Medium AI-901 practice questions

You are building a single-agent solution in the Microsoft Foundry portal for a retail company. The agent must answer customer questions and, when asked about product availability, look up live stock levels from the company's inventory REST API. Which capability should you configure for the agent so it can retrieve this real-time data?

A media company wants to automatically produce written descriptions of the visual content in its archive of video clips, including identifying objects, actions, and scenes so editors can search the archive by content. Which AI workload category best addresses this requirement?

A company wants to build a lightweight app that captures live audio from an in-person staff meeting through a microphone and displays a running text transcript on screen as people speak. The developers plan to use Azure Speech in Foundry Tools. Which Speech capability should they use to meet this requirement?

A developer wants to add a chat feature to an internal Python web app. A GPT model is already deployed in an Azure AI Foundry project. The developer needs the app to send user messages to the deployed model and receive responses programmatically, without building any infrastructure from scratch. Which approach best meets this requirement?

A developer at a logistics company has already created and tested a single-agent solution in the Azure AI Foundry portal that answers shipment-tracking questions. They now want to embed this agent's capabilities into their internal web application so employees can interact with it programmatically, without opening the Foundry portal. What is the most appropriate way to accomplish this?

A software company receives thousands of customer support emails daily. They want to automatically route each email to the correct department by determining which of five predefined categories (Billing, Technical, Sales, Legal, or General) best describes the email content. Which AI capability should they use to accomplish this?

A financial services company builds an AI assistant that answers customer questions about investment products. Compliance requires that whenever the assistant provides an answer, it must clearly indicate which internal documents the information came from and note that the response was generated by AI. Which responsible AI principle does this requirement primarily support?

A city government is deploying an AI-powered chatbot to help residents access public services. The team wants to ensure the chatbot works effectively for residents with visual impairments, limited literacy, and those who speak languages other than the region's primary language. Which responsible AI principle is the team primarily addressing?

A bank deploys a generative AI assistant to answer customer questions about account fees. During testing, the team notices that the same question sometimes produces conflicting fee amounts, which could mislead customers. Which responsible AI principle is the team primarily addressing when they work to ensure the assistant returns consistent, dependable answers?

A development team is building an internal knowledge search tool. They want user queries and documents to be converted into numeric vector representations so that documents with similar meaning to a query can be retrieved, even when they do not share the exact same words. Which type of model should they use to generate these representations?

A software development team wants to integrate an AI capability into their internal developer portal that generates working code snippets and functions from plain-language descriptions typed by engineers. Which type of generative AI model is best suited for this requirement?

A marketing analyst wants to automatically group thousands of customers into segments based on similarities in their purchasing behavior, without providing any predefined labels or categories. Which type of machine learning workload best fits this requirement?

You are building a single-agent solution in the Microsoft Foundry portal for a finance team. Users will paste raw quarterly sales numbers into the chat and ask the agent to calculate growth percentages, produce a bar chart, and return the results. The agent's language model cannot reliably perform arithmetic or generate images on its own. Which capability should you enable for the agent so it can run calculations and create the chart?

A media company is deploying a generative image model that lets marketing staff create promotional artwork from text prompts. Leadership is concerned that the model could occasionally produce harmful, violent, or otherwise offensive imagery from ambiguous prompts. Which responsible AI principle is most directly addressed by adding content filters that block such outputs?

An e-commerce company wants to automatically detect when two product listings submitted by different sellers describe essentially the same item, even when the wording differs (for example, 'wireless noise-cancelling headphones' vs 'cordless sound-blocking earphones'). Which AI technique is best suited to measure how closely the meaning of two text passages matches?

A medical imaging startup is building an AI system that flags suspicious lesions in X-ray images. The team decides that whenever the model's prediction confidence falls below a set threshold, the case is automatically routed to a human radiologist instead of being auto-classified. Which responsible AI principle does this design choice primarily support?

You are building a single-agent travel-booking assistant in Microsoft Foundry. The agent must be able to check real-time seat availability by calling your company's existing internal REST API during a conversation. Which capability should you configure on the agent to enable this?

A logistics company wants to build a lightweight information-extraction app that pulls the shipment ID, carrier name, and delivery date from thousands of scanned bill-of-lading documents. Before running any documents through Azure Content Understanding in Foundry Tools, what must the team define so the service knows which pieces of information to return?

A media company records weekly podcast episodes and wants to automatically build a searchable catalog. For each audio file, they need to extract structured data such as the episode title, list of guest names, and the main topics discussed. They want to use a prebuilt Foundry capability designed to pull structured information from unstructured content, including audio, without training a custom machine learning model. Which Foundry Tool should they use?

A logistics company receives a mix of shipping documents each day: some are scanned bills of lading, others are packing slips, and some are delivery photos taken by drivers. The team wants to build a lightweight information-extraction app in Azure AI Foundry that pulls structured fields (such as tracking number, weight, and destination) from all of these inputs. Which Foundry Tools capability should they use as the foundation of this app?

A legal operations team receives hundreds of scanned PDF contracts each week. They need to build a lightweight app that automatically pulls specific fields — such as the counterparty name, effective date, and total contract value — into a structured format for their database. Which Azure AI Foundry capability should they use to define the fields and extract this information from the documents?

A consumer electronics company receives thousands of photographed warranty registration cards from customers. Each card contains printed and handwritten fields such as product serial number, purchase date, and store name. The team wants to build a lightweight app in Microsoft Foundry that automatically pulls these specific field values into structured data for their database. Which Foundry capability should they use?

A media company records training webinars and needs an automated pipeline that extracts the presenter's spoken talking points, on-screen slide titles, and key visual moments from each recorded video file. Which Azure AI capability in Foundry Tools should the team use to build this single information-extraction solution?

A media company wants to build a lightweight app that automatically extracts spoken dialogue, on-screen text, and key visual moments from uploaded promotional videos so editors can quickly find relevant clips. The team wants to use a single Azure AI service in Foundry Tools that handles all these modalities from the video files. Which service should they use?

A logistics company wants to build a lightweight information-extraction app using Azure Content Understanding in Foundry Tools to pull specific fields (shipment ID, destination, weight) from thousands of scanned bill-of-lading documents. Before the app can process any documents, what must the team do first in the Content Understanding workflow?