50 Microsoft Azure AI Fundamentals Practice Questions: Question Bank 2025

You need to determine whether a proposed AI solution could produce unfair outcomes for different demographic groups. Which responsible AI principle is MOST directly related to this concern?

A retailer wants to predict next week's sales for each store based on historical sales data, promotions, and holidays. Which type of machine learning is this?

You want to detect and read printed text from scanned invoices and store the extracted text in a database. Which Azure AI capability best fits this requirement?

A support team wants to automatically identify whether an incoming email expresses a positive, negative, or neutral attitude. Which NLP workload does this describe?

A data scientist in Azure Machine Learning trained a model with great accuracy on training data but much worse performance on new, unseen data. What is the MOST likely issue?

You have a dataset of customer transactions with known outcomes (fraudulent or not fraudulent). You want to build a model in Azure Machine Learning to predict fraud for new transactions. Which approach should you choose?

A company wants to build a chatbot that answers employees’ HR questions by retrieving information from internal policy documents. They want the bot to ground answers in those documents and reduce hallucinations. Which solution is MOST appropriate?

A manufacturing company uses a vision system to spot defective products. The model suddenly starts missing obvious defects after a new camera and lighting setup is installed. What is the MOST likely cause?

You plan to fine-tune a generative language model to follow company-specific style guidelines. You must reduce the chance that training data containing employee personal data can be revealed through the model. Which action best supports this goal?

A healthcare organization is evaluating an AI model that recommends patient triage priority. The organization requires that clinicians can understand the key factors that influenced each recommendation. Which responsible AI principle is MOST relevant?

You are reviewing an AI solution that determines whether loan applications should be approved. You need to identify a key responsible AI consideration for this scenario. What should you focus on first?

You want to store and reuse features (for example, customer tenure, rolling spend, and churn label transformations) across multiple machine learning models in Azure Machine Learning. Which capability best fits this requirement?

You are building a model with Azure Machine Learning and you want to track parameters, metrics, and artifacts for each training run to compare experiments. What should you use?

A retail company wants to extract brand names and prices from images of product shelves. They do not need to classify the whole image, but they need text from many photos captured in stores. Which Azure AI capability should they use?

A support center wants to analyze thousands of customer chat transcripts to identify key topics and sentiment trends. Which Azure AI service is most appropriate?

You deployed an Azure Machine Learning model for real-time inferencing. The endpoint is returning HTTP 503 errors during peak traffic. What is the most likely remediation?

A team wants to implement Retrieval-Augmented Generation (RAG) so a chat application can answer questions using internal policy documents. Which design best matches RAG in Azure?

You built a conversational bot. Users frequently ask follow-up questions like "What about the next day?" and "Does that include shipping?" The bot often responds incorrectly because it does not remember earlier turns. What should you improve?

You are designing a generative AI assistant for employees. The assistant must reduce the risk of leaking sensitive information in prompts and completions. Which approach is the best practice?

A medical imaging team trained a model that performs well in one hospital but poorly in another hospital with different equipment and patient demographics. Which issue is the most likely cause?

A retailer wants to build an AI solution that recommends products to customers based on past purchases and browsing history. Which type of AI workload is this?

You are preparing data for a supervised machine learning model in Azure Machine Learning. Which statement best describes the role of labeled data?

A support team wants a quick way to convert recorded customer calls into text for search and review. Which Azure AI capability best fits this requirement?

A company is training a classification model and notices it performs very well on training data but significantly worse on new, unseen data. What is the most likely issue?

You plan to use Azure AI Vision to extract text from images of invoices. Some invoices are photographed at an angle and have skewed text. Which approach is most appropriate to improve text extraction accuracy?

A team is building a chatbot that should answer questions using a set of internal policy documents. They want responses grounded in those documents rather than general internet knowledge. Which solution pattern best meets this requirement?

An organization must ensure its AI solution provides explanations for automated decisions that affect customers (for example, loan approvals). Which responsible AI principle does this requirement primarily align with?

You deploy a model to production and later discover that customer behavior has changed, causing prediction quality to degrade over time. What is the best action to address this issue?

A developer uses a large language model to generate summaries of customer emails. During testing, some summaries include fabricated details that are not present in the source email. What is the best mitigation strategy?

You are building a multilingual customer-support bot. Users may ask questions in Spanish, French, or English, but the internal knowledge base articles are written only in English. You want the bot to return an answer in the user's language while still grounding responses in the English articles. Which approach is most appropriate?

A team is evaluating whether a new customer-support feature should use AI. The feature must provide a clear, traceable reason for each decision (approve/deny) to satisfy an internal audit. Which AI workload consideration is MOST important?

You have a dataset of product reviews labeled as positive, neutral, or negative. You want to predict the label for new reviews. Which machine learning task is this?

A retail company wants to extract the text from photos of receipts taken with mobile phones. Which Azure AI capability is the BEST fit?

You need to build a chatbot that can interpret user messages and extract structured details such as the city and travel date. Which Azure AI service capability is designed for this?

You trained a model in Azure Machine Learning and it performs well on training data but significantly worse on new, unseen data. Which issue is MOST likely occurring?

A company wants to evaluate an LLM-generated answer for whether it is grounded in the provided source documents and avoids including unsupported claims. Which evaluation concept BEST matches this requirement?

A manufacturer wants to detect defective items on an assembly line using images. Defects are rare, and there are very few labeled examples. Which approach is MOST appropriate?

Your team plans to use a cloud-hosted LLM to draft internal policy summaries. The policies contain sensitive employee information. Which practice BEST helps reduce the risk of exposing sensitive data in prompts and outputs?

You build a Q&A assistant that should answer using your company’s internal documents. Users report the assistant often provides confident answers that are not found in the documents. Which design change MOST directly addresses this problem?

A bank must demonstrate that its loan-approval model does not unfairly disadvantage a protected group. Which responsible AI principle is being addressed MOST directly?

You are advising a team that is building an AI solution to help approve small business loans. Which action best helps address the risk of unfair bias in the model outcomes?

You want to build a model that predicts the expected delivery time (in minutes) for food orders based on distance, restaurant load, and weather. What type of machine learning is this?

You need to extract printed text from photos of receipts taken by a mobile app. Which Azure AI capability is the most appropriate?

A data scientist trains a model in Azure Machine Learning and wants to ensure they can reproduce results later, including code, data references, environment, and metrics. What should they do?

You build an NLP model that classifies customer emails into categories. The model performs well in testing, but after deployment the accuracy drops because email writing style changes over time. What is the most likely cause?

A call center wants to automatically transcribe calls and then identify the key topics discussed (for example, "billing", "cancellation", "technical support"). Which approach best fits this requirement?

You are building a chatbot that must answer questions using only information from the company’s internal policy documents and should minimize unsupported claims. Which design is most appropriate?

You want to evaluate a text-generating model’s outputs for harmful content and policy violations before showing them to end users. What is the best practice in Azure AI solutions?

A team uses a large language model to summarize confidential internal emails. They must ensure the solution does not leak sensitive data and has strong access controls and auditability. Which combination best addresses these requirements?

You deployed an object detection model and notice it frequently misses small objects at a distance in surveillance footage. Which change is most likely to improve detection of small objects?