languages with a custom glossary— select —Azure AI SearchAzure Content UnderstandingAzure AI TranslatorAzure AI Speech

Transcribe recorded support calls and synthesize spoken agent replies— select —Azure AI SearchAzure Content UnderstandingAzure AI TranslatorAzure AI Speech

Items you miss here land in your weak spots, and a share of them appears in the timed mock exam.