Implement AI solutions by using Microsoft Foundry

A development team runs a nightly build pipeline that must deploy a new version of a language model, then automatically send a fixed set of regression prompts and score the replies, all with no human clicking through any screen. Which Microsoft Foundry surface should the team build this automation against?

• AThe Microsoft Foundry portal, driving each deployment and test prompt manually through its interactive playground each night.
• BAzure Content Understanding, using its extraction schema to deploy the model and grade each regression reply.
• CThe Foundry SDK, calling it from the pipeline to deploy the model and submit the regression prompts programmatically.check_circle Correct
• DAzure Speech, transcribing the regression prompts before the model deploys them on a schedule.

Why A is wrong: The portal is the interactive surface for exploring models and prototyping by hand, which is tempting because it is where a developer first meets a model, but it cannot run an unattended nightly pipeline with no human present.

Why B is wrong: Content Understanding extracts structured information from documents, images, audio and video, so it is tempting when the word 'score' appears, but it does not deploy models or run prompt pipelines.

Why C is correct: The SDK exposes deployment and inference as code that a pipeline can call without any manual interaction, which is exactly what an unattended nightly job requires.

Why D is wrong: Azure Speech handles speech to text and text to speech, which sounds relevant because prompts are involved, but the prompts here are text and Speech cannot deploy a model.

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.

A finance team must process thousands of scanned supplier invoices and pull out the same fields from each one, including the vendor name, invoice total, and every line item, returning them as consistent structured data for an accounting system. Which Foundry capability best meets this requirement?

• AA general multimodal chat model with vision, asked in free text to describe what each invoice image contains.
• BAzure Speech, transcribing the invoices so the field values become available as text.
• CAzure Content Understanding, configured to extract the defined invoice fields as structured output.check_circle Correct
• DAn embeddings model, encoding each invoice into a vector so the fields can be compared numerically.

Why A is wrong: A vision chat model can read an image and is tempting because it sees the invoice, but its free-text description is not reliable consistent structured output across thousands of documents.

Why B is wrong: Azure Speech converts audio to text and is tempting because transcription sounds like extraction, but invoices are documents and images, not audio, so Speech does not apply.

Why C is correct: Content Understanding is built to extract defined fields from documents into consistent structured data, which is exactly what the accounting system needs.

Why D is wrong: Embeddings enable similarity search and are tempting for large document volumes, but a vector does not return named fields like vendor or total as structured data.

A developer is building a hands-free assistant in which a user speaks a request aloud and the application must first turn that spoken audio into a written transcript before any further processing. Which Azure Speech in Foundry Tools capability performs this conversion?

• ASpeech-to-text, which transcribes spoken audio into written textcheck_circle Correct
• BText-to-speech, which generates spoken audio from written input
• CSpeech translation, which converts speech in one language into speech or text in another language
• DSentiment analysis, which scores the emotional tone of a passage of text

Why A is correct: Speech-to-text is the Azure Speech capability that converts spoken audio into a written transcript, which is exactly the first step the assistant needs.

Why B is wrong: Text-to-speech is a genuine Azure Speech capability, but it runs in the opposite direction by producing audio from text rather than transcribing speech into text.

Why C is wrong: Speech translation also processes audio, so it is tempting, but its purpose is crossing languages rather than simply transcribing a request in the same language.

Why D is wrong: Sentiment analysis works on text and reports tone, which is unrelated to turning an audio request into a transcript.