Advanced

Exam Day Tips & Resources

Your final preparation guide: a quick-reference review sheet, test-taking strategies, time management tips, frequently asked questions, and curated links to official Microsoft resources.

Quick Reference Review Sheet

Review these high-priority concepts the night before your exam:

Azure AI Services (15-20%)

• Multi-service resource = one key for all services; single-service = separate key and billing
• Managed Identity > Key Vault > subscription keys (security hierarchy)
• Private endpoints = strongest network isolation
• Containers = on-premises processing, still need Azure for billing
• Disconnected containers = fully offline, periodic billing reconciliation
• Two keys for zero-downtime rotation
• az cognitiveservices account create = CLI command

Computer Vision (15-20%)

• Image Analysis v4.0 = unified analyze endpoint with multiple VisualFeatures
• Custom Vision: Compact domain required for edge/offline export
• Face API: Verification (1:1) vs. Identification (1:N with PersonGroup)
• Read API = OCR for images and PDFs; Document Intelligence = structured extraction (invoices, receipts)
• Spatial Analysis = edge container, people counting, video processed locally
• High precision + low recall = model is accurate but misses many positives; add more diverse training data

NLP Solutions (25-30%)

• CLU (formerly LUIS) = intents + entities; always populate the None intent
• Text Analytics: sentiment, key phrases, NER, PII detection, entity linking, language detection
• PII Detection = detect AND redact; NER = detect only
• Question Answering replaces QnA Maker; part of Azure AI Language
• Document Translation = preserves formatting; Text Translation = plain text only
• Custom Translator = domain-specific translation models with parallel documents
• SSML = control pronunciation, pauses, rate, pitch, emphasis
• Phrase lists = quick fix for speech recognition of specific terms
• Custom Text Classification: single-label (one category) vs. multi-label (multiple categories)

Knowledge Mining (15-20%)

• Pipeline: Data Source → Indexer → Skillset → Index + Knowledge Store
• Built-in skills: OCR, NER, Key Phrases, Sentiment, Translation, Image Analysis
• Custom Web API Skill = call Azure Functions during indexing
• Knowledge Store projections: tables (Power BI), objects (JSON), files (images)
• Shaper skill = create projection shapes for knowledge store
• Hybrid search = keyword + vector; outperforms either alone
• Output field mappings = connect skillset outputs to index fields

Generative AI (10-15%)

• Azure OpenAI uses deployment names, not model names
• RAG = retrieval + generation; for adding knowledge to responses
• Fine-tuning = changing model behavior/style, not adding facts
• temperature: 0 = deterministic, 1+ = creative
• frequency_penalty reduces repetition; presence_penalty increases topic diversity
• JSON mode: response_format: {"type": "json_object"} + mention JSON in prompt
• Content filters: hate, sexual, violence, self-harm; cannot be fully disabled
• Provisioned Throughput (PTUs) = guaranteed capacity for production

Test-Taking Strategy

Time Management

You have 120 minutes for 40-60 questions. Here is how to manage your time:

Phase	Time	Action
First pass	80 minutes	Answer all questions you are confident about. Mark uncertain ones for review.
Case studies	20 minutes	Read scenarios carefully. You cannot go back to case studies once completed.
Review	20 minutes	Return to marked questions. Change answers only if you have a strong reason.

Question Strategy

1. Read the entire question before looking at answers. Identify what is being asked.
2. Eliminate obviously wrong answers first. Usually 1-2 options are clearly incorrect.
3. Look for Microsoft's best practice. The exam rewards the recommended Azure approach, not just any working solution.
4. Watch for absolute words like "always", "never", "only". These are often incorrect.
5. When two answers seem correct, choose the more secure, managed, or Azure-native option.
6. Do not overthink. Your first instinct is usually correct. Only change answers with clear justification.

Common Exam Traps

• LUIS vs. CLU — The exam uses "Conversational Language Understanding" (CLU). LUIS is the predecessor. If you see LUIS, it is likely a distractor.
• QnA Maker vs. Question Answering — QnA Maker is deprecated. The correct answer is "Question Answering" (part of Azure AI Language).
• Form Recognizer vs. Document Intelligence — Same service, new name. The exam may use either. "Azure AI Document Intelligence" is the current name.
• Cognitive Services vs. Azure AI Services — Azure AI Services is the current umbrella name. The exam may use either.
• Deployment name vs. model name — Azure OpenAI uses deployment names in API calls. This is a frequent exam question.

Frequently Asked Questions

How long is the AI-102 exam?

120 minutes for 40-60 questions. You also get additional time for the NDA agreement and post-exam survey, but these do not count toward your 120 minutes.

What is the passing score?

700 out of 1000. Microsoft uses a scaled scoring model, so the exact number of correct answers needed varies by exam form. Approximately 70% correct is a safe target, but aim for 80%+ in your practice to account for the scaling.

Can I take the exam online from home?

Yes. You can take it online via Pearson VUE or at a physical test center. For online exams, you need: a private room, webcam, microphone, stable internet (minimum 3 Mbps), and a government-issued ID. No second monitors, phones, or notes are allowed.

What if I fail? Can I retake it?

Yes. There is a 24-hour waiting period before your first retake. If you fail a second time, you must wait 14 days. The fee ($165) applies each time. There is no limit on the number of attempts per year.

Do I need hands-on Azure experience?

Strongly recommended. The exam includes code snippets (Python SDK, REST API calls), portal screenshots, and architecture scenarios. Microsoft Learn provides free sandbox environments where you can practice without an Azure subscription.

How long is the certification valid?

The Azure AI Engineer Associate certification is valid for 1 year. Microsoft sends a renewal reminder 6 months before expiration. Renewal is a free online assessment (not a full exam) that you can take on Microsoft Learn.

Are there exam discounts available?

Yes. Microsoft offers discounts through: Microsoft Learn Cloud Skills Challenges (often free exam vouchers), Microsoft Certified Trainer program, student pricing, and Microsoft Imagine Academy. Check the Microsoft Learn website for current promotions.

What programming language appears on the exam?

Primarily Python, with some C# examples. You do not need to write code, but you must be able to read and understand SDK calls, REST API requests, and JSON configurations. Know the Azure AI SDKs for Python.

Should I memorize Azure CLI commands?

Know the high-level commands (e.g., az cognitiveservices account create) but you do not need to memorize every flag. The exam focuses more on choosing the right service and configuration than exact CLI syntax.

How does the AI-102 relate to other Azure certifications?

AI-102 (AI Engineer Associate) focuses on implementing AI solutions with Azure AI Services. Related certifications: AZ-900 (Azure Fundamentals) is a good prerequisite, AI-900 (AI Fundamentals) covers AI concepts at a beginner level, and DP-100 (Data Scientist Associate) focuses on training custom ML models with Azure ML.

Official Microsoft Resources

Use these resources to supplement your preparation:

• Microsoft Learn AI-102 Learning Path — Free, self-paced modules covering every exam domain with hands-on exercises
• AI-102 Exam Skills Outline — The official document listing every topic that can appear on the exam (download from the certification page)
• Microsoft Learn Sandbox — Free Azure sandbox environments for hands-on practice (no Azure subscription needed)
• Azure AI Services Documentation — Official API references, quickstarts, and tutorials for every service
• Azure OpenAI Service Documentation — Models, deployment, prompt engineering, and RAG patterns
• Azure AI Search Documentation — Indexers, skillsets, vector search, and knowledge stores
• Microsoft Official Practice Assessment — Free practice questions on the AI-102 certification page

Course Complete

Congratulations on completing the Azure AI Engineer Associate (AI-102) Exam Prep course! You have covered all five exam domains, practiced with 50+ questions, and have a solid study strategy. When you are ready, schedule your exam and earn your certification.