50 Microsoft Certified: Azure AI Fundamentals Practice Questions: Question Bank 2025

A retail company wants to automatically answer common customer questions such as store hours and return policy on its website. The solution should handle natural language questions. Which Azure AI service is the best fit?

You need an AI solution that predicts whether a customer is likely to cancel a subscription next month based on historical customer behavior. What type of machine learning is this?

You are designing an AI system to help prioritize loan applications. Which action best supports responsible AI principles related to fairness and transparency?

A manufacturing company wants to detect defects in product images on an assembly line. Which computer vision workload is being performed?

A data scientist trains a model in Azure Machine Learning and wants to measure how well it performs on unseen data before deployment. Which approach is most appropriate?

An app must extract the total amount and vendor name from scanned receipts (images) and output structured fields. Which Azure service is best suited?

You build a chatbot that should identify whether a user message is about 'Billing', 'Technical support', or 'Sales'. The chatbot will then route the conversation accordingly. Which NLP capability is this?

A team is deploying an AI model for medical triage assistance. Which practice best addresses the 'reliability and safety' principle of responsible AI?

You are using a Large Language Model (LLM) to generate customer-support responses. Users report the model occasionally invents policies that do not exist. What is the most appropriate mitigation strategy?

A company wants to train a custom object detection model to identify specific machinery parts in images. They have 2,000 labeled images but labels are inconsistent between annotators. What is the biggest risk, and what is the best first step?

You are reviewing an AI solution used to screen loan applications. Which principle best helps ensure the model does not discriminate against protected groups?

You need to store training data that contains customer names and email addresses. Which action is the best practice to help protect the data while training a model in Azure?

You need to detect the language of short text messages submitted by users. Which Azure service capability is the best fit?

A team has a dataset with 1,000 labeled images of defective and non-defective products. They want to train a model without writing code. Which approach should they choose in Azure Machine Learning?

You trained a classification model in Azure Machine Learning. In production, you notice the distribution of incoming feature values has shifted significantly from the training data, and accuracy is dropping. What is this issue called?

A company wants to extract key-value pairs (for example, InvoiceNumber and TotalAmount) from invoices with varying layouts. Which Azure AI capability should they use?

You are designing a chatbot for internal IT helpdesk. The bot must answer questions using company policy documents and cite the source passages. Which approach best meets this requirement?

A model achieves 95% accuracy on a dataset where 95% of the examples are 'non-fraud' and 5% are 'fraud'. Stakeholders care most about catching fraud cases. Which metric is most important to review?

A computer vision model used for workplace safety must identify whether workers wear hard hats. The model performs well in testing but fails at night and in poorly lit areas. What is the best next step to improve performance?

You must deploy an Azure AI service that will only accept requests from your organization’s private network and must not be reachable from the public internet. Which configuration best meets the requirement?

You are designing an AI solution to help agents decide whether to approve or reject loan applications. Regulatory requirements state that applicants must be given an understandable reason for any rejection. Which consideration is MOST important to include in the solution design?

A team is building an image classification model in Azure Machine Learning. They have 2,000 labeled images, and many images include multiple objects, but only one label per image is available. Which type of learning problem are they primarily solving?

You need to extract printed text from scanned PDF pages and return the recognized text to an application. Which Azure service capability best fits this requirement?

A data scientist trained a model in Azure Machine Learning and notices the model performs significantly better on the training dataset than on the test dataset. What is the MOST likely issue?

You must identify whether product photos contain a company logo and highlight its location in the image. Which approach is MOST appropriate?

A company wants to build a customer support chatbot that answers questions using information from internal policy documents. They need the bot to respond in natural language and base answers on the provided documents. Which solution is the BEST fit?

You are evaluating a binary classification model that predicts whether a transaction is fraudulent. The business wants to minimize missed fraud cases, even if it means investigating more legitimate transactions. Which metric should you prioritize?

An organization collected customer feedback through a web form. They want to automatically redact personally identifiable information (PII) such as names, phone numbers, and email addresses before storing text for analytics. Which Azure capability should they use?

A team trains a model to predict equipment failure. In production, the model’s accuracy degrades over time even though the code and model version have not changed. The input sensor patterns are shifting due to new equipment calibration. What is the MOST likely cause?

You build an NLP pipeline to categorize incoming emails into 10 departments. After deployment, users report that emails in Spanish are frequently misrouted, while English emails are categorized correctly. The training data was almost entirely English. What is the BEST action to improve performance?

A team wants to build an AI solution that provides explanations for its predictions to help business users trust the results. Which principle is being emphasized?

You need to extract key phrases and detect the language of customer support tickets stored as text. Which Azure service is designed for this task without building a custom ML model?

A model predicts whether a patient is at high risk (Yes/No). Which machine learning task does this represent?

A retailer wants to identify the coordinates of multiple products within a single image (for example, boxes around each product). Which computer vision capability is most appropriate?

You train a model that performs well in training but poorly on new, unseen data. Which issue is the most likely cause?

A solution uses a prebuilt Azure OpenAI chat model to answer employee questions about internal policies. You must ensure responses are grounded only in your policy documents. What should you implement?

You build a dataset to classify emails as spam or not spam. The dataset contains 98% not spam and 2% spam. Which evaluation metric is typically more informative than accuracy for this scenario?

A manufacturing company wants to detect rare defects in images of products. They have many images of normal products and very few images of defects. Which approach is most appropriate?

You build a conversational bot that uses Azure AI Language question answering over a curated knowledge base. During testing, the bot gives confident but incorrect answers when it cannot find a relevant match. What is the best mitigation?

A bank wants to use an AI model to help approve loans. Regulators require proof the model does not unfairly disadvantage protected groups. What should you use in Azure Machine Learning to assess and help mitigate this risk?

You want to classify incoming emails as "spam" or "not spam". Each training example is labeled with the correct category. What type of machine learning is this?

You need an Azure AI solution that extracts printed text from scanned invoices to populate a database. Which workload is the best fit?

A chatbot must answer common questions using a curated list of answers written by your support team. Which NLP capability is most appropriate?

You are building a model to predict house prices. Your evaluation shows a low training error but a much higher test error. What is the most likely issue?

A team is building an AI system to assist with loan approvals. They must be able to explain which input factors most influenced a specific decision. What responsible AI principle does this requirement most directly support?

You have a dataset that contains 200,000 product images across 500 categories. Training locally is slow, and you want to scale training and track metrics and model versions. Which Azure service is best suited?

A retailer wants to monitor store entrances and count how many people enter per hour. Which computer vision capability best matches this requirement?

You are analyzing customer reviews to automatically identify whether each review is positive, negative, or neutral. Which NLP task is this?

You are training a binary classification model where 99% of the records are "non-fraud" and 1% are "fraud". Accuracy is 99% but the model misses most fraud cases. Which metric is most important to review to better understand performance on fraud detection?

A company wants to process multilingual customer messages and route them to the correct department. They need to first detect the language, then extract key information like order numbers and product names. Which approach best fits this requirement using Azure NLP capabilities?