Free AI-901 Practice Test Questions and Answers (2026)

The Azure AI Document Intelligence (formerly Form Recognizer) Python SDK uses a long-running-operation pattern. You submit the PDF with beginanalyzedocument() / beginanalyze() which immediately returns an LROPoller object. When processing is finished you call poller.result() (or poller.wait()) to obtain the extracted field values. Option A reflects exactly this required two-step pattern.

B. Extraction results are not returned in HTTP headers; you must poll the LRO for the JSON payload.

C. Manually invoking OCR and field mapping is unnecessary; the service already performs layout, key-value, and table extraction.

D. There is no synchronous analyze() that returns results in the same request; heavy analysis always runs as an LRO.

1. Microsoft Docs – Azure AI Document Intelligence Python SDK, “Quickstart: Analyze a document”, code block using poller = documentanalysisclient.beginanalyzedocument(...); result = poller.result() (v3.2.0). https://learn.microsoft.com/en-us/python/api/overview/azure/ai-formrecognizer-readme?view=azure-python#example-analyze-a-document

2. Microsoft Docs – REST API specification, “Analyze Document – This operation is long running … result is retrieved by GET on the operation-location URL”. (Document Intelligence REST API v2023-07-31).

3. Microsoft Learn Module “Extract key-value pairs with Document Intelligence”, Unit: Submit and retrieve results – explains LRO and poller.result pattern (Microsoft Learn collection AZ-AI-900).

You need a service that can read a document, map each detected key/value or table cell to a field you have explicitly declared, and return a hierarchical (nested) JSON object. Azure Content Understanding analyzers in Foundry Tools are built on Azure AI Document Intelligence (formerly Form Recognizer); they let you upload a schema describing every required field (including repeated or nested collections such as invoice line-items) and then run an Analyzer job that returns the populated structure. Neither OCR-only pipelines, Speech transcription, nor Azure AI Search provide schema-driven, nested field extraction.

A. Speech transcription converts spoken audio to text; it has no capability for document parsing or field-level extraction.

B. OCR returns unstructured text and coordinates; it does not map results to a user-defined schema or support nested data.

D. Azure AI Search indexes existing text for searchability; it does not analyze page images nor extract field/value pairs according to a schema.

1. Microsoft Learn – “Build a custom document model with Azure AI Document Intelligence”, section “Define fields and object hierarchy” (learn.microsoft.com, last paragraph before “Train the model”).

2. Azure AI Document Intelligence REST API v3.2 – “Analyze Invoices” response schema shows nested lineItems array (docs.microsoft.com/rest/api/document-intelligence/2023-07-31-preview).

3. Microsoft Learn – “Choose OCR or Document Intelligence for document solutions”, table 1: ‘Structured field extraction’ only supported by Document Intelligence (learn.microsoft.com).

4. Microsoft Docs – “Speech-to-text overview” §“What is speech transcription?” (no document analysis features) (learn.microsoft.com/azure/ai-services/speech-service).

5. Microsoft Docs – “What is Azure AI Search?” §“Core capabilities” (search, ranking; no field extraction) (learn.microsoft.com/azure/search).

Converting written text into lifelike spoken audio is the Text-to-Speech function of Azure Speech. In Azure documentation this capability is formally called speech synthesis. Text sent to the Speech service is synthesized into natural-sounding audio streams that can be routed to telephony systems, IVRs, or any other playback device, exactly matching the requirement in the scenario.

A. Speaker recognition identifies or verifies who is speaking; it does not create audio from text.

C. Speech recognition converts spoken audio into text (speech-to-text), the opposite direction required here.

D. Speech translation takes spoken audio in one language and outputs translated text or speech; it still starts with speech input, not written text.

1. Microsoft Azure AI Speech documentation – “Text-to-speech (speech synthesis) overview”, Paragraph 1 & 2, learn.microsoft.com/azure/ai-services/speech-service/text-to-speech (accessed 2026-05-26).

2. Microsoft Azure AI Speech documentation – “Compare speech capabilities”, Table “Speech synthesis”, learn.microsoft.com/azure/ai-services/speech-service/overview#speech-capabilities (shows text→speech role).

3. Microsoft Azure AI Speech documentation – “Speaker recognition”, Section “What is speaker recognition?”, learn.microsoft.com/azure/ai-services/speech-service/speaker-recognition (demonstrates purpose distinct from synthesis).

Azure OpenAI vision models accept images only in the content array of a user-role message. Microsoft’s request schema shows a single user message whose content array mixes a text item (type="text") and one imageurl item (type="imageurl"). Putting both modalities in the same request eliminates an extra round-trip and therefore yields the lowest latency. Images are explicitly disallowed in system messages, and sending the image in a second request would double network time, violating the “quickest response” requirement.

A. Two separate requests incur extra latency and complicate state management; slower than one mixed message.

B. Docs prohibit image content in system messages; request will be rejected.

C. Same two-request latency issue as A and still places image in disallowed system context.

1. Microsoft Learn – “Vision models in Azure OpenAI Service: request format”, Example JSON shows role:"user" with both text and imageurl in one content array (https://learn.microsoft.com/azure/ai-services/openai/vision/how-to) section “Send an image”.

2. Azure OpenAI REST API reference, Chat Completions – “Images must be supplied within a user role message; system messages may contain text only” (https://learn.microsoft.com/azure/ai-services/openai/reference#chat-completions).

3. Microsoft Build 2024 Session OD446 – slide 18 “Best performance: single user message containing text + imageurl” (public conference deck hosted on microsoft.com).

https://kxbjsyuhceggsyvxdkof.supabase.co/storage/v1/object/public/file-images/Microsoft_AI-901/page_17_img_2.jpg

The completed sentence is:

After deploying a vision-enabled GPT model in Microsoft Foundry, you can configure an application

to send requests to the endpoint of the model.

To call a deployed model from an application, the application must send requests to the deployed

model’s endpoint and include the required authentication credentials. The Foundry playground is

used for testing interactively, but application code calls the deployed model endpoint.

The other options are incorrect:

evaluation pipeline of model is used to evaluate model performance, not to receive inference

requests from an app.

Foundry playground is an interactive testing environment, not the production target for application

requests.

training dataset of the model is used for training or fine-tuning, not for sending inference requests.

Therefore, the correct answer is endpoint of the model.

https://kxbjsyuhceggsyvxdkof.supabase.co/storage/v1/object/public/file-images/Microsoft_AI-901/page_28_img_2.jpg

Statement 1: Voice Live returns only transcribed text. = No

Voice Live is not limited to transcription. Microsoft documentation states that the Voice Live API

supports real-time bidirectional voice applications, including speech recognition, text-to-speech

synthesis, avatar streaming, animation data, and audio processing.

Statement 2: Voice Live requires you to separately implement speech to text and text to speech

services. = No

Voice Live provides a single real-time voice API experience rather than requiring separate STT and

TTS implementations for the conversational loop. Microsoft describes live AI voice conversations as

combining speech capabilities for real-time interaction, and the Voice Live API includes speech

recognition and text-to-speech synthesis features.

Statement 3: Voice Live combines speech to text, reasoning, and text to speech into a single

conversational experience. = Yes

This is correct. Microsoft’s guidance explains that Azure OpenAI audio/realtime capabilities are for

scenarios that combine audio with language understanding, reasoning, or generation in a single

model call, and Voice Live supports real-time voice-enabled applications over WebSocket

connections.

https://kxbjsyuhceggsyvxdkof.supabase.co/storage/v1/object/public/file-images/Microsoft_AI-901/page_8_img_1.jpg

The correct Python method is:

recognizer.recognize_once()

Completed code:

import azure.cognitiveservices.speech as speechsdk

...

speech_config = speechsdk.SpeechConfig(subscription=key, region=region)

recognizer = speechsdk.SpeechRecognizer(speech_config=speech_config)

recognizer.recognize_once()

Microsoft’s Azure Speech documentation for Foundry Tools explains that Speech to text is used for

real-time speech recognition and converting spoken audio into text. The Python Speech SDK uses a

SpeechRecognizer object for recognition from audio input, such as a microphone.

Why the other options are incorrect:

recognizer.speak_text_async("Ready") is incorrect because speaking text is text-to-speech, not

speech-to-text recognition.

recognizer.start_continuous_recognition() can be used for continuous recognition, but the code

shown is asking for the basic method to recognize spoken input and convert it to text from the

SpeechRecognizer.

recognizer.start_keyword_recognition() is used for keyword/wake-word recognition, not general

speech-to-text transcription of spoken commands.

Therefore, the correct answer is:

recognizer.recognize_once()

https://kxbjsyuhceggsyvxdkof.supabase.co/storage/v1/object/public/file-images/Microsoft_AI-901/page_9_img_2.jpg

Azure Content Understanding analyzers define the type of content to process, such as documents,

images, audio, or video, and what to extract, such as text, layout, fields, transcripts, and structured

JSON output. Microsoft lists base analyzer types such as prebuilt-document, prebuilt-audio, prebuilt-

video, and prebuilt-image.

For scanned invoices in PDF format, the correct choice is document analyzer, because invoices are

document/form content. Microsoft’s Content Understanding quickstart specifically uses the prebuilt-

invoice analyzer to extract structured data from an invoice PDF.

For voicemail recordings in WAV format, the correct choice is audio analyzer, because voicemail is

audio content. Content Understanding supports structured data extraction from audio files, and the

analyzer reference identifies prebuilt-audio as the base analyzer for audio content.

Fairness

Privacy and security

Transparency

Reliability and safety

Comprehensive and Detailed Explanation with all Azure AI documents: =

Evaluating model outputs to ensure that decisions are NOT biased against specific demographic

groups: Fairness

Microsoft’s responsible AI principle of fairness means AI systems should treat people fairly and avoid

unfair bias across demographic groups.

Encrypting sensitive customer data and restricting system access to authorized personnel: Privacy

and security

The privacy and security principle requires AI systems to protect personal and sensitive data and

prevent unauthorized access.

Informing users when they are interacting with an AI system and explaining the system’s capabilities

and limitations: Transparency

The transparency principle requires that people understand when they are interacting with AI and

understand the system’s purpose, capabilities, and limitations.

Testing AI systems under different conditions to reduce unexpected failures: Reliability and safety

The reliability and safety principle means AI systems should operate reliably and safely, including

through testing and validation under expected and unexpected conditions.

client = OpenAI(

api_key="...",

base_url="https://resource1.openai.azure.com/openai/v1/",

)

response = client.responses.create(

model="my-mini-gpt",

...

)

For Azure OpenAI in Microsoft Foundry, the base_url uses the Azure OpenAI resource name in the

endpoint format:

https://.openai.azure.com/openai/v1/

In the question, the Azure OpenAI resource is named Resource1, so the first blank must be resource1.

Microsoft documentation for Azure OpenAI v1 endpoints confirms that the endpoint must use the

...openai.azure.com/openai/v1/ path.

For the model parameter, Azure OpenAI requires the deployment name, not the underlying model

name. Microsoft states that Azure OpenAI always requires the deployment name when calling APIs,

even when the parameter is named model.

The deployed model is gpt-4.1-mini, but the deployment name is my-mini-gpt. Therefore, the second

blank must be:

model="my-mini-gpt"

So the correct selections are:

base_url blank = resource1

model blank = my-mini-gpt