Microsoft Certified: Azure AI Fundamentals (AI-901) practice questions

A developer has finished prototyping a chat experience in the Microsoft Foundry portal and now needs to embed the same deployed model inside a C# web service so that inference requests are issued programmatically at runtime. Which Foundry surface is designed for this?

• AThe Foundry SDK, which exposes client libraries so an application can authenticate to and call the deployed model from its own code.check_circle Correct
• BThe Foundry portal, which is a browser-based workspace for building and testing projects interactively rather than for embedding calls in application code.
• CAzure Content Understanding, which extracts fields from documents and media rather than providing programmatic access to a deployed chat model.
• DThe model catalogue, which lists available models to browse and deploy but does not itself issue runtime inference requests from application code.

Why A is correct: The Foundry SDK provides client libraries for languages such as C# and Python, letting application code authenticate and send inference requests to a deployed model at runtime, which is exactly what embedding the model in a web service requires.

Why B is wrong: The portal is tempting because that is where the prototyping happened, but it is an interactive browser workspace for building and testing, not a mechanism for issuing calls from within a compiled application.

Why C is wrong: Azure Content Understanding is an information-extraction capability for documents, images, audio, and video, so it does not give an application programmatic access to a deployed conversational model.

Why D is wrong: The model catalogue helps you find and deploy a model, but it is a discovery surface, not the runtime client an application uses to call the model once deployed.

A team wants a solution that, given the goal "reconcile this invoice", will decide which steps to take, call a lookup tool to fetch purchase-order data, and then update a record, all without a person choosing each step. Which type of AI workload does this describe?

• AGenerative AI, which produces new content such as text or images in response to a prompt but does not plan or invoke tools to act on a goal.
• BAgentic AI, in which an agent plans a sequence of steps and takes actions such as calling tools or knowledge sources to complete a task.check_circle Correct
• CText analysis, which extracts information such as sentiment or key phrases from existing text rather than taking actions to complete a task.
• DComputer vision, which interprets images and video rather than orchestrating steps and tool calls towards a goal.

Why A is wrong: Generative AI is a close trap because agents are built on generative models, but generation alone only returns content for a prompt; it does not plan a sequence of steps or call tools to change external state.

Why B is correct: The defining trait of agentic AI is that the system plans steps and takes actions towards a goal, including invoking tools and knowledge sources, which matches deciding steps, calling a lookup tool, and updating a record autonomously.

Why C is wrong: Text analysis reads and characterises existing text, so it could describe the invoice but cannot plan steps, call a lookup tool, or update a record to complete the goal.

Why D is wrong: Computer vision interprets visual input; it does not plan multi-step actions or invoke tools, so it does not fit a task defined by autonomous reconciliation of data.

In a chat application built on a deployed Foundry model, a developer wants to set the assistant's persistent role and behaviour rules, such as "you are a legal assistant that answers only in British English", so that they apply to every turn regardless of what the end user types. Where should this instruction be placed?

• AIn the user prompt, since that field carries the end user's per-turn input rather than the assistant's standing role.
• BIn an embedding vector, since embeddings represent the meaning of text numerically and cannot hold behavioural rules.
• CIn the system prompt, which establishes the model's role and standing behaviour for the whole conversation.check_circle Correct
• DIn the model deployment name, since that value only identifies which deployed model is called and cannot carry behavioural rules.

Why A is wrong: The user prompt does carry text to the model, which makes it tempting, but it represents each turn of end-user input and is not the place to fix behaviour that must persist across the whole conversation.

Why B is wrong: Embeddings numerically represent the meaning of text for search and comparison; they are not a channel for instructing a model how to behave, so they cannot carry persistent role rules.

Why C is correct: The system prompt sets the assistant's role and standing rules for the entire conversation, so persistent behaviour such as acting as a legal assistant answering only in British English belongs there.

Why D is wrong: The deployment name simply identifies which deployed model an application calls; it is an identifier, not an instruction channel, so it cannot hold behavioural guidance.