AI-102 Exam Guide 2026: The Azure AI Engineer Credential Built Around Azure OpenAI and Generative AI
In 2026 there is no Microsoft role-based certification more in flux - or more worth sitting - than Exam AI-102: Designing and Implementing a Microsoft Azure AI Solution. Microsoft has revised the AI-102 skills outline multiple times since 2024 to add Azure OpenAI Service, retrieval-augmented generation (RAG), Azure AI Foundry (formerly Azure AI Studio), Azure AI Search (formerly Azure Cognitive Search), and agentic/generative workflows. The exam that existed in 2022 covered Cognitive Services and Bot Framework; the 2026 exam covers those plus a full generative AI chapter that candidates coming from older study guides completely miss.
This guide is written exclusively for the current 2026 exam window. It maps every skill domain against the April 2026 skills measured outline, calls out the Azure OpenAI / RAG material that used to be absent, and gives you an 8-week plan, hands-on lab checklist, and career/salary context for the Azure AI Engineer Associate credential.
AI-102 Exam At-a-Glance (2026)
| Item | Detail (2026) |
|---|---|
| Full Name | Exam AI-102: Designing and Implementing a Microsoft Azure AI Solution |
| Credential Earned | Microsoft Certified: Azure AI Engineer Associate |
| Delivery | Pearson VUE (online-proctored via OnVUE or test center) |
| Questions | ~40-60 items (multiple choice, multi-select, drag-and-drop, case studies, performance-based labs) |
| Time Limit | 100 minutes of exam time (~120 minutes total seat time) |
| Passing Score | 700 out of 1000 (scaled) |
| Exam Fee | $165 USD (varies by country; India ~$55, UK £113) |
| Prerequisites | None required; AI-900 + Python or C# strongly recommended |
| Languages | English, Japanese, Chinese (Simplified), Korean, German, French, Spanish, Portuguese (Brazil), Arabic (Saudi Arabia), Italian, Indonesian |
| Certification Validity | 1 year; renew FREE via Microsoft Learn assessment (6-month renewal window) |
| Retake Policy | 24-hour wait after 1st fail; 14-day wait for attempts 2-5; max 5 attempts per 12 months |
| Current Skills Outline | Revised multiple times 2024-2026 to add Azure OpenAI, Foundry, and RAG content |
| Related | AI-900 (AI Fundamentals), DP-100 (Data Scientist), AZ-305 (Solutions Architect), DP-700 (Fabric Data Engineer) |
Source: Microsoft Learn exam page (learn.microsoft.com/credentials/certifications/exams/ai-102), AI-102 skills measured outline, and Pearson VUE scheduling portal.
Start Your FREE AI-102 Prep Today
Who Should Sit AI-102
The Microsoft skills outline targets developers who build AI solutions using Azure services, including the Azure OpenAI Service, Azure AI services (formerly Cognitive Services), Azure AI Search, Azure AI Document Intelligence, Azure AI Vision, Azure AI Language, Azure AI Speech, and the Azure AI Foundry (SDK + portal) for generative and agentic workloads.
| Candidate Profile | Why AI-102 Fits |
|---|---|
| Python or C# developers adding AI to existing apps | Core target audience - most exam items are SDK-based |
| Full-stack engineers building RAG apps on Azure OpenAI | Generative AI chapter covers the exact patterns you ship |
| ML engineers with Databricks/PyTorch experience | Validates the Azure-specific deployment, security, and responsible AI layer |
| Cloud engineers moving into AI roles | Adds AI to your Azure credential stack (complements AZ-204, AZ-305) |
| Data engineers integrating AI with pipelines | Pairs with DP-700 for Fabric + AI Skill + embeddings |
| Solution architects designing Azure AI platforms | Provides hands-on depth beyond AI-900 fundamentals |
Recommended Prerequisites (Not Required)
AI-102 has no formal prerequisites, but Microsoft strongly recommends candidates complete or have equivalent experience with:
- Exam AI-900 (Azure AI Fundamentals) - the conceptual foundation
- 1-2 years developing in Python or C# - both languages appear in code-read items
- Familiarity with REST APIs, JSON, authentication patterns (API key, Entra ID / Managed Identity)
- Basic Azure portal, Azure CLI, and Bicep/ARM navigation
- Conceptual understanding of machine learning (training vs inference, embeddings, vector search)
If you are brand new to Azure AI, sit AI-900 first (2 weeks of prep, $99) before starting on AI-102.
Build AI-102 Mastery with FREE Practice Questions
AI-102 Skills Measured (2026 Domain Weights)
The current Microsoft Learn skills outline for AI-102 splits the exam into six weighted domains. Microsoft has historically updated AI-102 more frequently than any other role-based exam - four to five outline revisions since the exam launched. The 2026 revision reflects the maturation of Azure OpenAI and the rebranding of Cognitive Services into Azure AI services, and the introduction of Azure AI Foundry.
| Domain | 2026 Weight | What It Covers |
|---|---|---|
| 1. Plan and manage an Azure AI solution | 15-20% | Resource selection, security, responsible AI, monitoring, container deployment, cost |
| 2. Implement decision support solutions | 10-15% | Azure AI Content Safety, Anomaly Detector (deprecated), personalized recommendations |
| 3. Implement computer vision solutions | 15-20% | Azure AI Vision (image analysis, OCR/Read, Face), Video Indexer, Custom Vision |
| 4. Implement natural language processing solutions | 15-20% | Azure AI Language (NER, sentiment, PII, summarization), Translator, Speech (STT/TTS/translation) |
| 5. Implement knowledge mining and document intelligence solutions | 15-20% | Azure AI Search (vector + hybrid), Document Intelligence (prebuilt + custom models) |
| 6. Implement generative AI solutions | 10-15% | Azure OpenAI Service, prompt engineering, RAG, Azure AI Foundry, agents, fine-tuning |
IMPORTANT 2026 CAVEAT: Microsoft adjusts these weights periodically and has trended the generative AI domain higher. The 10-15% listed for Domain 6 is often the minimum - in practice, many candidates report that generative AI items appear across Domain 1 (security/responsible AI) and Domain 5 (RAG-based knowledge mining) as well. Always verify the current weights on learn.microsoft.com/credentials/certifications/exams/ai-102 within two weeks of your exam date.
Source: Microsoft AI-102 Study Guide, current 2026 revision.
Domain 1: Plan and Manage an Azure AI Solution (15-20%)
This domain is where DP-900/AI-900 alumni often under-prepare. It is architecture and operations, not model code.
Select the appropriate Azure AI service
- Choose between single-service vs multi-service (Azure AI services) resources for cost and key management.
- Decide between Azure AI services, Azure OpenAI, Azure Machine Learning, and Azure AI Foundry based on workload (prebuilt vs custom vs generative).
- Understand regional availability - Azure OpenAI is restricted to specific regions (East US, East US 2, Sweden Central, West Europe, Japan East, and others that rotate).
Plan, configure, and secure an AI solution
- Configure authentication: key-based vs Microsoft Entra ID (formerly Azure AD) with Managed Identity and Key Vault.
- Implement private endpoints and VNet integration for data exfiltration protection.
- Apply Customer-Managed Keys (CMK) with Key Vault for encryption at rest.
- Enable diagnostic logging to Log Analytics, Event Hub, or storage.
- Rotate keys programmatically and use Key Vault references in apps.
Implement responsible AI
- Apply the Microsoft Responsible AI Standard v2 six principles (fairness, reliability/safety, privacy/security, inclusiveness, transparency, accountability).
- Configure content filtering in Azure OpenAI (default low-severity thresholds, custom categories, jailbreak detection, protected material detection).
- Use Azure AI Content Safety for image/text moderation (hate, sexual, violence, self-harm) and prompt shield.
- Implement responsible AI metrics and human review workflows.
Manage, monitor, and optimize cost
- Use Azure Monitor and Application Insights for telemetry.
- Track token usage and throughput units (PTUs) in Azure OpenAI.
- Choose between pay-as-you-go and Provisioned Throughput Units (PTUs) for Azure OpenAI.
- Understand deployment options: Standard (shared capacity), Provisioned Managed, Global Standard, Global Batch.
Deploy Azure AI services in containers
- Deploy eligible services (Text Analytics, Language, Translator, Speech, Document Intelligence, AI Vision Read) as Docker containers on-prem or in Azure Kubernetes Service for data residency.
Domain 2: Implement Decision Support Solutions (10-15%)
The smallest domain. Microsoft has been phasing out Anomaly Detector (retirement announced October 2026), so most current items focus on Azure AI Content Safety and content moderation.
- Azure AI Content Safety: text moderation (hate/sexual/violence/self-harm categories with severity 0/2/4/6), image moderation, Prompt Shields (jailbreak + indirect prompt injection detection), Protected Material detection (text copyright + code from public repos), Groundedness detection (hallucination detection).
- Personalizer (in Limited Access as of 2026) - contextual bandit reinforcement learning for content ranking.
- Anomaly Detector - understand its role and retirement timeline if it still appears on items; Microsoft has guided migration toward custom time-series models on Azure Machine Learning or Azure AI Foundry.
Domain 3: Implement Computer Vision Solutions (15-20%)
Azure AI Vision (formerly Computer Vision)
- Image Analysis 4.0 (built on Florence foundation model): captioning, dense captioning, OCR/Read, object detection, people detection, background removal, smart crops, image retrieval via multimodal embeddings.
- Read API / OCR 4.0 for printed + handwritten text.
- Face API with Limited Access (requires approved registration) for detection, verification, identification, grouping.
- Spatial Analysis (container-only workload) for people counting, social-distancing analytics.
Custom Vision
- Build custom classification (multi-class or multi-label) and object detection models with a small training set (~50+ images per tag).
- Export models to ONNX, TensorFlow, CoreML, Docker for edge deployment.
- Understand iterations and performance metrics (precision, recall, mAP).
Azure AI Video Indexer
- Extract insights from video (speakers, topics, OCR, faces, sentiment, keywords, labels).
- Use custom models for language, brands, people.
- Integrate via widget embedding or API-driven pipelines.
Domain 4: Implement Natural Language Processing (15-20%)
Azure AI Language (formerly Text Analytics + LUIS + QnA Maker)
- Prebuilt features: sentiment analysis, opinion mining, key phrase extraction, language detection, PII detection + redaction, Named Entity Recognition (NER), text summarization (extractive + abstractive), healthcare text analytics.
- Custom models: custom NER, custom text classification (single + multi label), Conversational Language Understanding (CLU) (the successor to LUIS), Custom Question Answering (successor to QnA Maker), orchestration workflow for chaining CLU + QA.
Azure AI Translator
- Text translation (100+ languages), document translation, custom translator for domain-specific glossaries, transliteration.
Azure AI Speech
- Speech-to-Text (STT): real-time, batch, fast transcription, custom speech models, pronunciation assessment.
- Text-to-Speech (TTS): neural voices, custom neural voice (Limited Access), SSML for prosody.
- Speech translation: speech-to-speech translation across 50+ languages.
- Speaker recognition (Limited Access) for voice biometric identification/verification.
Note: LUIS is retired (April 1, 2026). Any study guide mentioning LUIS as current content is out of date. Use Conversational Language Understanding (CLU) instead.
Domain 5: Implement Knowledge Mining and Document Intelligence (15-20%)
Azure AI Search (formerly Azure Cognitive Search)
- Create indexes, indexers, data sources, skillsets for AI enrichment.
- Built-in cognitive skills: OCR, key phrase extraction, entity recognition, translation, image analysis.
- Custom skills: call Azure Functions or custom APIs from a skillset.
- Vector search: create vector fields, embed documents via Azure OpenAI text-embedding-3-large / text-embedding-ada-002, perform hybrid search (keyword + vector), use semantic ranker for reranking.
- Integrated vectorization (GA 2024-2025): one-step index creation with automatic embedding via Azure OpenAI.
- Security trimming using user group membership or Entra ID.
Azure AI Document Intelligence (formerly Form Recognizer)
- Prebuilt models: invoices, receipts, ID documents, business cards, W-2s, 1098s, 1099s, health insurance cards, US/UK bank checks, pay stubs, contracts.
- Layout model for tables + structure.
- Custom models: custom template (form-based with fixed layout) vs custom neural (variable layout, signature detection, checkboxes).
- Composed models to route documents through multiple custom models.
- Train with as few as 5 labeled samples (template) or ~100+ samples (neural).
Domain 6: Implement Generative AI Solutions (10-15%) - The Highest-Value Chapter
This is the hottest and most-tested area of AI-102 in 2026, even though its nominal weight is 10-15%. Generative AI concepts routinely bleed into Domains 1, 5, and even 3 (multimodal image generation and understanding).
Azure OpenAI Service
- Provision and deploy models: GPT-4o, GPT-4o mini, GPT-4 Turbo, GPT-3.5 Turbo, o1 / o3-mini reasoning models, embeddings (text-embedding-3-large, text-embedding-3-small, ada-002), DALL-E 3, Sora (video - Limited Access), Whisper (speech-to-text).
- Choose a deployment type: Standard, Global Standard (lowest latency across regions), Provisioned Managed (PTU - reserved capacity), Global Batch (24-hour async jobs at 50% cost).
- Apply content filtering (hate, sexual, violence, self-harm with severity thresholds, plus prompt shields and protected material detection).
- Integrate via Azure OpenAI SDK (Python, C#, Java, JavaScript) or REST; use Entra ID with Managed Identity for keyless authentication.
Prompt engineering
- System messages for role priming, grounding, few-shot examples.
- Parameter tuning: temperature (0-2), top_p, presence_penalty, frequency_penalty, max_tokens, stop sequences, seed (for deterministic replays).
- Chain-of-thought (CoT) and reasoning prompts (especially for o1/o3 models).
- Function / tool calling: JSON schema definitions, tool choice, parallel calls.
- Structured outputs via JSON mode or strict JSON schema.
Retrieval-Augmented Generation (RAG)
- End-to-end pattern: chunk documents -> embed -> index in Azure AI Search -> retrieve with hybrid search + semantic ranker -> pass top-K chunks into a grounded prompt -> generate response with citations.
- "Add your data" (on your data) feature in Azure AI Foundry to wire RAG with a few clicks against Azure AI Search, Azure Blob, Azure Cosmos DB, URL, or Microsoft Fabric OneLake.
- Chunking strategies: fixed size, sentence-based, semantic chunking.
- Evaluation: groundedness, relevance, coherence, fluency metrics; prompt flow evaluation in Azure AI Foundry.
Azure AI Foundry (formerly Azure AI Studio)
- Create hubs and projects for AI development teams.
- Use the model catalog (Azure OpenAI + Meta Llama + Mistral + Cohere + custom) with serverless API deployments.
- Build agents with the Azure AI Agent Service (function calling + tool integration + code interpreter + file search + Bing grounding).
- Build prompt flows (visual DAG for prompt orchestration with evaluation + deployment).
- Monitor with AI-assisted metrics and safety evaluators.
Fine-tuning
- Supervised fine-tuning (SFT), DPO (direct preference optimization), continuous fine-tuning on GPT-4o, GPT-4o mini, GPT-3.5 Turbo, o4-mini (when available).
- Data format: JSONL with messages arrays.
- Evaluate vs base model, deploy as dedicated model.
Azure OpenAI Quick Reference (High-Yield for 2026)
| Concept | What to Know |
|---|---|
| PTU vs Standard | PTU = reserved throughput (predictable latency, no rate limits, higher base cost); Standard = pay-per-token shared capacity with TPM/RPM limits |
| Global Standard | Routes requests to any region with capacity; lowest latency, same per-token price as Standard |
| Global Batch | Async file-based jobs, 24-hour SLA, 50% discount - ideal for summarization backfills |
| Content filters | Default on; configurable by severity (low/medium/high block) per category; customer-configured filters require application approval |
| Prompt Shield | Detects user-injected jailbreaks and indirect prompt injection in grounded documents |
| Protected Material | Detects generation of copyrighted song lyrics / book text (text) or GitHub-sourced code (code) |
| Groundedness | Checks whether a response is supported by provided sources; emits per-sentence citations |
| Entra ID auth | Preferred over API keys; use DefaultAzureCredential or Managed Identity |
| Responsible AI Limited Access | Face (liveness/ID verification), Custom Neural Voice, Speaker Recognition, Personalizer require approved registration |
Languages and SDKs on the Exam
AI-102 is an SDK-heavy exam. You will see code read and code-fill items in:
- Python (
azure-ai-textanalytics,azure-ai-vision-imageanalysis,azure-ai-documentintelligence,openai(Azure OpenAI),azure-search-documents,azure-ai-language-conversations,azure-ai-ml). - C# / .NET (
Azure.AI.OpenAI,Azure.AI.Vision.ImageAnalysis,Azure.AI.Language,Azure.AI.DocumentIntelligence,Azure.Search.Documents).
You do not need to write code on a blank screen. You need to recognize the correct SDK call, parameter, or authentication pattern among multiple choices.
Example Code Snippet (Azure OpenAI - Python)
from openai import AzureOpenAI
from azure.identity import DefaultAzureCredential, get_bearer_token_provider
token_provider = get_bearer_token_provider(
DefaultAzureCredential(),
"https://cognitiveservices.azure.com/.default",
)
client = AzureOpenAI(
azure_endpoint="https://my-aoai.openai.azure.com/",
api_version="2024-10-21",
azure_ad_token_provider=token_provider,
)
response = client.chat.completions.create(
model="gpt-4o", # deployment name
temperature=0.2,
response_format={"type": "json_object"},
messages=[
{"role": "system", "content": "Return strictly valid JSON."},
{"role": "user", "content": "Extract invoice fields from: ..."},
],
)
Example Code Snippet (Azure AI Search - Hybrid Vector Query, Python)
from azure.search.documents import SearchClient
from azure.search.documents.models import VectorizedQuery
vector_query = VectorizedQuery(
vector=embedding, # 3072-dim from text-embedding-3-large
k_nearest_neighbors=10,
fields="content_vector",
)
results = search_client.search(
search_text="privacy policy for minors",
vector_queries=[vector_query],
query_type="semantic",
semantic_configuration_name="default",
top=5,
)
8-Week AI-102 Study Plan (Working Developer)
This plan assumes ~8-10 hours per week and that you already know Python or C# plus basic Azure.
| Week | Focus | Deliverable |
|---|---|---|
| Week 1 | Azure AI services overview, resource creation, auth (key + Entra ID), Key Vault, private endpoints | Deploy an Azure AI services multi-service + Azure OpenAI resource with Managed Identity |
| Week 2 | Azure AI Vision (Image Analysis 4.0, OCR, Face), Custom Vision, Video Indexer | Build an image classifier + OCR pipeline on sample receipts |
| Week 3 | Azure AI Language (NER, sentiment, PII, summarization), Custom NER, CLU, Custom QA, Speech, Translator | Build a CLU bot + custom QA knowledge base |
| Week 4 | Document Intelligence (prebuilt + custom template + custom neural), composed models | Train a custom neural model for mixed-layout invoices |
| Week 5 | Azure AI Search (vector + hybrid), embeddings, integrated vectorization, semantic ranker | Index 500 docs, run hybrid semantic vector query |
| Week 6 | Azure OpenAI + prompt engineering + function calling + structured outputs | Build a JSON-schema-grounded assistant with tool calls |
| Week 7 | RAG end-to-end: Azure AI Foundry "on your data" + prompt flow + evaluation + Agent Service | Build + evaluate a RAG chatbot with groundedness metrics |
| Week 8 | Responsible AI, content filters, cost/PTU, full-length timed mocks, remediation | Score consistently >75% on two timed mocks before test day |
Time Allocation (Match the Blueprint)
| Domain | Weight | Share of Study Time |
|---|---|---|
| Computer vision | 15-20% | 17% |
| Natural language processing | 15-20% | 17% |
| Plan and manage AI solution | 15-20% | 16% |
| Knowledge mining / document intelligence | 15-20% | 16% |
| Generative AI (includes RAG) | 10-15% (often higher in practice) | 22% |
| Decision support | 10-15% | 12% |
Recommended AI-102 Resources (FREE-First)
| Resource | Type | Why It Helps |
|---|---|---|
| OpenExamPrep AI-102 Practice (FREE) | Free, unlimited | Scenario items mapped to the current 2026 skills outline with AI explanations |
| Microsoft Learn AI-102 learning path | Free | Official modules across all six domains with sandbox labs |
| Microsoft AI-102 Study Guide (aka.ms/AI102-StudyGuide) | Free PDF | Authoritative list of skills measured; print and tick off |
| Azure AI Foundry free portal access | Free | Experiment with model catalog, prompt flow, agent service without a subscription for initial playground |
| Azure free tier ($200 credit + 12 months) | Free-ish | Required for hands-on; most AI service tiers include a free monthly quota |
| John Savill's Technical Training (YouTube) | Free | Whiteboard explanations of Azure OpenAI, PTUs, RAG, Agent Service |
| Tim Warner - AI-102 Cert Prep (YouTube + Pluralsight) | Free + paid | Domain-by-domain walkthroughs aligned to outline |
| Microsoft AI-102 GitHub labs (MicrosoftLearning/mslearn-ai) | Free | Hands-on Python/C# labs for every domain |
| Azure OpenAI samples (Azure-Samples/openai) | Free | RAG, function calling, evaluation, fine-tuning reference apps |
| Azure-Samples/azure-search-openai-demo | Free | Canonical RAG reference implementation |
| Microsoft AI Skills Challenge | Free | Periodic campaigns award a 50% or 100% exam voucher |
| MeasureUp AI-102 Practice Tests | Paid ($129) | Closest to the official item style; full-length timed mocks |
| Pluralsight / A Cloud Guru AI-102 course | Paid | Video-led structure if you prefer that format |
| Scott Guthrie / Mark Russinovich Azure AI talks | Free (YouTube) | Strategic context for exam-level design trade-offs |
Hands-On Labs Are Non-Negotiable
AI-102 is not a memorization exam. Microsoft has moved aggressively toward case-study and performance-based items. Candidates who only watch videos routinely fail. Use the Azure free tier ($200 credit + 12 months of limited free services) or your employer's subscription to build:
- An Azure OpenAI chat app with Entra ID auth, content filters, function calling, and structured JSON outputs.
- An Azure AI Search index populated via integrated vectorization from a Blob container of 100+ PDFs.
- A RAG app using Azure AI Foundry's "on your data" wired to that search index, with prompt flow evaluation for groundedness.
- A Document Intelligence custom neural model trained on mixed-layout invoices, composed with a prebuilt receipt model.
- A Custom Vision object detection model exported to ONNX.
- An Azure AI Language CLU + Custom Question Answering orchestration bot deployed to a web app.
- A Speech translation app using the Speech SDK with a custom neural voice for TTS output.
If you have built all seven, you will recognize every SDK call on the exam.
Common Pitfalls That Sink First-Time Scores
- Studying an old (pre-2024) outline. Guides that still focus on LUIS, QnA Maker (old), Anomaly Detector as first-class, or skip Azure OpenAI entirely are out of date. Confirm the outline on learn.microsoft.com within 2 weeks of your exam.
- Under-studying Azure OpenAI. Even with a nominal 10-15% weight for Domain 6, generative AI content surfaces inside Domains 1 (responsible AI, content filtering), 5 (RAG), and more. Budget 25% of study time here.
- Skipping Azure AI Search vector + hybrid + semantic. Integrated vectorization, text-embedding-3-large, hybrid queries, and semantic ranker configuration are high-yield in Domain 5.
- Confusing custom template vs custom neural in Document Intelligence. Template = fixed layout, ~5 samples. Neural = variable layout, ~100 samples, supports signatures and selection marks. Exam items pick the cheaper/faster option when the layout allows.
- Missing Entra ID / Managed Identity auth. Key-based auth is still valid but Microsoft's exam philosophy rewards Managed Identity + Entra ID + Key Vault for production readiness.
- Ignoring Limited Access registrations. Face (verification/identification), Custom Neural Voice, Speaker Recognition, Personalizer require Microsoft approval. Items test whether you know the registration requirement.
- Case study time management. Case studies embed 2-4 pages of context plus 4-6 linked questions. Read the question first, skim the case for the specific detail; do not re-read the entire case twice.
- Weak on PTU vs Standard vs Global Standard vs Global Batch. Cost/latency/capacity trade-offs appear in every real exam report. Know when each deployment type wins.
- Not knowing content filter tiers and Prompt Shield. Default low-severity blocking, customer-configurable severity, Prompt Shield vs Protected Material vs Groundedness. Easy points if memorized.
- No timed full-length practice. 100 minutes across 40-60 items including case studies is tight; two timed mocks minimum before test day.
Test-Day Logistics and Strategies
Before you sit:
- Confirm your Microsoft Learn Profile matches your government ID exactly.
- If online-proctored, run the OnVUE system check 24 hours ahead; a single camera/mic issue can cost your slot.
- Clear your desk. Proctors will ask for a 360-degree room scan.
During the exam:
- You can flag and review within a section, but cannot return to prior sections once submitted.
- Case studies appear as standalone sections. Read the question first, skim the case for the relevant detail.
- When two answers look defensible, pick the Microsoft-native, managed-service option with least operational overhead (e.g., integrated vectorization beats hand-rolled embedding code; Managed Identity beats rotating keys).
- Target ~2 minutes per standalone item, ~4-5 minutes per case-study item.
After the exam:
- Pass/fail and scaled score (0-1000, pass = 700) displayed immediately.
- Skill-area breakdown emailed within 1-3 business days.
- If you fail: 24-hour wait for first retake, then 14-day wait for attempts 2-5, max 5 per 12 months.
Career Impact and Salary (Azure AI Engineer, 2026)
The AI engineering labor market is one of the tightest in tech. Microsoft-stack AI engineers with hands-on Azure OpenAI and RAG experience command a strong premium in 2026.
| Source (2026) | Azure AI Engineer Pay |
|---|---|
| Glassdoor (US, "Azure AI Engineer") | Median total comp ~$150,000/yr; range $120K-$190K |
| Levels.fyi (AI Engineer, Microsoft stack) | Entry $110K-$140K; mid $140K-$180K; senior $180K-$260K+ |
| Dice.com tech salary report (AI/ML) | Average $155,000/yr for AI/ML engineers with cloud certs |
| LinkedIn Talent Insights (Azure AI, US) | 25,000+ open roles citing Azure AI / Azure OpenAI; 55%+ YoY growth |
| Robert Half Tech Salary Guide 2026 | AI engineer range $130K-$200K with certification premium of 10-15% |
Azure AI Career Ladder
| Role | Typical 2026 US Pay | Next Step |
|---|---|---|
| Junior AI / ML Engineer (Azure) | $95K-$130K | AI-102 + AI-900 + 1-2 yrs hands-on |
| Azure AI Engineer (mid) | $130K-$180K | AI-102 + DP-100 + team lead scope |
| Senior Azure AI Engineer | $180K-$240K | AI-102 + AZ-305 + architecture ownership |
| Principal / Staff AI Engineer | $230K-$350K+ | Platform leadership in AI product teams |
| AI Solutions Architect | $200K-$300K | AI-102 + AZ-305 + consulting delivery |
Deep Dives on Five Topics Competitor Guides Skim
1. Azure OpenAI Deployment Types (Standard vs PTU vs Global)
Azure OpenAI offers four deployment tiers in 2026, and knowing when each wins is the single highest-yield Domain 1 / Domain 6 cross-over concept.
- Standard (regional): pay-per-token, shared regional capacity, rate-limited by TPM (tokens per minute) and RPM (requests per minute). Good for dev/test and bursty small apps.
- Global Standard: pay-per-token at the same rate as Standard, but Microsoft routes to any region with capacity. Lowest latency for most workloads; the default recommendation for production apps without strict data residency.
- Provisioned Managed (PTU): reserve throughput in PTUs (each PTU = a fixed slice of per-minute capacity); predictable latency, no rate limiting, higher base cost but lower per-token effective rate at high utilization. Good for SLA-bound production apps.
- Global Batch: async file-based jobs, 50% discount, 24-hour SLA. Ideal for document summarization backfills, embedding generation jobs, data labeling.
2. Retrieval-Augmented Generation (RAG) End-to-End
RAG is the highest-impact generative pattern on AI-102. The canonical pipeline:
- Ingest documents into Azure Blob, Cosmos DB, Fabric OneLake, or SharePoint.
- Chunk using fixed size (512-1024 tokens) or semantic chunking. Overlap 10-20%.
- Embed each chunk with
text-embedding-3-large(3072 dims) ortext-embedding-3-small(1536 dims). Integrated vectorization does this automatically. - Index in Azure AI Search with keyword + vector + semantic fields.
- Retrieve top-K (typically 5-10) via hybrid query (BM25 + vector + semantic ranker).
- Compose a grounded prompt with retrieved chunks as context and explicit citation instructions.
- Generate with GPT-4o / GPT-4o mini.
- Evaluate groundedness, relevance, coherence, fluency via prompt flow metrics.
Azure AI Foundry's "on your data" feature wires steps 4-7 automatically when pointed at an AI Search index, Blob container, Cosmos DB account, URL, or Fabric OneLake.
3. Content Filtering, Prompt Shields, and Groundedness
Azure OpenAI ships four families of safety checks:
- Content filters across hate, sexual, violence, self-harm with severity 0/2/4/6; blocked at low severity by default. Customer-configured filters (adjusting thresholds) require application approval.
- Prompt Shield - detects user-injected jailbreaks and indirect prompt injection within retrieved documents (RAG context).
- Protected Material detection - flags generation of copyrighted song lyrics/book text (text) or GitHub-sourced code (code).
- Groundedness detection - verifies whether a response is supported by provided sources, with per-sentence citations.
Azure AI Content Safety is the standalone service that exposes these capabilities plus image moderation and custom category models.
4. Azure AI Foundry, Agents, and Prompt Flow
Azure AI Foundry (rebranded from Azure AI Studio) is the unified workspace for AI development:
- Hubs organize security, networking, and shared resources at the team/org level.
- Projects are the unit of AI development (prompt flows, evaluations, agents, deployments).
- Model catalog exposes Azure OpenAI, Meta Llama, Mistral, Cohere, Phi, and custom registered models with one-click serverless or managed deployment.
- Prompt flow is a visual DAG-based tool for prompt orchestration, evaluation, and deployment. Nodes call LLMs, Python scripts, Azure AI Search, etc.
- Azure AI Agent Service provides managed agents with function calling, code interpreter, file search, and Bing grounding - the successor to Assistants API with enterprise security.
5. Fine-Tuning vs Prompt Engineering vs RAG
Candidates frequently over-apply fine-tuning. The exam rewards picking the right tool:
- Prompt engineering (system prompts + few-shot) solves the vast majority of behavior shaping. Free and instant to iterate.
- RAG solves knowledge freshness and grounding - the model cites recent proprietary data it was never trained on.
- Fine-tuning solves style, tone, structure, and token efficiency - when you need a compressed instruction footprint or a specific output format. Use SFT or DPO on GPT-4o mini / GPT-3.5 Turbo / GPT-4o. Requires high-quality JSONL training data.
The canonical exam answer to "how should I reduce hallucinations about our internal HR policies" is RAG, not fine-tuning. "How should I enforce a strict company-specific JSON schema output" can go either way but often starts with structured outputs + prompt engineering before fine-tuning.
How AI-102 Fits into the Broader Microsoft AI / Data Certification Path
| Exam | Role | When to Sit |
|---|---|---|
| AI-900 Azure AI Fundamentals | Entry | Foundational prerequisite (recommended) |
| AI-102 Azure AI Engineer Associate | AI Engineer | This exam - core developer credential |
| DP-100 Azure Data Scientist Associate | Data scientist | Pair with AI-102 if you train custom models |
| DP-900 Azure Data Fundamentals | Entry | Optional foundation |
| DP-700 Fabric Data Engineer Associate | Data engineer | Pair with AI-102 for Fabric + AI (AI Skill, embeddings) |
| AZ-305 Azure Solutions Architect Expert | Architect | After AI-102 + 2-3 yrs, for architecture scope |
| AI Agent Engineering (future) | Agents | Rumored; monitor Microsoft Learn roadmap |
AI-102 pairs especially well with DP-100 (custom ML on Azure ML) and DP-700 (Fabric + AI Skill) for a full Microsoft-stack AI practitioner profile.
Renewal and Continuing Competency
Like all Microsoft role-based certifications, the Azure AI Engineer Associate is valid for one year. Renewal is FREE via Microsoft Learn - a browser-based assessment opens 6 months before expiration. It is typically 25-35 questions covering what has changed in the Azure AI surface area (new Azure OpenAI models, Foundry features, Content Safety capabilities, AI Search features) since your initial pass. There is no proctoring; you can retake renewals unlimited times.
Do not let it lapse. An expired credential cannot be renewed and you would need to re-sit the full AI-102 at $165. Given Azure AI's velocity of change, renewal is genuine continuing education - plan ~4-6 hours before the renewal assessment to review Microsoft Learn's AI-102 "What's changed" guidance.
Total Cost of AI-102 Certification (2026)
| Item | Cost | Notes |
|---|---|---|
| Exam fee (US) | $165 | Varies by country |
| Azure free tier | $0 ($200 credit + free quotas) | Required for hands-on; plan for ~$50-$100 of Azure OpenAI spend if you do all labs |
| Microsoft Learn training | $0 | Free official path + sandbox labs |
| OpenExamPrep practice | $0 | Free scenario bank |
| MeasureUp practice tests (optional) | $129 | Closest to real item style |
| Pluralsight / A Cloud Guru (optional) | ~$29-$49/month | Video-led structure |
| Instructor-led bootcamp (optional) | $1,500-$3,500 | Microsoft partner training |
| Renewal | $0 | Free assessment on Microsoft Learn yearly |
| Typical all-in first-time cost | $165-$400 | Self-study + free tier |
Azure Free Tier Strategy for AI-102 Labs
Azure's free tier is generous but has gotchas for AI workloads:
- Azure AI services (F0 tier) - free monthly quotas for Text Analytics, Vision, Language, Translator, Speech. Sufficient for most AI-102 labs.
- Azure OpenAI - no always-free tier. Use the $200 Azure credit for the first 30 days, then pay-as-you-go. Budget ~$5-$20 for GPT-4o mini experimentation across the full study period; GPT-4o full is more expensive, save for final integration labs.
- Azure AI Search - free tier (50 MB index, 3 replicas, 3 partitions) supports small RAG labs. Upgrade to Basic ($75/month) only if you need vector > 50 MB; you can also delete/recreate free-tier services between labs.
- Document Intelligence F0 - 500 pages/month free on prebuilt; custom models require S0 (paid).
- Content Safety F0 - 5K transactions/month free.
- Application Insights - free tier is ample for monitoring labs.
Rule of thumb: plan for $30-$100 of total Azure spend across the 8-week study period if you use GPT-4o mini and free tiers aggressively.
Keep Training with FREE AI-102 Practice
Frequently Missed 2026 Details (Competitor Guides Get These Wrong)
- LUIS is retired (April 1, 2026). Use Conversational Language Understanding (CLU) and orchestration workflow.
- QnA Maker (old) is retired. Use Custom Question Answering in Azure AI Language.
- Cognitive Services rebranded to Azure AI services. Same APIs, new resource kind; the "multi-service" resource is now "Azure AI services".
- Azure AI Studio rebranded to Azure AI Foundry (late 2024/early 2025). Hubs and projects are the new mental model.
- text-embedding-3-large / text-embedding-3-small are the current recommended embedding models; ada-002 is legacy but still supported.
- Integrated vectorization in Azure AI Search (GA 2024-2025) automates embedding via Azure OpenAI skill inside a skillset.
- Anomaly Detector has a retirement timeline; if it appears, know it is being phased out in favor of custom Azure ML / Foundry solutions.
- Azure OpenAI Global Batch gives 50% off for async workloads - high-yield cost question.
- Prompt Shield is the official Microsoft term for jailbreak + indirect prompt injection detection.
- Groundedness detection (Content Safety) and groundedness metrics (prompt flow evaluation) are distinct features - know both.
- Azure AI Agent Service (successor to Assistants API) ships managed agents with function calling + code interpreter + file search + Bing grounding.
Official Sources Used
- Microsoft Learn - Exam AI-102: Designing and Implementing a Microsoft Azure AI Solution (skills outline, 2026 revision)
- Microsoft Learn - AI-102 Study Guide (aka.ms/AI102-StudyGuide)
- Microsoft Certified: Azure AI Engineer Associate credential page
- Azure AI services documentation (learn.microsoft.com/azure/ai-services)
- Azure OpenAI Service documentation and model availability pages
- Azure AI Search documentation (vector + hybrid + semantic ranker)
- Azure AI Foundry documentation
- Microsoft Responsible AI Standard v2
- Pearson VUE Microsoft exam scheduling portal (fee and retake policy)
- Microsoft Learn credential renewal policy (6-month window, free assessment)
- Glassdoor / Levels.fyi / Dice / Robert Half - 2026 salary references
- LinkedIn Talent Insights - Azure AI job demand signals
Certification details, fees, and skills measured may be revised by Microsoft. Always confirm current requirements directly on learn.microsoft.com before scheduling.