8 AI-102 Practice Questions That Mirror the Real Exam
Here are 8 multiple-choice questions designed to match the difficulty and style of the Microsoft AI-102 exam, covering key domains like solution design, implementation, and responsible AI.
Question 1
You are designing a solution that uses Azure AI Language to extract key phrases and entities from customer emails. The solution must identify the sentiment of each email and classify it into predefined categories (e.g., "Complaint", "Inquiry", "Praise"). Which two Azure AI Language features should you use? Related reading: 8 AWS CloudOps Engineer Practice Questions That Mirror the Real Exam.
A. Named Entity Recognition (NER) and Key Phrase Extraction
B. Sentiment Analysis and Custom Text Classification
C. Entity Linking and Conversational Language Understanding
D. Language Detection and Summarization
Answer: B
Explanation: Sentiment Analysis provides the sentiment score (positive/negative/neutral), while Custom Text Classification allows you to train a model to classify text into your specific categories (Complaint, Inquiry, Praise). NER and Key Phrase Extraction (A) extract entities and phrases but don't classify into custom categories. Entity Linking (C) and Language Detection (D) are not relevant to sentiment or custom classification.
Question 2
You are building a chatbot using Azure AI Language. The bot needs to understand user intents like "Book a flight" and "Check weather". You have a small set of labeled training data (e.g., 50 examples per intent). You want to minimize development effort and achieve high accuracy quickly. Which approach should you use?
A. Use the pre-built Conversational Language Understanding (CLU) model and fine-tune it with your data.
B. Use Azure AI Bot Service's QnA Maker to map intents to FAQ responses.
C. Use Azure AI Language's pre-configured "Domain" models for general conversation.
D. Use Azure AI Translator to convert user input into a standard format for intent matching.
Answer: A
Explanation: CLU is the correct service for building custom intent and entity models for a chatbot. It is designed to be trained with small amounts of labeled data and provides a straightforward authoring experience. QnA Maker (B) is for question-answering from a knowledge base, not complex intent classification. Pre-built domain models (C) are generic and won't match your specific "Book a flight" intents. Translator (D) is for language translation, not intent recognition.
Question 3
You are deploying a custom Azure AI Document Intelligence model to extract data from invoices. You need to ensure that the model is retrained automatically when new invoice formats are introduced by vendors. What is the most efficient way to handle this?
A. Manually retrain the model each month using a new sample of invoices.
B. Use the "Custom extraction" model type and schedule a retraining job using Azure Machine Learning pipelines.
C. Use a pre-built invoice model and rely on its built-in capabilities to handle all format variations.
D. Create a separate model for each vendor and manually update them via the API.
Answer: B You can explore this further in 8 AWS Developer Associate Practice Questions That Mirror the Real Exam.
Explanation: Azure Machine Learning pipelines can be used to automate the retraining process for Document Intelligence models. This is the most efficient way to handle evolving formats. Manually retraining (A) is not efficient. Pre-built models (C) are for standard layouts and won't handle custom vendor formats well. Creating separate models per vendor (D) is complex and still requires manual updates.
Question 4
You are using Azure OpenAI Service to generate product descriptions. Your application frequently receives the same prompts from users. To reduce cost and improve response latency, what should you implement?
A. Enable content filtering on the Azure OpenAI endpoint.
B. Implement a caching layer using Azure Cache for Redis to store responses for identical prompts.
C. Use the "text-embedding-ada-002" model instead of a generation model.
D. Increase the max_tokens parameter to allow for longer responses.
Answer: B
Explanation: Caching identical prompts and their responses (e.g., in Redis) eliminates the need to call the model repeatedly, reducing both cost and latency. Content filtering (A) is a safety measure, not a performance optimization. Embedding models (C) are for vector search, not text generation. Increasing max_tokens (D) would increase cost and latency, not decrease it.
Question 5
You are building a solution that uses Azure AI Vision to read text from images of handwritten notes. The notes contain a mix of printed and handwritten text. Which API should you use to extract the text?
You can also try our free study plans and readiness checks for the exam you are preparing for.
A. The OCR (Optical Character Recognition) API.
B. The Read API.
C. The Image Analysis API for tags and objects.
D. The Face API for detecting individuals.
Answer: B
Explanation: The Read API is specifically designed for extracting printed and handwritten text from images and documents, with support for mixed text. The OCR API (A) is for printed text only. Image Analysis (C) identifies objects and tags, not text extraction. Face API (D) is for facial recognition, unrelated to text.
Question 6
You are developing a Q&A bot that uses Azure AI Search. The bot must provide answers from a large corpus of PDF documents. You need to ensure that the search index includes the content of these PDFs. What should you do first?
A. Manually upload the PDFs to the search index as JSON documents.
B. Use an indexer to connect to a data source (e.g., Azure Blob Storage) and automatically extract text from the PDFs.
C. Use Azure AI Translator to convert PDFs to text files before indexing. This topic pairs well with 10 AWS Security Specialty Practice Questions That Mirror the Real Exam.
D. Use the Azure AI Language service to summarize the PDFs and index the summaries.
Answer: B
Explanation: An indexer is the correct tool to automatically connect to a data source, extract text from documents (including PDFs), and load it into the search index. Manually uploading (A) is inefficient and error-prone. Translator (C) is for language translation, not text extraction. Summarizing (D) would lose critical detail needed for accurate answers.
Question 7
You are building an application that uses Azure AI Language to detect personally identifiable information (PII) in user-generated content. The application must redact this information before storing the content. You need to ensure that the redaction process is performed automatically as part of the data ingestion pipeline. Which service should you use?
A. Azure AI Content Safety.
B. Azure AI Language PII detection with the redact action.
C. Azure AI Translator with a custom dictionary.
D. Azure AI Document Intelligence to redact text on the fly.
Answer: B
Explanation: Azure AI Language's PII detection feature can be configured to automatically redact detected PII entities in the response. Content Safety (A) is for harmful content, not PII. Translator (C) is for translation. Document Intelligence (D) is for document analysis, not real-time text redaction.
Question 8
You are creating a multi-modal solution that analyzes both images and text to provide insights. You need to use a single Azure AI service that can process both modalities. Which service should you use?
A. Azure AI Vision.
B. Azure AI Language.
C. Azure OpenAI Service (e.g., GPT-4 with vision).
D. Azure AI Search.
Answer: C
Explanation: Azure OpenAI Service, specifically models like GPT-4 with vision capabilities, can process both images and text in a single request. Azure AI Vision (A) is image-only. Azure AI Language (B) is text-only. Azure AI Search (D) is for indexing and search, not multi-modal analysis.
These questions cover a range of core topics you'll encounter on the AI-102 exam, from selecting the right service for a task to understanding implementation details and optimization strategies. Remember to focus not just on what each service does, but when and why to use it.
Skip the uncertainty — Designing and Implementing a Microsoft Azure AI Solution (AI-102) mock exam give you a clear readiness signal before you book.