Tag: Microsoft Certification

AI, AI-901, Generative AI, Microsoft Certification

Create and test a single-agent solution in the Foundry Portal (AI-901 Exam Prep)

 
This post is a part of the AI-901: Microsoft Azure AI Fundamentals Exam Prep Hub. 
This topic falls under these sections:
Implement AI solutions by using Microsoft Foundry (55–60%)
   --> Implement generative AI apps and agents by using Foundry
      --> Create and test a single-agent solution in the Foundry Portal

Note that there are 10 practice questions (with answers and explanations) for each section to help you solidify your knowledge of the material. Also, there are 2 practice tests with 60 questions each available on the hub below the exam topics section.

AI agents are an increasingly important part of modern AI applications. Microsoft Azure AI Foundry provides tools that allow developers to create, configure, test, and manage AI agents directly within the Foundry portal.

For the AI-901 certification exam, candidates should understand the basic concepts behind creating and testing a single-agent AI solution using Azure AI Foundry.

This topic falls under the “Implement generative AI apps and agents by using Foundry” section of the AI-901 exam objectives.

What Is an AI Agent?

An AI agent is an AI-powered system designed to perform tasks, answer questions, and interact with users autonomously or semi-autonomously.

Agents often use:

  • Large Language Models (LLMs)
  • Prompt engineering
  • External tools
  • Memory
  • Data sources

to complete tasks.

What Is a Single-Agent Solution?

A single-agent solution uses one AI agent to manage interactions and tasks.

The agent receives input, processes requests, and generates responses.

Examples of Single-Agent Solutions

Common examples include:

  • Customer support assistants
  • FAQ bots
  • IT help desk assistants
  • Educational tutors
  • Internal knowledge assistants

AI Agent vs. Traditional Chatbot

Traditional ChatbotAI Agent
Often rule-basedAI-driven reasoning
Limited flexibilityMore adaptive
Predefined responsesDynamic responses
Basic workflowsCan perform complex tasks

Azure AI Foundry

Azure AI Foundry provides tools for creating and managing AI agents and generative AI applications.

The portal allows developers to:

  • Configure agents
  • Test prompts
  • Connect models
  • Evaluate responses
  • Monitor behavior

Basic Components of a Single-Agent Solution

A single-agent solution often includes:

  • AI model
  • System instructions
  • User interaction interface
  • Memory/context handling
  • Optional tools or data connections

AI Models in Agents

Agents typically use generative AI models such as large language models.

The model processes prompts and generates responses.

System Instructions

System instructions define how the agent should behave.

These instructions influence:

  • Tone
  • Personality
  • Safety
  • Response style
  • Allowed behavior

Example System Instruction

“You are a professional customer support assistant. Provide concise and helpful answers.”

User Prompts

Users interact with the agent by entering prompts or questions.

Example User Prompt

“How do I reset my password?”

Context and Memory

Many agents maintain conversational context.

This allows the agent to remember previous interactions during a session.

Example

User

“Tell me about Azure AI.”

User Later

“Can it support chatbots?”

The agent remembers the conversation topic.

Creating a Single-Agent Solution in Foundry

The general workflow includes:

  1. Open Azure AI Foundry
  2. Create or select a project
  3. Choose an AI model
  4. Configure the agent
  5. Define system instructions
  6. Test the agent
  7. Refine prompts and settings

Selecting a Model

Developers choose a model based on:

  • Performance
  • Cost
  • Speed
  • Language support
  • Context window size

Configuring the Agent

Agent configuration may include:

  • Name
  • Instructions
  • Model selection
  • Safety settings
  • Tool connections

Testing the Agent

The Foundry portal allows interactive testing.

Users can:

  • Enter prompts
  • Review responses
  • Adjust settings
  • Refine instructions

Playground Testing

Foundry includes playground environments for experimentation.

Developers can test:

  • Prompt quality
  • Tone
  • Accuracy
  • Context handling

before deploying applications.

Example Testing Scenario

System Instruction

“You are a helpful study assistant.”

User Prompt

“Explain supervised learning.”

The agent generates a response according to its instructions.

Prompt Engineering for Agents

Effective prompts improve agent behavior.

Helpful techniques include:

  • Clear instructions
  • Specific tasks
  • Output formatting
  • Context inclusion

Model Parameters

Developers may configure model settings such as:

  • Temperature
  • Maximum tokens
  • Top-p sampling

Temperature

Temperature controls response creativity.

Low TemperatureHigh Temperature
More predictableMore creative
More focusedMore varied

Maximum Tokens

Maximum tokens limit response length.

Lower values create shorter responses.

Tool Integration

Some agents can connect to external tools or data sources.

Examples include:

  • Databases
  • Search systems
  • APIs
  • Knowledge bases

Example Tool Usage

An IT support agent may retrieve information from a company knowledge base.

Grounding

Grounding connects AI responses to trusted data sources.

Grounded responses are generally more accurate and reliable.

Hallucinations

AI agents may occasionally produce incorrect or fabricated information.

These errors are called hallucinations.

Testing and grounding help reduce hallucinations.

Responsible AI Considerations

Single-agent solutions should follow Responsible AI principles.

Important considerations include:

  • Fairness
  • Privacy
  • Security
  • Transparency
  • Safety
  • Accountability

Content Filtering

Content filtering helps reduce:

  • Harmful outputs
  • Offensive content
  • Unsafe instructions

Authentication and Access Control

Organizations should secure access to AI agents using:

  • API keys
  • Identity management
  • Role-based access controls

Monitoring and Evaluation

Organizations should monitor agents for:

  • Accuracy
  • Performance
  • Bias
  • Safety
  • Usage patterns

Common Real-World Use Cases

Scenario 1: Customer Support Agent

Goal

Answer customer questions automatically.

Capabilities

  • Conversational responses
  • Knowledge retrieval
  • Escalation guidance

Scenario 2: Educational Tutor

Goal

Help students learn technical concepts.

Capabilities

  • Step-by-step explanations
  • Personalized tutoring
  • Interactive Q&A

Scenario 3: Internal Company Assistant

Goal

Help employees find company information.

Capabilities

  • Policy lookup
  • Document summarization
  • Search assistance

Advantages of Single-Agent Solutions

Benefits include:

  • Simpler architecture
  • Easier management
  • Faster deployment
  • Lower complexity
  • Natural interactions

Limitations of Single-Agent Solutions

Challenges may include:

  • Limited specialization
  • Hallucinations
  • Context limitations
  • Dependency on prompt quality

More complex systems may require multiple agents.

Single-Agent vs. Multi-Agent Systems

Single-AgentMulti-Agent
One agent handles tasksMultiple specialized agents
Simpler designMore complex
Easier managementBetter specialization
Lower overheadGreater coordination

Important AI-901 Exam Tips

For the exam, remember these key points:

  • AI agents use generative AI models to interact with users.
  • A single-agent solution uses one agent for interactions and tasks.
  • Azure AI Foundry provides tools for creating and testing agents.
  • System instructions guide agent behavior.
  • User prompts define tasks and questions.
  • Playground environments allow interactive testing.
  • Temperature controls creativity.
  • Grounding improves reliability.
  • Hallucinations are incorrect AI-generated outputs.
  • Responsible AI principles apply to AI agents.

Quick Knowledge Check

Question 1

What is a single-agent solution?

Answer

An AI system that uses one agent to process interactions and tasks.

Question 2

What is the purpose of system instructions?

Answer

To guide agent behavior, tone, and safety.

Question 3

What does grounding help improve?

Answer

Accuracy and reliability of AI responses.

Question 4

What are hallucinations?

Answer

Incorrect or fabricated AI-generated information.

Practice Exam Questions

Question 1

What is a single-agent solution?

A. A system that uses multiple AI agents simultaneously
B. A system that uses one AI agent to handle interactions and tasks
C. A database clustering solution
D. A networking security appliance

Correct Answer

B. A system that uses one AI agent to handle interactions and tasks

Explanation

A single-agent solution uses one AI-powered agent to process user requests and generate responses.

Why the Other Answers Are Incorrect

A. A system that uses multiple AI agents simultaneously

This describes a multi-agent system.

C. A database clustering solution

This is unrelated to AI agents.

D. A networking security appliance

This is unrelated to AI systems.

Question 2

Which Microsoft platform provides tools for creating and testing AI agents?

A. Microsoft Word
B. Azure AI Foundry
C. Microsoft Paint
D. Azure Virtual Desktop

Correct Answer

B. Azure AI Foundry

Explanation

Azure AI Foundry provides tools for building, testing, configuring, and managing AI agents and generative AI applications.

Why the Other Answers Are Incorrect

A. Microsoft Word

Word is a document editor.

C. Microsoft Paint

Paint is a graphics application.

D. Azure Virtual Desktop

This provides virtual desktop infrastructure services.

Question 3

What is the PRIMARY purpose of system instructions in an AI agent?

A. To physically store AI models
B. To define the agent’s behavior, tone, and rules
C. To improve monitor resolution
D. To compress training data

Correct Answer

B. To define the agent’s behavior, tone, and rules

Explanation

System instructions guide how the AI agent behaves and responds to users.

Why the Other Answers Are Incorrect

A. To physically store AI models

System instructions do not store models.

C. To improve monitor resolution

This is unrelated to AI agents.

D. To compress training data

This is unrelated to prompting.

Question 4

Which statement BEST describes grounding in AI systems?

A. Permanently deleting unused prompts
B. Connecting AI responses to trusted data sources
C. Increasing image brightness automatically
D. Compressing API requests

Correct Answer

B. Connecting AI responses to trusted data sources

Explanation

Grounding improves reliability by helping AI generate responses based on trusted information.

Why the Other Answers Are Incorrect

A. Permanently deleting unused prompts

This is unrelated to grounding.

C. Increasing image brightness automatically

This is unrelated to generative AI.

D. Compressing API requests

Grounding is unrelated to network compression.

Question 5

What is the PRIMARY purpose of playground testing in Azure AI Foundry?

A. Managing payroll systems
B. Experimenting with prompts and evaluating AI responses
C. Compressing video files
D. Managing physical servers

Correct Answer

B. Experimenting with prompts and evaluating AI responses

Explanation

Playgrounds allow developers to interactively test prompts, instructions, and AI behavior.

Why the Other Answers Are Incorrect

A. Managing payroll systems

This is unrelated to AI Foundry.

C. Compressing video files

Playgrounds are not media tools.

D. Managing physical servers

Playgrounds focus on AI interaction and testing.

Question 6

Which parameter controls how creative or random an AI agent’s responses will be?

A. Temperature
B. OCR threshold
C. Pixel density
D. Frame rate

Correct Answer

A. Temperature

Explanation

Temperature controls randomness and creativity in generated responses.

Why the Other Answers Are Incorrect

B. OCR threshold

This relates to text extraction from images.

C. Pixel density

This relates to image quality.

D. Frame rate

This relates to video playback.

Question 7

What are hallucinations in generative AI systems?

A. Hardware failures in cloud servers
B. Incorrect or fabricated AI-generated information
C. Authentication timeouts
D. Network bandwidth limitations

Correct Answer

B. Incorrect or fabricated AI-generated information

Explanation

Hallucinations occur when AI systems generate false or invented information.

Why the Other Answers Are Incorrect

A. Hardware failures in cloud servers

This is unrelated to hallucinations.

C. Authentication timeouts

This is a security or networking issue.

D. Network bandwidth limitations

This is unrelated to AI-generated accuracy.

Question 8

Why is conversation context important in AI agents?

A. It increases monitor resolution
B. It helps the agent remember previous interactions during a session
C. It permanently stores training datasets
D. It reduces internet costs

Correct Answer

B. It helps the agent remember previous interactions during a session

Explanation

Conversation context allows the AI agent to generate more coherent and relevant responses across multiple prompts.

Why the Other Answers Are Incorrect

A. It increases monitor resolution

Context does not affect displays.

C. It permanently stores training datasets

Context is session-related, not training storage.

D. It reduces internet costs

Context does not directly affect networking costs.

Question 9

Which Responsible AI feature helps reduce harmful or offensive AI-generated outputs?

A. Content filtering
B. Image compression
C. Database replication
D. Spreadsheet formatting

Correct Answer

A. Content filtering

Explanation

Content filtering helps block unsafe or inappropriate AI-generated responses.

Why the Other Answers Are Incorrect

B. Image compression

This reduces file size.

C. Database replication

This copies database data.

D. Spreadsheet formatting

This is unrelated to AI safety.

Question 10

What is one advantage of a single-agent solution compared to a multi-agent system?

A. Greater architectural complexity
B. Easier management and simpler design
C. Requires no prompts
D. Eliminates all hallucinations

Correct Answer

B. Easier management and simpler design

Explanation

Single-agent solutions are generally simpler to configure, deploy, and manage.

Why the Other Answers Are Incorrect

A. Greater architectural complexity

Multi-agent systems are usually more complex.

C. Requires no prompts

AI agents still rely on prompts and instructions.

D. Eliminates all hallucinations

Hallucinations can still occur in single-agent systems.

Final Thoughts

Creating and testing single-agent solutions in Azure AI Foundry is an important topic for the AI-901 certification exam. Microsoft expects candidates to understand the core concepts behind AI agents, prompt configuration, testing workflows, grounding, and Responsible AI practices.

Azure AI Foundry provides an accessible environment for building and experimenting with conversational AI agents that can support a wide variety of real-world business scenarios.

Go to the AI-901 Exam Prep Hub main page

AI, AI-901, Generative AI, Microsoft Certification, Natural Language Processing (NLP)

Create a lightweight chat client application by using the Foundry SDK (AI-901 Exam Prep)

 
This post is a part of the AI-901: Microsoft Azure AI Fundamentals Exam Prep Hub. 
This topic falls under these sections:
Implement AI solutions by using Microsoft Foundry (55–60%)
   --> Implement generative AI apps and agents by using Foundry
      --> Create a lightweight chat client application by using the Foundry SDK

Note that there are 10 practice questions (with answers and explanations) for each section to help you solidify your knowledge of the material. Also, there are 2 practice tests with 60 questions each available on the hub below the exam topics section.

Modern generative AI applications often include chat-based interfaces that allow users to interact naturally with AI models. Microsoft Azure AI Foundry provides SDKs (Software Development Kits) that developers can use to build lightweight chat applications that connect to deployed AI models.

For the AI-901 certification exam, candidates should understand the basic concepts behind creating chat client applications using the Foundry SDK and how these applications interact with deployed generative AI models.

This topic falls under the “Implement generative AI apps and agents by using Foundry” section of the AI-901 exam objectives.

What Is a Chat Client Application?

A chat client application is a software application that allows users to communicate with an AI model using conversational prompts and responses.

Users type messages, and the AI model generates replies.

Common Chat Application Examples

Examples include:

  • AI assistants
  • Customer support bots
  • Internal company copilots
  • Study assistants
  • Virtual agents
  • Help desk chatbots

What Is an SDK?

SDK stands for Software Development Kit.

An SDK provides tools and libraries that help developers build applications more easily.

SDKs typically include:

  • APIs
  • Authentication tools
  • Code libraries
  • Documentation
  • Example code

What Is the Foundry SDK?

The Foundry SDK allows developers to connect applications to deployed AI models within Azure AI Foundry.

Developers can use SDKs to:

  • Send prompts
  • Receive AI-generated responses
  • Manage conversations
  • Configure requests
  • Handle authentication

Why Use an SDK?

Using an SDK simplifies development.

Without an SDK, developers would need to manually handle:

  • Network requests
  • Authentication
  • Error handling
  • API formatting

SDKs abstract much of this complexity.

Lightweight Chat Applications

A lightweight chat client is a simple application focused on core chat functionality.

It usually includes:

  • User input field
  • Conversation display
  • AI response generation
  • Basic session management

Basic Chat Workflow

A typical AI chat application workflow includes:

  1. User enters a prompt
  2. Application sends request to deployed model
  3. AI model processes prompt
  4. Model generates response
  5. Application displays response

Connecting to a Deployed Model

Chat applications connect to deployed AI models using:

  • API endpoints
  • Authentication credentials
  • SDK libraries

The deployed model processes incoming prompts.

Authentication

Applications typically authenticate using:

  • API keys
  • Azure credentials
  • Managed identities

Authentication ensures only authorized users and applications can access AI services.

Example Chat Interaction

User

“Explain machine learning in simple terms.”

AI Model

“Machine learning is a type of AI where computers learn patterns from data instead of being explicitly programmed.”

Conversation History

Many chat applications maintain conversation history.

This allows the AI model to remember context during the session.

Example of Context Retention

User

“Who founded Microsoft?”

AI

“Microsoft was founded by Bill Gates and Paul Allen.”

User

“When was it founded?”

Because conversation history is maintained, the AI understands the second question refers to Microsoft.

System Prompts in Chat Applications

Chat applications often include system prompts that guide model behavior.

Example System Prompt

“You are a helpful technical tutor. Explain topics clearly for beginners.”

This influences:

  • Tone
  • Style
  • Behavior
  • Safety

User Prompts

User prompts represent the questions or requests entered during the conversation.

Example User Prompt

“Explain neural networks.”

Model Responses

The deployed AI model generates responses based on:

  • System prompt
  • User prompt
  • Conversation history
  • Model parameters

Model Parameters

Chat applications may configure parameters such as:

  • Temperature
  • Maximum tokens
  • Top-p sampling

Temperature

Temperature controls response creativity.

Low TemperatureHigh Temperature
More focusedMore creative
More predictableMore varied

Maximum Tokens

Maximum tokens limit response length.

Smaller values create shorter responses.

Streaming Responses

Some chat applications support streaming responses.

Streaming displays generated text gradually as the model produces it.

This improves user experience by reducing perceived waiting time.

Error Handling

Applications should handle errors gracefully.

Common issues include:

  • Network failures
  • Invalid credentials
  • Rate limits
  • Timeout errors

Rate Limits

AI services may limit request frequency.

Applications should be designed to handle:

  • Request throttling
  • Retry logic
  • Usage quotas

Responsible AI Considerations

Chat applications should follow Responsible AI principles.

Important considerations include:

  • Content filtering
  • Privacy
  • Safety
  • Bias reduction
  • Transparency

Content Filtering

Content filters help reduce:

  • Harmful responses
  • Offensive content
  • Unsafe outputs

Privacy and Security

Applications should protect:

  • User conversations
  • Authentication credentials
  • Sensitive information

Logging and Monitoring

Organizations may monitor chat applications for:

  • Performance
  • Usage
  • Errors
  • Safety concerns

Azure AI Foundry

Azure AI Foundry provides tools for deploying models and managing generative AI applications.

Developers can:

  • Deploy models
  • Test prompts
  • Monitor applications
  • Manage AI resources

Azure OpenAI Service

Azure OpenAI Service provides access to generative AI models used in chat applications.

High-Level SDK Workflow

A simplified workflow for a lightweight chat application typically includes:

  1. Install SDK
  2. Configure credentials
  3. Connect to deployed model
  4. Send prompts
  5. Receive responses
  6. Display conversation

Example High-Level Pseudocode

connect_to_model()
while True:
    user_prompt = get_user_input()
    response = send_prompt(user_prompt)
    display_response(response)

For AI-901, understanding the overall workflow is more important than memorizing syntax.

Common Real-World Scenarios

Scenario 1: Customer Support Chatbot

Goal

Answer customer questions automatically.

Features

  • Conversational interface
  • Context retention
  • Safe responses

Scenario 2: Internal Knowledge Assistant

Goal

Help employees search company information.

Features

  • Question answering
  • Document summarization
  • Secure access

Scenario 3: Educational Tutor

Goal

Provide interactive learning assistance.

Features

  • Step-by-step explanations
  • Conversational learning
  • Prompt customization

Advantages of Chat-Based AI Applications

Benefits include:

  • Natural user interaction
  • Faster information access
  • Automation of repetitive tasks
  • Improved customer experience
  • Scalability

Challenges and Limitations

Organizations should consider:

  • Hallucinations
  • Incorrect responses
  • Cost management
  • Privacy concerns
  • Latency
  • Prompt injection risks

Hallucinations

Generative AI models may occasionally generate incorrect or fabricated information.

These incorrect outputs are called hallucinations.

Applications should not assume all AI-generated responses are accurate.

Prompt Injection Risks

Malicious users may attempt to manipulate prompts to bypass safety controls.

Applications should implement safeguards against unsafe behavior.

Important AI-901 Exam Tips

For the exam, remember these key points:

  • SDKs simplify application development.
  • Chat clients communicate with deployed AI model endpoints.
  • System prompts define AI behavior.
  • User prompts represent user requests.
  • Conversation history helps maintain context.
  • Temperature controls response randomness.
  • Maximum tokens limit response length.
  • Streaming responses improve user experience.
  • Responsible AI principles apply to chat applications.
  • Authentication secures access to AI services.

Quick Knowledge Check

Question 1

What is the purpose of an SDK?

Answer

To simplify application development using tools and libraries.

Question 2

Why is conversation history important in chat applications?

Answer

It helps maintain context across multiple user interactions.

Question 3

What does temperature control in a generative AI model?

Answer

The creativity and randomness of responses.

Question 4

Why are content filters important?

Answer

They help reduce harmful or unsafe AI-generated outputs.

Practice Exam Questions

Question 1

What is the PRIMARY purpose of a chat client application in generative AI?

A. To physically store servers
B. To allow users to interact conversationally with an AI model
C. To compress database files
D. To manage network hardware

Correct Answer

B. To allow users to interact conversationally with an AI model

Explanation

A chat client application enables users to send prompts and receive AI-generated conversational responses.

Why the Other Answers Are Incorrect

A. To physically store servers

Chat clients are software applications, not physical infrastructure.

C. To compress database files

This is unrelated to chat applications.

D. To manage network hardware

This is unrelated to generative AI chat systems.

Question 2

What does SDK stand for?

A. Secure Data Kernel
B. Software Development Kit
C. System Deployment Key
D. Structured Data Kit

Correct Answer

B. Software Development Kit

Explanation

An SDK provides tools, libraries, and documentation that help developers build applications more efficiently.

Why the Other Answers Are Incorrect

A. Secure Data Kernel

This is not the correct definition.

C. System Deployment Key

This is incorrect terminology.

D. Structured Data Kit

This is not the meaning of SDK.

Question 3

Why do developers commonly use SDKs when building AI applications?

A. SDKs eliminate the need for internet access
B. SDKs simplify communication with AI services and APIs
C. SDKs permanently store all prompts automatically
D. SDKs replace AI models entirely

Correct Answer

B. SDKs simplify communication with AI services and APIs

Explanation

SDKs help developers handle authentication, requests, responses, and integration more easily.

Why the Other Answers Are Incorrect

A. SDKs eliminate the need for internet access

Cloud AI services still require connectivity.

C. SDKs permanently store all prompts automatically

SDKs do not inherently provide permanent storage.

D. SDKs replace AI models entirely

SDKs connect applications to models; they do not replace them.

Question 4

What allows a chat application to remember previous user interactions during a conversation?

A. OCR
B. Conversation history
C. Image classification
D. Regression analysis

Correct Answer

B. Conversation history

Explanation

Conversation history preserves context so the AI can respond appropriately across multiple prompts.

Why the Other Answers Are Incorrect

A. OCR

OCR extracts text from images.

C. Image classification

This categorizes images.

D. Regression analysis

Regression predicts numeric values.

Question 5

Which prompt type defines the AI assistant’s behavior and communication style?

A. User prompt
B. System prompt
C. SQL prompt
D. OCR prompt

Correct Answer

B. System prompt

Explanation

System prompts establish behavior rules, tone, style, and safety guidelines.

Why the Other Answers Are Incorrect

A. User prompt

User prompts contain requests or questions.

C. SQL prompt

SQL is related to databases.

D. OCR prompt

OCR is unrelated to conversational behavior.

Question 6

What is the PRIMARY purpose of authentication in a chat client application?

A. To improve image resolution
B. To ensure only authorized users or applications access AI services
C. To increase response creativity
D. To summarize conversations

Correct Answer

B. To ensure only authorized users or applications access AI services

Explanation

Authentication protects AI resources and controls access to deployed services.

Why the Other Answers Are Incorrect

A. To improve image resolution

Authentication does not affect graphics.

C. To increase response creativity

Temperature settings influence creativity.

D. To summarize conversations

Authentication does not summarize data.

Question 7

Which configuration parameter controls how creative or random a generative AI response will be?

A. Temperature
B. OCR threshold
C. Frame rate
D. Compression ratio

Correct Answer

A. Temperature

Explanation

Temperature controls response randomness and creativity.

Why the Other Answers Are Incorrect

B. OCR threshold

This relates to text extraction.

C. Frame rate

This relates to video processing.

D. Compression ratio

This relates to file compression.

Question 8

What is the benefit of streaming AI responses in a chat application?

A. It improves monitor resolution
B. It allows responses to appear gradually as they are generated
C. It permanently stores all conversations
D. It disables content filtering

Correct Answer

B. It allows responses to appear gradually as they are generated

Explanation

Streaming improves user experience by showing generated text incrementally instead of waiting for the entire response.

Why the Other Answers Are Incorrect

A. It improves monitor resolution

Streaming does not affect displays.

C. It permanently stores all conversations

Streaming does not automatically store data.

D. It disables content filtering

Streaming does not remove safety controls.

Question 9

Which Responsible AI feature helps reduce harmful or offensive AI-generated responses?

A. Content filtering
B. Data compression
C. Video rendering
D. File indexing

Correct Answer

A. Content filtering

Explanation

Content filters help prevent unsafe or inappropriate AI outputs.

Why the Other Answers Are Incorrect

B. Data compression

Compression reduces file size.

C. Video rendering

Rendering creates visual output.

D. File indexing

Indexing organizes data for search.

Question 10

What are hallucinations in generative AI systems?

A. Hardware overheating events
B. Incorrect or fabricated AI-generated information
C. Authentication failures
D. Video processing delays

Correct Answer

B. Incorrect or fabricated AI-generated information

Explanation

Hallucinations occur when AI models generate inaccurate or invented information.

Why the Other Answers Are Incorrect

A. Hardware overheating events

This is unrelated to AI hallucinations.

C. Authentication failures

This is a security issue.

D. Video processing delays

This relates to media performance, not AI accuracy.

Final Thoughts

Creating lightweight chat applications with the Foundry SDK is an important concept for the AI-901 certification exam. Microsoft expects candidates to understand the basic architecture and workflow of AI-powered chat applications, including prompts, endpoints, authentication, conversation management, and Responsible AI considerations.

Azure AI Foundry and Azure OpenAI Service provide powerful tools that allow developers to build conversational AI experiences quickly and efficiently.

Go to the AI-901 Exam Prep Hub main page

AI, AI-901, Artificial Intelligence (AI), Microsoft Certification

Deploy a model and interact with it in the Foundry Portal (AI-901 Exam Prep)

 
This post is a part of the AI-901: Microsoft Azure AI Fundamentals Exam Prep Hub. 
This topic falls under these sections:
Implement AI solutions by using Microsoft Foundry (55–60%)
   --> Implement generative AI apps and agents by using Foundry
      --> Deploy a model and interact with it in the Foundry Portal

Note that there are 10 practice questions (with answers and explanations) for each section to help you solidify your knowledge of the material. Also, there are 2 practice tests with 60 questions each available on the hub below the exam topics section.

Microsoft Azure AI Foundry provides a centralized environment for building, testing, deploying, and managing generative AI models and AI-powered applications. For the AI-901 certification exam, candidates should understand the basic process of deploying AI models and interacting with them through the Foundry portal.

This topic focuses on how developers and AI practitioners use Azure AI Foundry to deploy generative AI models, test prompts, configure model settings, and interact with deployed AI endpoints.

This topic falls under the “Implement generative AI apps and agents by using Foundry” section of the AI-901 exam objectives.

What Is Azure AI Foundry?

Azure AI Foundry is Microsoft’s platform for building and managing AI applications and agents.

Azure AI Foundry provides tools to:

  • Explore AI models
  • Deploy models
  • Test prompts
  • Configure AI behavior
  • Evaluate responses
  • Monitor AI applications
  • Manage AI resources

It supports generative AI development using Azure-hosted AI services and models.

What Does “Deploying a Model” Mean?

Deploying a model means making the AI model available for use.

A deployed model can:

  • Receive prompts
  • Process requests
  • Generate responses
  • Be accessed through applications or APIs

Deployment creates an endpoint that applications can use to interact with the model.

What Is a Model Endpoint?

An endpoint is a network-accessible interface that allows applications or users to communicate with a deployed AI model.

Applications send requests to the endpoint and receive AI-generated responses.

Common Deployment Scenarios

Organizations deploy models for many purposes, including:

  • Chatbots
  • AI assistants
  • Document summarization
  • Content generation
  • Customer support systems
  • Code generation
  • Data extraction

Azure AI Foundry Workflow

A simplified workflow in Azure AI Foundry typically includes:

  1. Create or access an Azure AI resource
  2. Open Azure AI Foundry portal
  3. Select a model
  4. Configure deployment settings
  5. Deploy the model
  6. Test prompts
  7. Interact with the model
  8. Integrate the endpoint into applications

Accessing the Foundry Portal

Users access Azure AI Foundry through a web-based portal.

The portal provides graphical tools for:

  • Model selection
  • Prompt testing
  • Deployment management
  • Performance monitoring

Exploring Available Models

Azure AI Foundry allows users to browse available models.

Examples may include:

  • Large Language Models (LLMs)
  • Image-generation models
  • Embedding models
  • Speech models

Models may vary by:

  • Size
  • Performance
  • Cost
  • Supported capabilities

Selecting a Model

Users choose models based on application requirements.

Factors may include:

  • Accuracy
  • Speed
  • Cost
  • Context window size
  • Multimodal support
  • Language support

Example Scenario

A company building a customer support chatbot may choose a conversational large language model.

Deploying a Model in Foundry

The deployment process usually involves:

  • Selecting a model
  • Naming the deployment
  • Choosing deployment settings
  • Allocating resources
  • Creating the endpoint

Deployment Names

Deployments are typically assigned unique names.

Example

support-chat-model

Applications use deployment names when sending requests.

Model Configuration Options

During deployment, users may configure:

  • Model version
  • Scaling options
  • Authentication settings
  • Content filters
  • Region
  • Resource allocation

Content Filtering and Safety

Azure AI Foundry includes Responsible AI safety features.

Content filtering helps reduce:

  • Harmful outputs
  • Offensive content
  • Unsafe responses

This is important for enterprise AI applications.

Interacting with a Deployed Model

After deployment, users can interact with the model directly within the Foundry portal.

This often includes:

  • Entering prompts
  • Viewing responses
  • Adjusting settings
  • Testing behavior

Playground Interfaces

Azure AI Foundry provides playground environments for experimentation.

Playgrounds allow users to:

  • Test prompts
  • Compare outputs
  • Tune settings
  • Evaluate model behavior

Prompt Testing

Users can experiment with:

  • System prompts
  • User prompts
  • Formatting instructions
  • Role prompting

Prompt testing helps improve AI response quality.

Example Prompt Interaction

User Prompt

“Summarize this customer feedback in three bullet points.”

Model Response

The model generates a summarized response.

Model Parameters

Foundry portals may allow adjustment of model parameters such as:

  • Temperature
  • Maximum tokens
  • Top-p sampling

Temperature

Temperature controls response randomness.

Low TemperatureHigh Temperature
More predictableMore creative
More focusedMore varied

Maximum Tokens

Maximum tokens limit response length.

Smaller limits create shorter responses.

System Prompts in Foundry

Users can configure system prompts to guide AI behavior.

Example System Prompt

“You are a professional technical support assistant. Keep responses concise and helpful.”

System prompts influence:

  • Tone
  • Style
  • Safety
  • Formatting

Evaluating Responses

Users should evaluate AI outputs for:

  • Accuracy
  • Relevance
  • Safety
  • Bias
  • Hallucinations

AI-generated content should be reviewed carefully.

Hallucinations

Generative AI models can produce incorrect or fabricated information.

These incorrect outputs are called hallucinations.

Prompt engineering and grounding techniques help reduce hallucinations.

API Access

Once deployed, applications can connect to the model endpoint using APIs.

This allows developers to integrate AI into applications.

Common Integration Scenarios

Applications may use deployed models for:

  • Chat interfaces
  • Search assistants
  • Document analysis
  • AI copilots
  • Workflow automation

Monitoring and Management

Azure AI Foundry supports monitoring deployed models.

Monitoring may include:

  • Usage tracking
  • Performance analysis
  • Error monitoring
  • Cost management

Scaling AI Deployments

Organizations may scale deployments to support:

  • More users
  • Higher request volumes
  • Faster response times

Cloud-based deployments support elastic scaling.

Responsible AI Considerations

When deploying AI models, organizations should consider:

  • Privacy
  • Security
  • Fairness
  • Transparency
  • Safety
  • Compliance

Generative AI applications should include safeguards against misuse.

Authentication and Security

Deployed models typically require secure authentication.

Security features may include:

  • API keys
  • Identity management
  • Access control

Common Challenges

Organizations may encounter challenges such as:

  • High usage costs
  • Latency
  • Hallucinations
  • Unsafe outputs
  • Poor prompt quality

Proper testing and monitoring are important.

Azure OpenAI Service

Azure OpenAI Service provides access to powerful generative AI models that can be deployed and managed through Azure AI Foundry.

Real-World Scenarios

Scenario 1: Customer Support Chatbot

Goal

Deploy a conversational AI assistant.

Activities

  • Deploy language model
  • Configure system prompts
  • Test responses in the playground

Scenario 2: Internal Knowledge Assistant

Goal

Allow employees to ask questions about company documentation.

Activities

  • Deploy AI model
  • Configure prompts
  • Integrate with enterprise systems

Scenario 3: Marketing Content Generator

Goal

Generate product descriptions automatically.

Activities

  • Deploy generative AI model
  • Test prompt variations
  • Evaluate response quality

Important AI-901 Exam Tips

For the exam, remember these key points:

  • Deploying a model makes it available for use.
  • Deployments create accessible endpoints.
  • Azure AI Foundry provides tools for testing and managing models.
  • Playgrounds allow prompt experimentation.
  • System prompts guide model behavior.
  • Temperature controls creativity and randomness.
  • Maximum tokens control response length.
  • AI outputs should be evaluated for accuracy and safety.
  • Content filtering supports Responsible AI practices.
  • APIs allow applications to connect to deployed models.

Quick Knowledge Check

Question 1

What does deploying a model do?

Answer

It makes the AI model available for use through an endpoint.

Question 2

What is the purpose of a playground in Azure AI Foundry?

Answer

To test prompts and interact with deployed models.

Question 3

What does the temperature setting control?

Answer

The randomness and creativity of model responses.

Question 4

Why are content filters important?

Answer

They help reduce harmful or unsafe AI-generated outputs.

Practice Exam Questions

Question 1

What is the PRIMARY purpose of deploying an AI model?

A. To permanently delete the model
B. To make the model available for use through an endpoint
C. To compress training data
D. To convert images into text

Correct Answer

B. To make the model available for use through an endpoint

Explanation

Deploying a model makes it accessible so applications and users can interact with it.

Why the Other Answers Are Incorrect

A. To permanently delete the model

Deployment does not delete models.

C. To compress training data

Deployment is unrelated to data compression.

D. To convert images into text

This describes OCR.

Question 2

What is an endpoint in the context of AI model deployment?

A. A physical server room
B. A network-accessible interface for interacting with a deployed model
C. A type of database backup
D. A computer vision algorithm

Correct Answer

B. A network-accessible interface for interacting with a deployed model

Explanation

Endpoints allow applications and users to send requests to deployed AI models and receive responses.

Why the Other Answers Are Incorrect

A. A physical server room

Endpoints are logical interfaces, not physical locations.

C. A type of database backup

This is unrelated to AI deployment.

D. A computer vision algorithm

Endpoints are not algorithms.

Question 3

Which Azure tool provides playgrounds for testing prompts and interacting with deployed AI models?

A. Azure SQL Database
B. Azure AI Foundry
C. Microsoft Excel
D. Azure Virtual Desktop

Correct Answer

B. Azure AI Foundry

Explanation

Azure AI Foundry provides tools for model deployment, prompt testing, evaluation, and management.

Why the Other Answers Are Incorrect

A. Azure SQL Database

This is a database service.

C. Microsoft Excel

Excel is not an AI deployment platform.

D. Azure Virtual Desktop

This provides desktop virtualization services.

Question 4

What is the PRIMARY purpose of a playground in Azure AI Foundry?

A. Hosting multiplayer games
B. Experimenting with prompts and testing model behavior
C. Managing employee payroll
D. Compressing image files

Correct Answer

B. Experimenting with prompts and testing model behavior

Explanation

Playgrounds allow users to interact with models, test prompts, and evaluate responses.

Why the Other Answers Are Incorrect

A. Hosting multiplayer games

This is unrelated to AI Foundry.

C. Managing employee payroll

This is unrelated to AI development.

D. Compressing image files

Playgrounds are not image utilities.

Question 5

Which configuration setting controls how creative or random AI-generated responses are?

A. OCR level
B. Temperature
C. Resolution scaling
D. Data indexing

Correct Answer

B. Temperature

Explanation

Temperature controls randomness and creativity in generative AI responses.

Why the Other Answers Are Incorrect

A. OCR level

OCR extracts text from images.

C. Resolution scaling

This relates to images, not text generation randomness.

D. Data indexing

Indexing is unrelated to generative response creativity.

Question 6

What is the effect of setting a lower temperature value in a generative AI model?

A. More random responses
B. More predictable and focused responses
C. Faster internet speeds
D. Larger image generation sizes

Correct Answer

B. More predictable and focused responses

Explanation

Lower temperature settings reduce randomness and produce more deterministic outputs.

Why the Other Answers Are Incorrect

A. More random responses

Higher temperatures increase randomness.

C. Faster internet speeds

Temperature does not affect networking.

D. Larger image generation sizes

Temperature is unrelated to image dimensions.

Question 7

Which prompt type defines the AI assistant’s behavior, tone, and rules?

A. User prompt
B. System prompt
C. SQL query
D. OCR prompt

Correct Answer

B. System prompt

Explanation

System prompts provide high-level behavioral instructions to the AI model.

Why the Other Answers Are Incorrect

A. User prompt

User prompts specify tasks or requests.

C. SQL query

SQL queries interact with databases.

D. OCR prompt

OCR is unrelated to conversational AI behavior.

Question 8

Why are content filters important when deploying generative AI models?

A. They improve internet bandwidth
B. They help reduce harmful or unsafe outputs
C. They increase monitor resolution
D. They replace system prompts entirely

Correct Answer

B. They help reduce harmful or unsafe outputs

Explanation

Content filtering supports Responsible AI by helping prevent harmful or inappropriate AI-generated content.

Why the Other Answers Are Incorrect

A. They improve internet bandwidth

Content filters do not affect networking performance.

C. They increase monitor resolution

This is unrelated to AI safety.

D. They replace system prompts entirely

Content filters complement prompts; they do not replace them.

Question 9

What are hallucinations in generative AI?

A. Physical hardware failures
B. Incorrect or fabricated AI-generated information
C. Database replication errors
D. Unauthorized user logins

Correct Answer

B. Incorrect or fabricated AI-generated information

Explanation

Hallucinations occur when AI generates inaccurate or invented information.

Why the Other Answers Are Incorrect

A. Physical hardware failures

This is unrelated to AI hallucinations.

C. Database replication errors

This is a database issue.

D. Unauthorized user logins

This is a security issue.

Question 10

After deploying a model, how do external applications typically interact with it?

A. Through handwritten forms
B. Through APIs connected to the deployment endpoint
C. Through spreadsheet imports only
D. Through local USB connections

Correct Answer

B. Through APIs connected to the deployment endpoint

Explanation

Applications commonly communicate with deployed AI models using APIs and endpoints.

Why the Other Answers Are Incorrect

A. Through handwritten forms

This is unrelated to AI deployment.

C. Through spreadsheet imports only

Spreadsheets are not the primary integration mechanism.

D. Through local USB connections

Cloud AI services typically use network-based APIs, not USB connections.

Final Thoughts

Deploying and interacting with AI models in Azure AI Foundry is an important skill area for the AI-901 certification exam. Microsoft expects candidates to understand the basic deployment workflow, prompt testing process, model configuration options, and Responsible AI considerations involved in building generative AI applications.

Azure AI Foundry simplifies AI development by providing a centralized environment for deploying, testing, and managing AI models and agents.

Go to the AI-901 Exam Prep Hub main page

AI, AI-901, Artificial Intelligence (AI), Generative AI, Microsoft Certification

Create effective system and user prompts for Generative AI models (AI-901 Exam Prep)

 
This post is a part of the AI-901: Microsoft Azure AI Fundamentals Exam Prep Hub. 
This topic falls under these sections:
Implement AI solutions by using Microsoft Foundry (55–60%)
   --> Implement generative AI apps and agents by using Foundry
      --> Create effective system and user prompts for Generative AI models

Note that there are 10 practice questions (with answers and explanations) for each section to help you solidify your knowledge of the material. Also, there are 2 practice tests with 60 questions each available on the hub below the exam topics section.

Prompting is one of the most important skills when working with generative AI systems. Microsoft expects AI-901 candidates to understand how to create effective prompts that guide generative AI models toward useful, accurate, and safe outputs.

This topic focuses on how system prompts and user prompts influence the behavior of generative AI models and how prompt engineering techniques improve AI-generated responses.

This topic falls under the “Implement generative AI apps and agents by using Foundry” section of the AI-901 exam objectives.

What Is a Prompt?

A prompt is an instruction or input provided to a generative AI model.

Prompts guide the model’s response and influence:

  • Content
  • Tone
  • Format
  • Style
  • Accuracy
  • Level of detail

The quality of the prompt strongly affects the quality of the output.

What Is Prompt Engineering?

Prompt engineering is the process of designing and refining prompts to improve AI-generated responses.

Effective prompt engineering helps:

  • Produce more accurate answers
  • Reduce ambiguity
  • Improve consistency
  • Control response format
  • Reduce hallucinations
  • Improve safety and Responsible AI behavior

Types of Prompts

For the AI-901 exam, two important prompt types are:

  • System prompts
  • User prompts

What Is a System Prompt?

A system prompt provides high-level instructions that define how the AI model should behave.

System prompts often control:

  • Personality
  • Tone
  • Rules
  • Safety boundaries
  • Formatting requirements
  • Behavior expectations

The system prompt typically has higher priority than user prompts.

Example of a System Prompt

“You are a professional technical support assistant. Provide concise and accurate troubleshooting guidance. Do not provide harmful or unsafe instructions.”

This system prompt defines:

  • The assistant’s role
  • Communication style
  • Safety expectations

What Is a User Prompt?

A user prompt is the direct request or question submitted by the user.

User prompts specify the task the model should perform.

Example of a User Prompt

“How do I reset my router?”

The AI model combines:

  • System instructions
  • User request
  • Context information

to generate a response.

Relationship Between System and User Prompts

System prompts establish behavior rules, while user prompts define the immediate task.

Example

System Prompt

“You are a helpful travel assistant. Always provide answers in bullet points.”

User Prompt

“Suggest three family-friendly attractions in Orlando.”

The model responds according to both prompts.

Characteristics of Effective Prompts

Good prompts are usually:

  • Clear
  • Specific
  • Contextual
  • Structured
  • Goal-oriented

Clear Prompts

Clear prompts reduce confusion and ambiguity.

Weak Prompt

“Tell me about databases.”

Better Prompt

“Explain the differences between relational and non-relational databases for beginners.”

The second prompt provides:

  • Specific topic
  • Audience
  • Scope

Specific Prompts

Specific prompts improve response accuracy.

Weak Prompt

“Write a report.”

Better Prompt

“Write a 300-word summary of cloud computing benefits for small businesses.”

Specific prompts define:

  • Length
  • Topic
  • Audience

Providing Context

Context helps the model generate more relevant answers.

Example

“I am studying for the AI-901 exam. Explain OCR in simple terms with one real-world example.”

The additional context improves response quality.

Requesting Output Format

Prompts can specify desired formatting.

Example

“Provide the answer as a table.”

or

“Summarize the information in bullet points.”

Role Prompting

Role prompting assigns the AI a specific role or perspective.

Example

“Act as a cybersecurity consultant.”

or

“You are an experienced data analyst.”

Role prompting helps guide tone and expertise.

Step-by-Step Prompting

Prompts can request step-by-step explanations.

Example

“Explain how machine learning works step-by-step for beginners.”

This improves clarity and educational usefulness.

Few-Shot Prompting

Few-shot prompting provides examples within the prompt.

This helps the model understand expected patterns.

Example

Positive review → Positive sentiment
Negative review → Negative sentiment
“The service was excellent.” →

The model learns the desired output structure.

Zero-Shot Prompting

Zero-shot prompting asks the model to perform a task without examples.

Example

“Classify this review as positive or negative.”

Chain-of-Thought Prompting

Chain-of-thought prompting encourages step-by-step reasoning.

Example

“Explain your reasoning step-by-step before providing the final answer.”

This can improve reasoning accuracy for complex tasks.

Prompting for Summarization

Generative AI models can summarize content using prompts.

Example

“Summarize this article in three bullet points.”

Prompting for Content Generation

Prompts can generate new content such as:

  • Emails
  • Reports
  • Stories
  • Marketing copy
  • Code

Example

“Write a professional email requesting a project update.”

Prompting for Transformation Tasks

AI models can transform content into different formats.

Examples

  • Translate text
  • Rewrite text
  • Simplify technical content
  • Convert paragraphs into tables

Example

“Rewrite this paragraph for a non-technical audience.”

Prompting for Code Generation

Generative AI can assist with programming tasks.

Example

“Write a Python function that calculates sales tax.”

Prompting for Data Extraction

Prompts can request structured data extraction.

Example

“Extract all dates and company names from this document.”

Prompt Injection Risks

Prompt injection occurs when users attempt to override system instructions.

Example

A malicious user prompt may attempt to bypass safety rules.

Organizations should implement safeguards against unsafe prompting behavior.

Responsible AI Considerations

Effective prompting should follow Responsible AI principles.

Important considerations include:

  • Safety
  • Fairness
  • Privacy
  • Transparency
  • Content moderation
  • Harm prevention

Hallucinations

Generative AI models can sometimes produce incorrect or fabricated information.

These errors are called hallucinations.

Good prompting can reduce hallucinations but may not eliminate them completely.

Example of a Hallucination

An AI model inventing a fake citation or incorrect fact.

Techniques to Reduce Hallucinations

Helpful strategies include:

  • Providing clear context
  • Using specific instructions
  • Asking for sources
  • Limiting scope
  • Using grounded data

Temperature and Creativity

Some generative AI systems allow configuration settings such as temperature.

Temperature affects randomness and creativity.

Low TemperatureHigh Temperature
More predictableMore creative
More focusedMore varied
Better for factual tasksBetter for brainstorming

Azure AI Foundry

Azure AI Foundry helps developers build, test, and manage generative AI applications and agents.

Developers can:

  • Experiment with prompts
  • Evaluate AI responses
  • Configure AI models
  • Implement safety controls

Azure OpenAI Service

Azure OpenAI Service provides access to powerful generative AI models that support prompt-based interactions.

Real-World Prompting Scenarios

Scenario 1: Customer Support Assistant

System Prompt

“You are a professional support assistant. Be polite and concise.”

User Prompt

“How do I reset my password?”

Scenario 2: Study Assistant

System Prompt

“Explain technical topics for beginners.”

User Prompt

“Explain neural networks in simple terms.”

Scenario 3: Marketing Content Generator

System Prompt

“Generate professional marketing copy.”

User Prompt

“Create a product description for a smartwatch.”

Best Practices for Effective Prompting

  • Be specific
  • Provide context
  • Define output format
  • Use examples when helpful
  • Keep instructions clear
  • Test and refine prompts
  • Avoid ambiguity
  • Include Responsible AI safeguards

Common Prompting Mistakes

Common mistakes include:

  • Vague instructions
  • Missing context
  • Conflicting requirements
  • Overly broad requests
  • Unclear formatting expectations

Important AI-901 Exam Tips

For the exam, remember these key points:

  • System prompts define AI behavior and rules.
  • User prompts specify the task to perform.
  • Effective prompts are clear and specific.
  • Prompt engineering improves AI outputs.
  • Few-shot prompting includes examples.
  • Zero-shot prompting provides no examples.
  • Chain-of-thought prompting encourages reasoning.
  • Hallucinations are incorrect AI-generated outputs.
  • Temperature settings affect creativity and randomness.
  • Responsible AI principles apply to prompting.

Quick Knowledge Check

Question 1

What is the difference between a system prompt and a user prompt?

Answer

A system prompt defines AI behavior and rules, while a user prompt requests a specific task.

Question 2

What is prompt engineering?

Answer

The process of designing prompts to improve AI-generated responses.

Question 3

What is few-shot prompting?

Answer

Providing examples within prompts to guide the model.

Question 4

What are hallucinations in generative AI?

Answer

Incorrect or fabricated AI-generated information.

Practice Exam Questions

Question 1

What is the PRIMARY purpose of a system prompt in a generative AI application?

A. To store images generated by the model
B. To define the AI model’s behavior, rules, and tone
C. To increase internet speed
D. To encrypt database records

Correct Answer

B. To define the AI model’s behavior, rules, and tone

Explanation

System prompts provide high-level instructions that guide how the AI assistant behaves and responds.

Why the Other Answers Are Incorrect

A. To store images generated by the model

System prompts do not store data.

C. To increase internet speed

This is unrelated to AI prompting.

D. To encrypt database records

Encryption is unrelated to prompting.

Question 2

Which statement BEST describes a user prompt?

A. A hidden configuration file for servers
B. A direct instruction or request submitted by the user
C. A database backup mechanism
D. A type of neural network architecture

Correct Answer

B. A direct instruction or request submitted by the user

Explanation

User prompts contain the specific task or question the user wants the AI model to perform.

Why the Other Answers Are Incorrect

A. A hidden configuration file for servers

This is unrelated to generative AI prompting.

C. A database backup mechanism

This is unrelated to prompting.

D. A type of neural network architecture

Prompts are instructions, not architectures.

Question 3

Which prompt is MOST effective?

A. “Tell me stuff.”
B. “Write something about technology.”
C. “Explain cloud computing for beginners in 5 bullet points.”
D. “Do work.”

Correct Answer

C. “Explain cloud computing for beginners in 5 bullet points.”

Explanation

Effective prompts are clear, specific, and include formatting or audience requirements.

Why the Other Answers Are Incorrect

A. “Tell me stuff.”

This is too vague.

B. “Write something about technology.”

This lacks detail and direction.

D. “Do work.”

This is ambiguous and unclear.

Question 4

What is prompt engineering?

A. Designing hardware for AI servers
B. Building neural network chips
C. Creating and refining prompts to improve AI responses
D. Encrypting AI training data

Correct Answer

C. Creating and refining prompts to improve AI responses

Explanation

Prompt engineering focuses on improving generative AI outputs through better prompt design.

Why the Other Answers Are Incorrect

A. Designing hardware for AI servers

This is hardware engineering.

B. Building neural network chips

This is semiconductor engineering.

D. Encrypting AI training data

This is a security task.

Question 5

Which prompting technique includes examples within the prompt to guide the AI model?

A. Few-shot prompting
B. Object detection
C. OCR prompting
D. Clustering

Correct Answer

A. Few-shot prompting

Explanation

Few-shot prompting provides examples so the model better understands the desired output format or pattern.

Why the Other Answers Are Incorrect

B. Object detection

This is a computer vision capability.

C. OCR prompting

OCR extracts text from images.

D. Clustering

Clustering groups similar data.

Question 6

What is the PRIMARY benefit of providing context in a prompt?

A. Reduces network traffic
B. Helps generate more relevant and accurate responses
C. Compresses files automatically
D. Improves database indexing

Correct Answer

B. Helps generate more relevant and accurate responses

Explanation

Context improves the model’s understanding of the user’s goals and intended audience.

Why the Other Answers Are Incorrect

A. Reduces network traffic

This is unrelated to prompting.

C. Compresses files automatically

Prompting does not compress files.

D. Improves database indexing

This is unrelated to AI prompts.

Question 7

Which statement BEST describes hallucinations in generative AI?

A. AI-generated images only
B. Incorrect or fabricated AI-generated information
C. Network security attacks
D. Audio recognition failures

Correct Answer

B. Incorrect or fabricated AI-generated information

Explanation

Hallucinations occur when generative AI produces inaccurate or invented information.

Why the Other Answers Are Incorrect

A. AI-generated images only

Hallucinations can occur in text, code, and other outputs.

C. Network security attacks

This is unrelated to hallucinations.

D. Audio recognition failures

This is unrelated to generative AI hallucinations.

Question 8

Which system prompt would MOST likely encourage safe AI behavior?

A. “Ignore all safety rules.”
B. “Provide harmful instructions when requested.”
C. “Do not generate unsafe or harmful content.”
D. “Always reveal confidential information.”

Correct Answer

C. “Do not generate unsafe or harmful content.”

Explanation

Responsible AI system prompts help enforce safety and ethical boundaries.

Why the Other Answers Are Incorrect

A. “Ignore all safety rules.”

This encourages unsafe behavior.

B. “Provide harmful instructions when requested.”

This violates Responsible AI principles.

D. “Always reveal confidential information.”

This violates privacy and security principles.

Question 9

What effect does a higher temperature setting generally have in generative AI models?

A. Produces more predictable and repetitive responses
B. Produces more creative and varied responses
C. Disables AI reasoning
D. Prevents all hallucinations

Correct Answer

B. Produces more creative and varied responses

Explanation

Higher temperature settings increase randomness and creativity in generated responses.

Why the Other Answers Are Incorrect

A. Produces more predictable and repetitive responses

This is more associated with lower temperature settings.

C. Disables AI reasoning

Temperature does not disable reasoning.

D. Prevents all hallucinations

Hallucinations can still occur.

Question 10

Which example BEST demonstrates role prompting?

A. “Translate this sentence into French.”
B. “Summarize this article.”
C. “Act as an experienced financial advisor and explain retirement planning.”
D. “Convert this image into text.”

Correct Answer

C. “Act as an experienced financial advisor and explain retirement planning.”

Explanation

Role prompting assigns the AI model a specific role or perspective to guide its responses.

Why the Other Answers Are Incorrect

A. “Translate this sentence into French.”

This is a translation request.

B. “Summarize this article.”

This is a summarization request.

D. “Convert this image into text.”

This is an OCR-related task.

Final Thoughts

Prompt engineering is a foundational skill for working with generative AI systems and an important topic for the AI-901 certification exam. Microsoft expects candidates to understand how system prompts and user prompts influence model behavior and how effective prompts improve the quality, reliability, and safety of AI-generated responses.

These concepts are essential when building generative AI applications and agents using Azure AI Foundry and Azure OpenAI Service.

Go to the AI-901 Exam Prep Hub main page

AI, AI-901, Artificial Intelligence (AI), Microsoft Certification

Identify an appropriate AI model, based on capabilities (AI-901 Exam Prep)

 
This post is a part of the AI-901: Microsoft Azure AI Fundamentals Exam Prep Hub. 
This topic falls under these sections:
Identify AI concepts and capabilities (40–45%)
   --> Identify AI model components and configurations
      --> Identify an appropriate AI model, based on capabilities

Note that there are 10 practice questions (with answers and explanations) for each section to help you solidify your knowledge of the material. Also, there are 2 practice tests with 60 questions each available on the hub below the exam topics section.

Selecting the correct AI model for a specific business problem is an important skill and a key topic for the AI-901 certification exam. Microsoft expects candidates to understand the capabilities of common AI model types and recognize which model is appropriate for different scenarios.

This topic falls under the “Identify AI model components and configurations” section of the exam objectives.

Why Choosing the Right AI Model Matters

Different AI models are designed for different types of tasks.

Choosing the wrong model may lead to:

  • Poor accuracy
  • Inefficient processing
  • Increased costs
  • Unusable results
  • Poor user experiences

Understanding model capabilities helps organizations build effective AI solutions.

Major Categories of AI Models

For AI-901, you should understand the capabilities of several major AI model categories:

  • Classification models
  • Regression models
  • Clustering models
  • Computer vision models
  • Natural language processing (NLP) models
  • Generative AI models
  • Recommendation systems
  • Anomaly detection models

Classification Models

Classification models predict categories or labels.

They answer questions such as:

  • “What type is this?”
  • “Which category does this belong to?”

Common Use Cases

  • Spam email detection
  • Fraud detection
  • Sentiment analysis
  • Medical diagnosis classification
  • Image categorization

Example

A model predicts whether an email is:

  • Spam
  • Not spam

This is a classification problem.

Binary Classification

Binary classification predicts one of two possible outcomes.

Examples

  • Fraud or not fraud
  • Approved or denied
  • Positive or negative sentiment

Multiclass Classification

Multiclass classification predicts one of several categories.

Example

An AI model identifies whether an image contains:

  • A dog
  • A cat
  • A bird
  • A horse

Regression Models

Regression models predict numeric values.

They answer questions such as:

  • “How much?”
  • “How many?”
  • “What value?”

Common Use Cases

  • House price prediction
  • Sales forecasting
  • Temperature prediction
  • Demand estimation

Example

Predicting the selling price of a house based on:

  • Size
  • Location
  • Number of bedrooms

This is a regression problem.

Clustering Models

Clustering models group similar items together without predefined labels.

Clustering is a type of unsupervised learning.

Common Use Cases

  • Customer segmentation
  • Market analysis
  • Pattern discovery
  • Grouping similar documents

Example

A retailer groups customers based on purchasing behavior.

The model discovers patterns automatically.

Computer Vision Models

Computer vision models analyze images and video.

Common Capabilities

  • Object detection
  • Facial recognition
  • Image classification
  • Optical Character Recognition (OCR)
  • Image tagging

Example Use Cases

  • Self-driving cars
  • Security systems
  • Medical imaging
  • Product identification

Image Classification

Image classification identifies what appears in an image.

Example

Determining whether an image contains:

  • A cat
  • A dog
  • A car

Object Detection

Object detection identifies and locates objects within an image.

Example

A traffic monitoring system detects:

  • Cars
  • Pedestrians
  • Traffic lights

and determines their positions.

Optical Character Recognition (OCR)

OCR extracts text from images or scanned documents.

Example

Reading text from:

  • Receipts
  • Invoices
  • Forms
  • License plates

Natural Language Processing (NLP) Models

NLP models work with human language.

Common Capabilities

  • Sentiment analysis
  • Translation
  • Text summarization
  • Chatbots
  • Speech recognition
  • Named entity recognition

Example Use Cases

  • Customer support chatbots
  • Language translation apps
  • Voice assistants

Sentiment Analysis

Sentiment analysis identifies emotional tone in text.

Example

Determining whether a product review is:

  • Positive
  • Negative
  • Neutral

Translation Models

Translation models convert text between languages.

Example

Converting English text into Spanish.

Speech Recognition

Speech recognition converts spoken language into text.

Example

Voice assistants converting speech commands into written text.

Generative AI Models

Generative AI models create new content.

Common Outputs

  • Text
  • Images
  • Audio
  • Video
  • Code

Example Use Cases

  • AI chatbots
  • Content generation
  • Image creation
  • Coding assistants

Large Language Models (LLMs)

LLMs are generative AI models focused on language tasks.

Capabilities

  • Conversations
  • Summarization
  • Question answering
  • Content generation
  • Code generation

Example

An AI assistant answering user questions in natural language.

Recommendation Systems

Recommendation systems suggest items users may prefer.

Common Use Cases

  • Product recommendations
  • Movie recommendations
  • Music recommendations
  • Online advertising

Example

An online retailer recommends products based on browsing history.

Anomaly Detection Models

Anomaly detection models identify unusual patterns or behaviors.

Common Use Cases

  • Fraud detection
  • Cybersecurity monitoring
  • Equipment failure prediction
  • Network intrusion detection

Example

A bank identifies suspicious credit card transactions.

Supervised vs. Unsupervised Learning

Understanding learning types helps identify appropriate models.

Learning TypeDescription
Supervised LearningUses labeled data
Unsupervised LearningFinds patterns without labels

Supervised Examples

  • Classification
  • Regression

Unsupervised Examples

  • Clustering
  • Some anomaly detection systems

Choosing the Right AI Model

To select an appropriate AI model, ask:

What Type of Output Is Needed?

GoalModel Type
Predict categoriesClassification
Predict numbersRegression
Group similar itemsClustering
Generate contentGenerative AI
Analyze imagesComputer Vision
Process languageNLP

Is the Data Labeled?

Data TypeAppropriate Learning Type
Labeled dataSupervised learning
Unlabeled dataUnsupervised learning

What Content Is Being Processed?

Content TypeAppropriate Model
TextNLP or LLM
ImagesComputer Vision
AudioSpeech models
Numerical dataRegression or classification

Real-World Examples

Scenario 1: Email Spam Detection

Goal

Identify whether emails are spam.

Best Model

Classification model

Scenario 2: Predicting House Prices

Goal

Estimate home values.

Best Model

Regression model

Scenario 3: Grouping Customers by Buying Behavior

Goal

Identify customer segments.

Best Model

Clustering model

Scenario 4: AI Chatbot

Goal

Generate conversational responses.

Best Model

Large Language Model (LLM)

Scenario 5: Reading Text from Scanned Documents

Goal

Extract printed text.

Best Model

OCR computer vision model

Scenario 6: Detecting Fraudulent Transactions

Goal

Identify suspicious activity.

Best Model

Anomaly detection model

Azure AI Services and Model Types

Microsoft Azure AI Services provide many prebuilt AI capabilities, including:

  • Vision services
  • Speech services
  • Language services
  • Generative AI tools
  • Document intelligence
  • Recommendation capabilities

Microsoft Azure helps organizations apply the correct AI models to different business scenarios.

Responsible AI Considerations

When selecting AI models, organizations should also consider:

  • Fairness
  • Transparency
  • Privacy
  • Reliability
  • Inclusiveness
  • Accountability

A technically accurate model may still create ethical or operational concerns if deployed improperly.

Important AI-901 Exam Tips

For the exam, remember these key points:

  • Classification predicts categories.
  • Regression predicts numeric values.
  • Clustering groups similar items.
  • NLP models process language.
  • Computer vision models process images and video.
  • Generative AI creates new content.
  • Recommendation systems suggest relevant items.
  • Anomaly detection identifies unusual behavior.
  • LLMs are generative AI models for language tasks.
  • OCR extracts text from images or documents.

Quick Knowledge Check

Question 1

Which model type is best for predicting numeric values?

Answer

Regression models.

Question 2

Which AI capability is used to extract text from scanned documents?

Answer

Optical Character Recognition (OCR).

Question 3

What type of model is typically used for chatbots that generate responses?

Answer

Large Language Models (LLMs).

Question 4

Which learning type uses unlabeled data?

Answer

Unsupervised learning.

Practice Exam Questions

Question 1

A company wants to predict future monthly sales revenue based on historical sales data.

Which type of AI model is MOST appropriate?

A. Classification
B. Regression
C. Clustering
D. Computer vision

Correct Answer

B. Regression

Explanation

Regression models are used to predict numeric values such as revenue, prices, or temperatures.

Why the Other Answers Are Incorrect

A. Classification

Classification predicts categories, not numeric values.

C. Clustering

Clustering groups similar items.

D. Computer vision

Computer vision processes images and video.

Question 2

An organization wants to identify whether emails are spam or not spam.

Which type of AI model should be used?

A. Regression
B. Clustering
C. Classification
D. OCR

Correct Answer

C. Classification

Explanation

Spam detection is a classification problem because the output belongs to predefined categories: spam or not spam.

Why the Other Answers Are Incorrect

A. Regression

Regression predicts numeric values.

B. Clustering

Clustering groups unlabeled data.

D. OCR

OCR extracts text from images.

Question 3

Which AI capability is MOST appropriate for extracting text from scanned documents?

A. Object detection
B. OCR
C. Regression
D. Recommendation system

Correct Answer

B. OCR

Explanation

Optical Character Recognition (OCR) extracts printed or handwritten text from images or scanned documents.

Why the Other Answers Are Incorrect

A. Object detection

Object detection identifies objects within images.

C. Regression

Regression predicts numeric values.

D. Recommendation system

Recommendation systems suggest items to users.

Question 4

A retailer wants to group customers based on purchasing behavior without predefined labels.

Which type of AI model is MOST appropriate?

A. Classification
B. Regression
C. Clustering
D. Translation

Correct Answer

C. Clustering

Explanation

Clustering models group similar data points together without labeled categories.

Why the Other Answers Are Incorrect

A. Classification

Classification requires labeled categories.

B. Regression

Regression predicts numbers.

D. Translation

Translation converts text between languages.

Question 5

Which type of AI model is BEST suited for generating natural language responses in a chatbot?

A. Large Language Model (LLM)
B. Regression model
C. Clustering model
D. Decision tree only

Correct Answer

A. Large Language Model (LLM)

Explanation

LLMs are generative AI models designed for language tasks such as conversation, summarization, and question answering.

Why the Other Answers Are Incorrect

B. Regression model

Regression predicts numeric values.

C. Clustering model

Clustering groups similar data.

D. Decision tree only

Decision trees are not specialized for conversational text generation.

Question 6

A bank wants to identify suspicious credit card transactions that differ from normal spending patterns.

Which AI capability is MOST appropriate?

A. Sentiment analysis
B. Anomaly detection
C. OCR
D. Image classification

Correct Answer

B. Anomaly detection

Explanation

Anomaly detection models identify unusual or abnormal behavior that may indicate fraud or security issues.

Why the Other Answers Are Incorrect

A. Sentiment analysis

Sentiment analysis evaluates emotional tone in text.

C. OCR

OCR extracts text from images.

D. Image classification

Image classification categorizes images.

Question 7

What is the PRIMARY capability of a computer vision model?

A. Predicting stock prices
B. Processing and analyzing visual content such as images and video
C. Translating text between languages
D. Generating database queries

Correct Answer

B. Processing and analyzing visual content such as images and video

Explanation

Computer vision models work with images and video to identify objects, text, faces, and other visual information.

Why the Other Answers Are Incorrect

A. Predicting stock prices

This is typically a regression problem.

C. Translating text between languages

Translation is an NLP task.

D. Generating database queries

This is not the primary role of computer vision.

Question 8

A streaming service suggests movies based on a user’s viewing history.

Which AI capability is being used?

A. Recommendation system
B. OCR
C. Regression
D. Object detection

Correct Answer

A. Recommendation system

Explanation

Recommendation systems suggest products, movies, music, or other items based on user behavior and preferences.

Why the Other Answers Are Incorrect

B. OCR

OCR extracts text from images.

C. Regression

Regression predicts numeric values.

D. Object detection

Object detection identifies objects in images.

Question 9

Which type of AI model would MOST likely be used for language translation?

A. NLP model
B. Clustering model
C. Regression model
D. Computer vision model

Correct Answer

A. NLP model

Explanation

Natural Language Processing (NLP) models are designed to process and understand human language, including translation tasks.

Why the Other Answers Are Incorrect

B. Clustering model

Clustering groups similar items.

C. Regression model

Regression predicts numeric outputs.

D. Computer vision model

Computer vision analyzes images and video.

Question 10

Which statement BEST describes the difference between classification and regression models?

A. Classification predicts categories, while regression predicts numeric values
B. Classification uses images, while regression uses text only
C. Regression groups data, while classification predicts prices
D. Regression and classification are identical

Correct Answer

A. Classification predicts categories, while regression predicts numeric values

Explanation

Classification models predict labels or categories, while regression models predict continuous numeric values.

Why the Other Answers Are Incorrect

B. Classification uses images, while regression uses text only

Both models can work with many data types.

C. Regression groups data, while classification predicts prices

Grouping data is clustering, not regression.

D. Regression and classification are identical

They solve different types of problems.

Final Thoughts

Understanding AI model capabilities is a critical foundational skill for the AI-901 certification exam. Microsoft expects candidates to recognize which AI model types are appropriate for different business scenarios and understand the strengths of common AI approaches.

Knowing how to match business problems to the correct AI capabilities is essential for designing effective AI solutions on Azure and beyond.

Go to the AI-901 Exam Prep Hub main page

AI, AI Governance, AI-901, Artificial Intelligence (AI), Microsoft Certification

Identify scenarios for common AI workloads, Including Generative and Agentic AI, Text Analysis, Speech, Computer Vision, and Information Extraction (AI-901 Exam Prep)

 
This post is a part of the AI-901: Microsoft Azure AI Fundamentals Exam Prep Hub. 
This topic falls under these sections:
Identify AI concepts and capabilities (40–45%)
   --> Identify AI workloads
      --> Identify scenarios for common AI workloads, Including Generative and Agentic AI, Text Analysis, Speech, Computer Vision, and Information Extraction

Note that there are 10 practice questions (with answers and explanations) for each section to help you solidify your knowledge of the material. Also, there are 2 practice tests with 60 questions each available on the hub below the exam topics section.

Understanding common AI workloads is one of the foundational concepts in artificial intelligence and a major focus area of the AI-901 certification exam. Microsoft expects candidates to recognize different types of AI workloads and identify appropriate real-world scenarios for each.

This topic falls under the “Identify AI workloads” section of the exam objectives.

What Is an AI Workload?

An AI workload is a category of AI tasks designed to solve a particular type of problem.

Different workloads specialize in processing different types of data such as:

  • Text
  • Speech
  • Images
  • Documents
  • Audio
  • Video

Understanding AI workloads helps organizations choose the correct AI technologies for business solutions.

Major AI Workloads for AI-901

For the AI-901 exam, you should understand these common AI workloads:

  • Generative AI
  • Agentic AI
  • Text analysis
  • Speech AI
  • Computer vision
  • Information extraction

Generative AI

Generative AI creates new content based on patterns learned from training data.

Common Outputs

  • Text
  • Images
  • Audio
  • Video
  • Code

Common Scenarios

  • AI chatbots
  • Content creation
  • Email drafting
  • Code generation
  • Image generation
  • Text summarization

Example

A marketing team uses AI to generate product descriptions automatically.

Large Language Models (LLMs)

Many generative AI systems use Large Language Models (LLMs).

LLMs are trained on massive text datasets and can:

  • Answer questions
  • Summarize content
  • Generate text
  • Translate languages
  • Assist with coding

Example

An AI assistant generates meeting summaries from conversation transcripts.

Agentic AI

Agentic AI refers to AI systems that can autonomously plan, reason, and take actions to accomplish goals.

Agentic AI systems may:

  • Make decisions
  • Perform multi-step tasks
  • Use tools
  • Interact with applications
  • Adapt based on feedback

Unlike simple chatbots, agentic AI systems can perform actions and workflows.

Agentic AI Scenarios

Examples

  • AI travel planning assistants
  • Autonomous customer support agents
  • AI workflow automation systems
  • AI research assistants
  • Scheduling assistants

Example

An AI assistant receives a request to schedule a meeting, checks calendars, sends invitations, and updates schedules automatically.

Text Analysis

Text analysis is an AI workload focused on understanding and processing written language.

Text analysis is part of Natural Language Processing (NLP).

Common Capabilities

  • Sentiment analysis
  • Key phrase extraction
  • Language detection
  • Named entity recognition
  • Text classification

Sentiment Analysis

Sentiment analysis identifies emotional tone in text.

Example Scenarios

  • Product review analysis
  • Social media monitoring
  • Customer feedback analysis

Example

An organization analyzes customer reviews to determine whether feedback is positive or negative.

Key Phrase Extraction

Key phrase extraction identifies important terms or phrases in text.

Example Scenarios

  • Document summarization
  • Search indexing
  • Topic identification

Example

An AI system extracts important keywords from support tickets.

Language Detection

Language detection identifies the language used in text.

Example Scenarios

  • Multilingual applications
  • Translation routing
  • Global customer support

Example

A website detects whether incoming text is English, Spanish, or French.

Named Entity Recognition (NER)

NER identifies important entities in text such as:

  • People
  • Organizations
  • Locations
  • Dates

Example

An AI system extracts company names and locations from contracts.

Speech AI

Speech AI works with spoken language and audio.

Common Capabilities

  • Speech-to-text
  • Text-to-speech
  • Speech translation
  • Speaker recognition

Speech-to-Text

Speech-to-text converts spoken audio into written text.

Example Scenarios

  • Voice transcription
  • Meeting captions
  • Voice assistants

Example

A meeting platform generates live captions during conferences.

Text-to-Speech

Text-to-speech converts written text into spoken audio.

Example Scenarios

  • Accessibility tools
  • Virtual assistants
  • Audiobooks
  • Navigation systems

Example

A navigation app reads driving directions aloud.

Speech Translation

Speech translation converts spoken language into another language.

Example Scenarios

  • International meetings
  • Travel applications
  • Multilingual support systems

Example

A conference tool translates spoken English into Spanish in real time.

Computer Vision

Computer vision enables AI systems to analyze images and video.

Common Capabilities

  • Image classification
  • Object detection
  • Facial recognition
  • OCR
  • Image tagging

Image Classification

Image classification identifies the contents of an image.

Example Scenarios

  • Medical image analysis
  • Product categorization
  • Wildlife monitoring

Example

An AI system identifies whether an image contains a cat or a dog.

Object Detection

Object detection identifies and locates objects within an image.

Example Scenarios

  • Traffic monitoring
  • Security surveillance
  • Manufacturing inspection

Example

A self-driving car detects pedestrians and vehicles.

Optical Character Recognition (OCR)

OCR extracts text from images or scanned documents.

Example Scenarios

  • Invoice processing
  • Form digitization
  • Receipt scanning

Example

An AI system extracts totals and dates from receipts.

Facial Recognition

Facial recognition identifies or verifies people using facial features.

Example Scenarios

  • Building access systems
  • Smartphone authentication
  • Security systems

Example

A mobile phone unlocks using facial recognition.

Information Extraction

Information extraction identifies and retrieves structured information from unstructured content.

This workload often combines:

  • OCR
  • NLP
  • Document analysis

Information Extraction Scenarios

Examples

  • Invoice processing
  • Contract analysis
  • Insurance claims processing
  • Healthcare form processing

Example

An AI system extracts invoice numbers, dates, and totals from scanned invoices automatically.

Structured vs. Unstructured Data

AI workloads often process unstructured data.

Structured DataUnstructured Data
TablesDocuments
DatabasesImages
SpreadsheetsAudio
Defined formatsVideos

Many AI workloads specialize in converting unstructured data into structured information.

Choosing the Correct AI Workload

Understanding the business problem helps determine the correct AI workload.

ScenarioAppropriate Workload
Generate contentGenerative AI
Perform autonomous tasksAgentic AI
Analyze written reviewsText analysis
Convert speech to textSpeech AI
Analyze imagesComputer vision
Extract data from formsInformation extraction

Real-World Examples

Scenario 1: Customer Support Chatbot

Goal

Answer customer questions naturally.

Appropriate Workload

Generative AI

Scenario 2: AI Scheduling Assistant

Goal

Manage appointments automatically.

Appropriate Workload

Agentic AI

Scenario 3: Review Analysis System

Goal

Determine customer sentiment.

Appropriate Workload

Text analysis

Scenario 4: Live Meeting Captions

Goal

Convert speech into text in real time.

Appropriate Workload

Speech AI

Scenario 5: Self-Driving Vehicle

Goal

Detect objects and surroundings.

Appropriate Workload

Computer vision

Scenario 6: Invoice Data Extraction

Goal

Extract invoice information automatically.

Appropriate Workload

Information extraction

Azure AI Services for Common Workloads

Microsoft Azure AI Services provide prebuilt tools for many AI workloads, including:

  • Azure AI Language
  • Azure AI Speech
  • Azure AI Vision
  • Azure AI Document Intelligence
  • Azure OpenAI Service

These services help organizations build AI solutions without creating models from scratch.

Responsible AI Considerations

All AI workloads should follow Responsible AI principles, including:

  • Fairness
  • Privacy
  • Transparency
  • Reliability
  • Inclusiveness
  • Accountability

Organizations should ensure AI systems are used ethically and safely.

Important AI-901 Exam Tips

For the exam, remember these key points:

  • Generative AI creates new content.
  • Agentic AI can autonomously perform tasks and workflows.
  • Text analysis processes written language.
  • Speech AI works with spoken language and audio.
  • Computer vision processes images and video.
  • OCR extracts text from images.
  • Information extraction converts unstructured data into structured information.
  • Sentiment analysis determines emotional tone in text.
  • Named Entity Recognition identifies important entities in text.

Quick Knowledge Check

Question 1

Which AI workload is best for generating marketing content?

Answer

Generative AI.

Question 2

Which AI workload converts spoken language into written text?

Answer

Speech AI.

Question 3

What does OCR do?

Answer

Extracts text from images or scanned documents.

Question 4

Which workload is designed to autonomously complete tasks and workflows?

Answer

Agentic AI.

Practice Exam Questions

Question 1

A company wants an AI system that can automatically generate marketing emails and product descriptions.

Which AI workload is MOST appropriate?

A. Computer vision
B. Generative AI
C. OCR
D. Regression analysis

Correct Answer

B. Generative AI

Explanation

Generative AI creates new content such as text, images, audio, and code based on learned patterns.

Why the Other Answers Are Incorrect

A. Computer vision

Computer vision analyzes images and video.

C. OCR

OCR extracts text from images.

D. Regression analysis

Regression predicts numeric values.

Question 2

An organization wants an AI assistant that can schedule meetings, send invitations, and update calendars automatically.

Which AI workload BEST fits this scenario?

A. Speech AI
B. Agentic AI
C. Clustering
D. OCR

Correct Answer

B. Agentic AI

Explanation

Agentic AI systems can autonomously perform multi-step tasks, make decisions, and interact with tools or applications.

Why the Other Answers Are Incorrect

A. Speech AI

Speech AI processes spoken language.

C. Clustering

Clustering groups similar data.

D. OCR

OCR extracts text from images.

Question 3

Which AI workload is MOST appropriate for determining whether customer reviews are positive or negative?

A. Sentiment analysis
B. Object detection
C. Regression
D. Facial recognition

Correct Answer

A. Sentiment analysis

Explanation

Sentiment analysis is a text analysis capability that identifies emotional tone in written text.

Why the Other Answers Are Incorrect

B. Object detection

Object detection identifies objects in images.

C. Regression

Regression predicts numeric values.

D. Facial recognition

Facial recognition analyzes faces in images or video.

Question 4

A company needs to convert spoken customer service calls into written transcripts.

Which AI workload should be used?

A. Computer vision
B. Speech-to-text
C. OCR
D. Recommendation system

Correct Answer

B. Speech-to-text

Explanation

Speech-to-text converts spoken audio into written text.

Why the Other Answers Are Incorrect

A. Computer vision

Computer vision processes images and video.

C. OCR

OCR extracts text from images, not audio.

D. Recommendation system

Recommendation systems suggest items to users.

Question 5

Which AI workload is MOST appropriate for identifying objects such as cars and pedestrians in traffic camera footage?

A. Text analysis
B. Object detection
C. Speech translation
D. Key phrase extraction

Correct Answer

B. Object detection

Explanation

Object detection identifies and locates objects within images or video.

Why the Other Answers Are Incorrect

A. Text analysis

Text analysis processes written language.

C. Speech translation

Speech translation converts spoken language between languages.

D. Key phrase extraction

Key phrase extraction identifies important terms in text.

Question 6

What is the PRIMARY purpose of OCR?

A. Translating spoken language
B. Extracting text from images or scanned documents
C. Detecting emotions in speech
D. Generating new images

Correct Answer

B. Extracting text from images or scanned documents

Explanation

Optical Character Recognition (OCR) converts printed or handwritten text in images into machine-readable text.

Why the Other Answers Are Incorrect

A. Translating spoken language

This is speech translation.

C. Detecting emotions in speech

This is speech or sentiment analysis.

D. Generating new images

This is a generative AI capability.

Question 7

Which workload is MOST associated with analyzing and processing human language?

A. Natural Language Processing (NLP)
B. Computer vision
C. Regression
D. Clustering

Correct Answer

A. Natural Language Processing (NLP)

Explanation

NLP focuses on understanding, analyzing, and generating human language.

Why the Other Answers Are Incorrect

B. Computer vision

Computer vision works with images and video.

C. Regression

Regression predicts numeric values.

D. Clustering

Clustering groups similar items.

Question 8

A business wants to automatically extract invoice numbers, totals, and dates from scanned invoices.

Which AI workload is MOST appropriate?

A. Recommendation system
B. Information extraction
C. Speech recognition
D. Regression

Correct Answer

B. Information extraction

Explanation

Information extraction retrieves structured information from unstructured documents and often combines OCR and NLP technologies.

Why the Other Answers Are Incorrect

A. Recommendation system

Recommendation systems suggest items.

C. Speech recognition

Speech recognition processes audio.

D. Regression

Regression predicts numbers rather than extracting document data.

Question 9

Which scenario BEST represents a computer vision workload?

A. Translating English text into Spanish
B. Detecting defects on a manufacturing assembly line using cameras
C. Summarizing documents automatically
D. Predicting monthly sales revenue

Correct Answer

B. Detecting defects on a manufacturing assembly line using cameras

Explanation

Computer vision systems analyze visual content such as images and video to identify objects, defects, and patterns.

Why the Other Answers Are Incorrect

A. Translating English text into Spanish

This is an NLP task.

C. Summarizing documents automatically

This is a generative AI or NLP task.

D. Predicting monthly sales revenue

This is a regression task.

Question 10

Which statement BEST describes agentic AI?

A. AI systems that only classify images
B. AI systems that autonomously perform tasks and make decisions
C. AI systems that store relational databases
D. AI systems that only process audio recordings

Correct Answer

B. AI systems that autonomously perform tasks and make decisions

Explanation

Agentic AI systems can reason, plan, interact with tools, and complete multi-step workflows with limited human intervention.

Why the Other Answers Are Incorrect

A. AI systems that only classify images

This describes computer vision tasks.

C. AI systems that store relational databases

Databases are not AI workloads.

D. AI systems that only process audio recordings

Speech AI handles audio processing, not autonomous task execution.

Final Thoughts

Understanding common AI workloads is essential for the AI-901 certification exam and for designing effective AI solutions. Microsoft expects candidates to recognize how different AI technologies solve different business problems and when each workload is most appropriate.

These foundational concepts help build a strong understanding of modern AI systems and Azure AI services.

Go to the AI-901 Exam Prep Hub main page

AI, AI Governance, AI Strategy, AI-901, Microsoft Certification

Describe common Text Analysis techniques, including Keyword Extraction, Entity Detection, Sentiment Analysis, and Summarization (AI-901 Exam Prep)

 
This post is a part of the AI-901: Microsoft Azure AI Fundamentals Exam Prep Hub. 
This topic falls under these sections:
Identify AI concepts and capabilities (40–45%)
   --> Identify AI workloads
      --> Describe common Text Analysis techniques, including Keyword Extraction, Entity Detection, Sentiment Analysis, and Summarization

Note that there are 10 practice questions (with answers and explanations) for each section to help you solidify your knowledge of the material. Also, there are 2 practice tests with 60 questions each available on the hub below the exam topics section.

Text analysis is one of the most common and important AI workloads covered in the AI-901 certification exam. Microsoft expects candidates to understand how AI systems analyze and interpret written language using Natural Language Processing (NLP) techniques.

This topic falls under the “Identify AI workloads” section of the AI-901 exam objectives.

What Is Text Analysis?

Text analysis is an AI workload that uses Natural Language Processing (NLP) to analyze, interpret, and extract meaning from written text.

Text analysis helps organizations process large amounts of unstructured textual data automatically.

Common Sources of Text Data

Organizations analyze text from many sources, including:

  • Emails
  • Customer reviews
  • Social media posts
  • Chat messages
  • Support tickets
  • Surveys
  • Documents
  • Articles

What Is Natural Language Processing (NLP)?

Natural Language Processing (NLP) is a branch of AI focused on helping computers understand and work with human language.

NLP combines:

  • Machine learning
  • Linguistics
  • Statistical analysis
  • Deep learning

NLP enables systems to interpret meaning, emotion, intent, and context within text.

Common Text Analysis Techniques

For the AI-901 exam, important text analysis techniques include:

  • Keyword extraction
  • Entity detection
  • Sentiment analysis
  • Summarization

Additional related techniques include:

  • Language detection
  • Translation
  • Text classification

Keyword Extraction

Keyword extraction identifies the most important words or phrases within text.

The goal is to determine the primary topics or themes.

How Keyword Extraction Works

AI systems analyze text and identify terms that appear most significant based on:

  • Frequency
  • Relevance
  • Context
  • Relationships to other words

Keyword Extraction Examples

Input Text

“The customer was very satisfied with the fast delivery and excellent product quality.”

Extracted Keywords

  • customer
  • fast delivery
  • product quality

Common Use Cases for Keyword Extraction

Search Optimization

Improve document indexing and search engines.

Document Categorization

Identify major document topics automatically.

Customer Feedback Analysis

Detect common issues or themes.

Content Tagging

Automatically assign tags to articles or documents.

Entity Detection

Entity detection identifies important entities mentioned within text.

This technique is often called Named Entity Recognition (NER).

Common Entity Types

AI systems may identify:

  • People
  • Organizations
  • Locations
  • Dates
  • Phone numbers
  • Email addresses
  • Products
  • Currency amounts

Entity Detection Example

Input Text

“Microsoft announced a conference in Seattle on June 15.”

Detected Entities

  • Microsoft → Organization
  • Seattle → Location
  • June 15 → Date

Common Use Cases for Entity Detection

Document Processing

Extract important business information from contracts or forms.

Compliance Monitoring

Identify sensitive information.

Customer Relationship Management

Track companies, customers, or products mentioned in communications.

Search and Analytics

Improve document filtering and organization.

Sentiment Analysis

Sentiment analysis identifies emotional tone or opinion within text.

It determines whether text expresses:

  • Positive sentiment
  • Negative sentiment
  • Neutral sentiment

How Sentiment Analysis Works

AI models analyze words, phrases, and context to estimate emotional tone.

Example Positive Words

  • Excellent
  • Great
  • Amazing

Example Negative Words

  • Poor
  • Terrible
  • Frustrating

Context is important because words can have different meanings depending on usage.

Sentiment Analysis Example

Input Text

“The product quality was excellent, but shipping was slow.”

Possible Sentiment Results

  • Product quality → Positive
  • Shipping experience → Negative

Some systems provide:

  • Overall sentiment
  • Sentence-level sentiment
  • Confidence scores

Common Use Cases for Sentiment Analysis

Customer Feedback Monitoring

Analyze reviews and surveys.

Brand Monitoring

Track public opinion on social media.

Customer Service Improvement

Identify dissatisfied customers.

Market Research

Understand consumer opinions.

Summarization

Summarization creates shorter versions of longer text while preserving key information.

AI summarization helps users quickly understand large amounts of information.

Types of Summarization

Extractive Summarization

Extractive summarization selects important sentences directly from the original text.

Abstractive Summarization

Abstractive summarization generates new sentences that summarize the meaning of the text.

This approach is more similar to how humans summarize information.

Summarization Example

Original Text

“The company reported increased sales this quarter due to strong online demand and improved supply chain performance.”

Summary

“The company experienced increased sales driven by online demand.”

Common Use Cases for Summarization

Meeting Summaries

Condense meeting transcripts.

News Summaries

Provide quick article overviews.

Customer Support

Summarize long support conversations.

Research Assistance

Condense lengthy documents or reports.

Language Detection

Language detection identifies the language used in text.

Example

An AI system determines whether text is:

  • English
  • Spanish
  • French
  • German

Common Use Cases

  • Multilingual applications
  • Translation routing
  • International customer support

Text Classification

Text classification assigns categories or labels to text.

Examples

  • Spam detection
  • Topic categorization
  • Support ticket routing

Real-World Examples

Scenario 1: Customer Review Analysis

Goal

Understand customer opinions.

Techniques Used

  • Sentiment analysis
  • Keyword extraction

Scenario 2: Legal Contract Processing

Goal

Identify important contract information.

Techniques Used

  • Entity detection
  • Summarization

Scenario 3: News Aggregation Platform

Goal

Provide short summaries of articles.

Techniques Used

  • Summarization
  • Keyword extraction

Scenario 4: Customer Support Ticket System

Goal

Automatically categorize and prioritize tickets.

Techniques Used

  • Text classification
  • Sentiment analysis

Azure AI Language Services

Azure AI Language Services provide prebuilt NLP capabilities such as:

  • Sentiment analysis
  • Entity recognition
  • Summarization
  • Language detection
  • Key phrase extraction

These services help developers add text analysis features without building models from scratch.

Structured vs. Unstructured Text Data

Text analysis commonly processes unstructured data.

Structured DataUnstructured Data
DatabasesEmails
TablesDocuments
SpreadsheetsSocial media posts
Defined fieldsReviews

AI systems help convert unstructured text into usable structured information.

Responsible AI Considerations

Organizations using text analysis should consider:

  • Privacy
  • Bias
  • Transparency
  • Security
  • Accuracy
  • Responsible handling of personal data

Text analysis systems may process sensitive information and should be designed carefully.

Important AI-901 Exam Tips

For the exam, remember these key points:

  • Keyword extraction identifies important terms or phrases.
  • Entity detection identifies items such as people, places, organizations, and dates.
  • Sentiment analysis determines emotional tone.
  • Summarization creates shorter versions of text.
  • NLP enables computers to process human language.
  • OCR extracts text from images but is different from text analysis.
  • Summarization may be extractive or abstractive.
  • Text classification assigns categories to text.

Quick Knowledge Check

Question 1

Which text analysis technique identifies emotional tone?

Answer

Sentiment analysis.

Question 2

What does Named Entity Recognition (NER) identify?

Answer

Entities such as people, organizations, locations, and dates.

Question 3

What is the purpose of keyword extraction?

Answer

To identify important words or phrases in text.

Question 4

What does summarization do?

Answer

Creates shorter versions of longer text while preserving key information.

Practice Exam Questions

Question 1

Which text analysis technique identifies the emotional tone of written text?

A. OCR
B. Sentiment analysis
C. Object detection
D. Regression

Correct Answer

B. Sentiment analysis

Explanation

Sentiment analysis determines whether text expresses positive, negative, or neutral emotions or opinions.

Why the Other Answers Are Incorrect

A. OCR

OCR extracts text from images or scanned documents.

C. Object detection

Object detection identifies objects within images.

D. Regression

Regression predicts numeric values.

Question 2

A company wants to automatically identify important phrases from customer feedback forms.

Which text analysis technique is MOST appropriate?

A. Speech synthesis
B. Keyword extraction
C. Facial recognition
D. Image classification

Correct Answer

B. Keyword extraction

Explanation

Keyword extraction identifies the most important words or phrases within text.

Why the Other Answers Are Incorrect

A. Speech synthesis

Speech synthesis converts text into spoken audio.

C. Facial recognition

Facial recognition analyzes faces in images.

D. Image classification

Image classification categorizes images.

Question 3

What is the PRIMARY purpose of Named Entity Recognition (NER)?

A. Predicting future sales
B. Identifying important entities such as people, organizations, and locations in text
C. Translating languages automatically
D. Detecting objects in images

Correct Answer

B. Identifying important entities such as people, organizations, and locations in text

Explanation

NER extracts structured information from text by identifying entities like names, places, dates, and organizations.

Why the Other Answers Are Incorrect

A. Predicting future sales

This is typically a regression task.

C. Translating languages automatically

Translation is a separate NLP capability.

D. Detecting objects in images

This is a computer vision task.

Question 4

Which AI capability creates a shorter version of a document while preserving key information?

A. OCR
B. Summarization
C. Clustering
D. Object detection

Correct Answer

B. Summarization

Explanation

Summarization condenses long text into shorter, meaningful summaries.

Why the Other Answers Are Incorrect

A. OCR

OCR extracts text from images.

C. Clustering

Clustering groups similar data.

D. Object detection

Object detection identifies items within images.

Question 5

A business analyzes product reviews to determine whether customers are satisfied or dissatisfied.

Which AI technique is being used?

A. Sentiment analysis
B. Recommendation system
C. OCR
D. Regression

Correct Answer

A. Sentiment analysis

Explanation

Sentiment analysis evaluates emotional tone and opinions expressed in text.

Why the Other Answers Are Incorrect

B. Recommendation system

Recommendation systems suggest products or content.

C. OCR

OCR extracts text from images.

D. Regression

Regression predicts numeric outcomes.

Question 6

Which statement BEST describes keyword extraction?

A. It converts speech into text
B. It identifies important words or phrases in text
C. It translates text between languages
D. It predicts future trends

Correct Answer

B. It identifies important words or phrases in text

Explanation

Keyword extraction helps determine the main topics or themes within text documents.

Why the Other Answers Are Incorrect

A. It converts speech into text

This is speech recognition.

C. It translates text between languages

This is machine translation.

D. It predicts future trends

This is unrelated to keyword extraction.

Question 7

Which text analysis technique would MOST likely identify “Microsoft” as an organization and “Seattle” as a location?

A. Entity detection
B. Sentiment analysis
C. Speech recognition
D. Image segmentation

Correct Answer

A. Entity detection

Explanation

Entity detection (NER) identifies named entities such as organizations, locations, dates, and people within text.

Why the Other Answers Are Incorrect

B. Sentiment analysis

Sentiment analysis evaluates emotional tone.

C. Speech recognition

Speech recognition processes audio.

D. Image segmentation

Image segmentation is a computer vision task.

Question 8

What is the difference between extractive and abstractive summarization?

A. Extractive summarization uses images, while abstractive summarization uses text
B. Extractive summarization selects sentences from the original text, while abstractive summarization generates new summary wording
C. Extractive summarization only works with speech
D. There is no difference

Correct Answer

B. Extractive summarization selects sentences from the original text, while abstractive summarization generates new summary wording

Explanation

Extractive summarization pulls existing sentences directly from text, while abstractive summarization creates newly generated summaries.

Why the Other Answers Are Incorrect

A. Extractive summarization uses images, while abstractive summarization uses text

Both methods work with text.

C. Extractive summarization only works with speech

Summarization is generally text-based.

D. There is no difference

The two methods are different approaches.

Question 9

Which AI workload category includes keyword extraction, sentiment analysis, and summarization?

A. Computer vision
B. Text analysis
C. Robotics
D. Regression analysis

Correct Answer

B. Text analysis

Explanation

These techniques are part of Natural Language Processing (NLP) and text analysis workloads.

Why the Other Answers Are Incorrect

A. Computer vision

Computer vision focuses on images and video.

C. Robotics

Robotics involves physical machines and automation.

D. Regression analysis

Regression predicts numeric values.

Question 10

A company wants to process thousands of support tickets and automatically identify the most common customer complaints.

Which AI technique would be MOST useful?

A. Object detection
B. Keyword extraction
C. Facial recognition
D. Speech synthesis

Correct Answer

B. Keyword extraction

Explanation

Keyword extraction identifies recurring important phrases and themes within large collections of text.

Why the Other Answers Are Incorrect

A. Object detection

Object detection analyzes images.

C. Facial recognition

Facial recognition identifies people in images or video.

D. Speech synthesis

Speech synthesis converts text into audio.

Final Thoughts

Text analysis is a foundational AI workload and an important topic for the AI-901 certification exam. Microsoft expects candidates to understand common NLP techniques and recognize real-world scenarios where text analysis provides value.

These capabilities help organizations transform large volumes of unstructured text into actionable insights using Azure AI technologies.

Go to the AI-901 Exam Prep Hub main page

AI, AI-901, Artificial Intelligence (AI), Microsoft Certification

Identify features and capabilities of Speech Recognition and Speech Synthesis (AI-901 Exam Prep)

 
This post is a part of the AI-901: Microsoft Azure AI Fundamentals Exam Prep Hub. 
This topic falls under these sections:
Identify AI concepts and capabilities (40–45%)
   --> Identify AI workloads
      --> Identify features and capabilities of Speech Recognition and Speech Synthesis

Note that there are 10 practice questions (with answers and explanations) for each section to help you solidify your knowledge of the material. Also, there are 2 practice tests with 60 questions each available on the hub below the exam topics section.

AI-901: Microsoft Azure AI Fundamentals (beta)

Speech AI is one of the major AI workloads covered in the AI-901 certification exam. Microsoft expects candidates to understand how AI systems process spoken language using technologies such as speech recognition and speech synthesis.

These capabilities allow computers to listen to, understand, and generate human speech, enabling more natural human-computer interaction.

This topic falls under the “Identify AI workloads” section of the AI-901 exam objectives.

What Is Speech AI?

Speech AI refers to AI technologies that process spoken language and audio.

Speech AI enables systems to:

  • Recognize spoken words
  • Convert speech into text
  • Generate spoken responses
  • Translate spoken language
  • Identify speakers

Speech technologies are commonly used in modern AI assistants and accessibility tools.

Major Speech AI Capabilities

For the AI-901 exam, important speech AI capabilities include:

  • Speech recognition
  • Speech synthesis
  • Speech translation
  • Speaker recognition

The primary focus of this topic is speech recognition and speech synthesis.

What Is Speech Recognition?

Speech recognition converts spoken language into written text.

It is often called:

  • Speech-to-text
  • Automatic Speech Recognition (ASR)

Speech recognition allows computers to “listen” to human speech and interpret it as text.

How Speech Recognition Works

Speech recognition systems typically perform these steps:

  1. Capture audio input
  2. Analyze sound patterns
  3. Identify spoken words
  4. Convert speech into text output

Modern speech recognition systems often use:

  • Machine learning
  • Deep learning
  • Neural networks
  • Large speech datasets

Speech Recognition Example

Spoken Input

“Schedule a meeting for tomorrow at 2 PM.”

Text Output

Schedule a meeting for tomorrow at 2 PM.

Common Features of Speech Recognition

Speech recognition systems may support:

  • Real-time transcription
  • Multiple languages
  • Noise reduction
  • Speaker identification
  • Continuous speech recognition
  • Command recognition

Real-Time Transcription

Real-time transcription converts speech into text immediately as someone speaks.

Common Use Cases

  • Live captions
  • Meeting transcription
  • Accessibility tools

Example

A video conferencing platform generates live subtitles during meetings.

Continuous Speech Recognition

Continuous speech recognition processes natural conversation without requiring pauses between words.

Example

Voice assistants processing full spoken sentences naturally.

Command Recognition

Some speech systems focus on recognizing specific spoken commands.

Example Commands

  • “Play music”
  • “Turn on the lights”
  • “Call John”

These systems are commonly used in smart devices.

Noise Reduction

Speech recognition systems often include noise filtering capabilities.

This helps improve accuracy in noisy environments.

Example

Recognizing speech in a crowded airport.

Multilingual Speech Recognition

Many modern speech systems support multiple languages and accents.

Example

An AI assistant understanding English, Spanish, and French speakers.

Common Use Cases for Speech Recognition

Virtual Assistants

Examples include voice-controlled assistants that answer questions or perform actions.

Example

A user asks a smart speaker about the weather.

Accessibility Solutions

Speech recognition helps users who cannot type easily.

Example

Voice dictation software for users with disabilities.

Meeting Transcription

Organizations convert meetings into searchable text records.

Example

Automatic meeting notes.

Customer Service Systems

Interactive voice response (IVR) systems process spoken customer requests.

Example

A phone system asks customers to describe their issue verbally.

Hands-Free Applications

Speech recognition supports hands-free operation.

Example

Voice-controlled navigation while driving.

What Is Speech Synthesis?

Speech synthesis converts written text into spoken audio.

It is often called:

  • Text-to-speech (TTS)

Speech synthesis allows computers to “speak” naturally to users.

How Speech Synthesis Works

Speech synthesis systems:

  1. Receive text input
  2. Analyze words and pronunciation
  3. Generate spoken audio output

Modern systems use AI-generated voices that sound increasingly human-like.

Speech Synthesis Example

Text Input

“Your appointment is scheduled for Monday at 10 AM.”

Spoken Output

The system reads the message aloud.

Features of Speech Synthesis

Speech synthesis systems may support:

  • Natural-sounding voices
  • Multiple languages
  • Adjustable speaking speed
  • Voice customization
  • Emotional tone control

Natural Neural Voices

Modern AI systems use neural text-to-speech technology to create more human-like speech.

Benefits include:

  • Improved pronunciation
  • Better intonation
  • More natural rhythm

Voice Customization

Some systems allow organizations to customize voices.

Example

A company creates a branded AI voice for customer support systems.

Adjustable Speech Settings

Speech synthesis systems may allow changes to:

  • Speed
  • Pitch
  • Volume
  • Pronunciation

Common Use Cases for Speech Synthesis

Accessibility Tools

Text-to-speech helps visually impaired users consume written content.

Example

Screen readers reading web pages aloud.

Navigation Systems

GPS applications provide spoken directions.

Example

A navigation app announcing upcoming turns.

Virtual Assistants

AI assistants respond using synthesized speech.

Example

A smart assistant answers spoken questions aloud.

Customer Service Bots

Automated phone systems communicate using AI-generated voices.

Example

A banking system reads account information to customers.

Audiobooks and Learning

Speech synthesis converts written content into audio.

Example

Educational content read aloud automatically.

Speech Translation

Speech translation combines:

  • Speech recognition
  • Language translation
  • Speech synthesis

Example

A conference tool translates spoken English into spoken Spanish.

Speaker Recognition

Speaker recognition identifies or verifies individuals based on voice characteristics.

Types

  • Speaker identification
  • Speaker verification

Example

Voice-based authentication systems.

Challenges in Speech AI

Speech AI systems may face challenges such as:

  • Background noise
  • Strong accents
  • Multiple simultaneous speakers
  • Poor audio quality
  • Specialized vocabulary

Responsible AI Considerations

Speech AI systems should be designed responsibly.

Important considerations include:

  • Privacy
  • Consent
  • Security
  • Accessibility
  • Bias reduction
  • Transparency

Voice data may contain sensitive personal information.

Azure AI Speech Services

Azure AI Speech Services provide cloud-based speech AI capabilities including:

  • Speech-to-text
  • Text-to-speech
  • Speech translation
  • Speaker recognition

These services help developers integrate speech AI into applications without building models from scratch.

Speech Recognition vs. Speech Synthesis

CapabilityDescription
Speech RecognitionConverts speech into text
Speech SynthesisConverts text into spoken audio

Real-World Examples

Scenario 1: Live Meeting Captions

Goal

Convert spoken conversations into text.

Capability Used

Speech recognition

Scenario 2: GPS Navigation App

Goal

Read directions aloud.

Capability Used

Speech synthesis

Scenario 3: Voice-Controlled Smart Home

Goal

Understand spoken commands and respond verbally.

Capabilities Used

  • Speech recognition
  • Speech synthesis

Scenario 4: Audiobook Generator

Goal

Convert books into spoken audio.

Capability Used

Speech synthesis

Important AI-901 Exam Tips

For the exam, remember these key points:

  • Speech recognition converts speech into text.
  • Speech synthesis converts text into spoken audio.
  • Speech-to-text is another term for speech recognition.
  • Text-to-speech is another term for speech synthesis.
  • Real-time transcription supports live captions.
  • Neural voices produce more natural speech.
  • Speech translation combines multiple speech technologies.
  • Speaker recognition identifies individuals using voice characteristics.
  • Speech AI is commonly used in assistants, accessibility tools, and customer service systems.

Quick Knowledge Check

Question 1

What does speech recognition do?

Answer

Converts spoken language into written text.

Question 2

What does speech synthesis do?

Answer

Converts text into spoken audio.

Question 3

What is another name for speech synthesis?

Answer

Text-to-speech (TTS).

Question 4

Which speech capability is used for live meeting captions?

Answer

Speech recognition.

Practice Exam Questions

Question 1

What is the PRIMARY function of speech recognition?

A. Converting images into text
B. Converting spoken language into written text
C. Generating images from prompts
D. Translating text into code

Correct Answer

B. Converting spoken language into written text

Explanation

Speech recognition, also called speech-to-text, converts spoken audio into written text.

Why the Other Answers Are Incorrect

A. Converting images into text

This is OCR functionality.

C. Generating images from prompts

This is a generative AI capability.

D. Translating text into code

This is unrelated to speech recognition.

Question 2

Which capability converts written text into spoken audio?

A. OCR
B. Speech synthesis
C. Object detection
D. Clustering

Correct Answer

B. Speech synthesis

Explanation

Speech synthesis, also called text-to-speech (TTS), generates spoken audio from text.

Why the Other Answers Are Incorrect

A. OCR

OCR extracts text from images.

C. Object detection

Object detection identifies objects in images.

D. Clustering

Clustering groups similar data.

Question 3

A company wants to generate live subtitles during online meetings.

Which AI capability should be used?

A. Speech recognition
B. Speech synthesis
C. Facial recognition
D. Image segmentation

Correct Answer

A. Speech recognition

Explanation

Speech recognition converts spoken conversations into text in real time, enabling live captions and subtitles.

Why the Other Answers Are Incorrect

B. Speech synthesis

Speech synthesis creates spoken audio from text.

C. Facial recognition

Facial recognition analyzes faces in images.

D. Image segmentation

Image segmentation is a computer vision task.

Question 4

What is another common name for speech synthesis?

A. Object detection
B. Text-to-speech
C. Speech-to-text
D. Named Entity Recognition

Correct Answer

B. Text-to-speech

Explanation

Speech synthesis is commonly referred to as text-to-speech (TTS).

Why the Other Answers Are Incorrect

A. Object detection

Object detection identifies objects in images.

C. Speech-to-text

Speech-to-text refers to speech recognition.

D. Named Entity Recognition

NER identifies entities in text.

Question 5

Which scenario BEST demonstrates speech synthesis?

A. A chatbot reading answers aloud to users
B. A camera identifying vehicles on a road
C. A system categorizing customer emails
D. A database sorting sales records

Correct Answer

A. A chatbot reading answers aloud to users

Explanation

Speech synthesis converts text responses into spoken audio for users.

Why the Other Answers Are Incorrect

B. A camera identifying vehicles on a road

This is computer vision.

C. A system categorizing customer emails

This is text classification.

D. A database sorting sales records

This is not a speech AI task.

Question 6

Which feature helps speech recognition systems perform better in noisy environments?

A. Image enhancement
B. Noise reduction
C. OCR optimization
D. Regression tuning

Correct Answer

B. Noise reduction

Explanation

Noise reduction filters background sounds to improve speech recognition accuracy.

Why the Other Answers Are Incorrect

A. Image enhancement

Image enhancement relates to visual processing.

C. OCR optimization

OCR works with images and text extraction.

D. Regression tuning

Regression is unrelated to speech audio processing.

Question 7

A navigation application reads driving directions aloud to users.

Which AI capability is being used?

A. Sentiment analysis
B. Speech synthesis
C. Object detection
D. Language detection

Correct Answer

B. Speech synthesis

Explanation

Speech synthesis converts written navigation instructions into spoken audio.

Why the Other Answers Are Incorrect

A. Sentiment analysis

Sentiment analysis evaluates emotional tone in text.

C. Object detection

Object detection analyzes images.

D. Language detection

Language detection identifies languages in text.

Question 8

Which statement BEST describes speech translation?

A. It converts images into searchable text
B. It combines speech recognition, translation, and speech synthesis
C. It identifies objects in audio recordings
D. It predicts future speech patterns

Correct Answer

B. It combines speech recognition, translation, and speech synthesis

Explanation

Speech translation systems convert spoken language into another language and often generate translated spoken output.

Why the Other Answers Are Incorrect

A. It converts images into searchable text

This is OCR.

C. It identifies objects in audio recordings

This is not a standard speech AI capability.

D. It predicts future speech patterns

This is unrelated to translation systems.

Question 9

What is the PRIMARY purpose of speaker recognition?

A. Generating synthetic voices
B. Identifying or verifying individuals using voice characteristics
C. Translating speech into multiple languages
D. Extracting keywords from documents

Correct Answer

B. Identifying or verifying individuals using voice characteristics

Explanation

Speaker recognition systems use voice patterns to identify or authenticate users.

Why the Other Answers Are Incorrect

A. Generating synthetic voices

This is speech synthesis.

C. Translating speech into multiple languages

This is speech translation.

D. Extracting keywords from documents

This is keyword extraction.

Question 10

Which pair correctly matches the capability with its function?

A. Speech recognition → Converts text into speech
B. Speech synthesis → Converts speech into text
C. Speech recognition → Converts speech into text
D. OCR → Generates spoken audio

Correct Answer

C. Speech recognition → Converts speech into text

Explanation

Speech recognition converts spoken language into written text, while speech synthesis converts text into spoken audio.

Why the Other Answers Are Incorrect

A. Speech recognition → Converts text into speech

This describes speech synthesis.

B. Speech synthesis → Converts speech into text

This describes speech recognition.

D. OCR → Generates spoken audio

OCR extracts text from images.

Final Thoughts

Speech AI technologies are essential components of modern AI systems and are an important topic for the AI-901 certification exam. Microsoft expects candidates to understand how speech recognition and speech synthesis work, along with common business scenarios where these technologies are applied.

These capabilities help organizations build more natural, accessible, and interactive AI-powered experiences using Azure AI services.

Go to the AI-901 Exam Prep Hub main page

AI, AI-901, Computer Vision, Generative AI, Microsoft Certification

Identify features and capabilities of Computer Vision and Image-Generation models (AI-901 Exam Prep)

 
This post is a part of the AI-901: Microsoft Azure AI Fundamentals Exam Prep Hub. 
This topic falls under these sections:
Identify AI concepts and capabilities (40–45%)
   --> Identify AI workloads
      --> Identify features and capabilities of Computer Vision and Image-Generation models

Note that there are 10 practice questions (with answers and explanations) for each section to help you solidify your knowledge of the material. Also, there are 2 practice tests with 60 questions each available on the hub below the exam topics section.

Computer vision and image-generation AI models are important AI workloads covered in the AI-901 certification exam. Microsoft expects candidates to understand how AI systems analyze visual information and generate new images using machine learning and deep learning technologies.

These AI capabilities are widely used in healthcare, manufacturing, security, retail, entertainment, accessibility, and many other industries.

This topic falls under the “Identify AI workloads” section of the AI-901 exam objectives.

What Is Computer Vision?

Computer vision is an AI workload that enables computers to analyze and interpret images and video.

Computer vision systems attempt to simulate human visual understanding.

These systems can:

  • Identify objects
  • Detect faces
  • Read text
  • Analyze scenes
  • Track movement
  • Recognize patterns

How Computer Vision Works

Computer vision models are typically trained using large collections of labeled images.

The models learn patterns such as:

  • Shapes
  • Colors
  • Textures
  • Edges
  • Spatial relationships

Modern computer vision systems commonly use:

  • Deep learning
  • Neural networks
  • Convolutional Neural Networks (CNNs)

Common Computer Vision Capabilities

For the AI-901 exam, important computer vision capabilities include:

  • Image classification
  • Object detection
  • Facial recognition
  • Optical Character Recognition (OCR)
  • Image analysis
  • Image tagging

Image Classification

Image classification identifies the primary subject or category of an image.

The model assigns labels to entire images.

Image Classification Example

Input

An image of a dog.

Output

“Dog”

Common Use Cases for Image Classification

Medical Imaging

Classifying medical scans.

Retail

Categorizing products automatically.

Agriculture

Identifying plant diseases.

Wildlife Monitoring

Recognizing animal species.

Object Detection

Object detection identifies and locates multiple objects within an image.

Unlike image classification, object detection can identify several objects and their positions.

Object Detection Example

Input

Street traffic image.

Output

  • Car
  • Pedestrian
  • Traffic light

with location boundaries around each object.

Common Use Cases for Object Detection

Autonomous Vehicles

Detecting vehicles and pedestrians.

Manufacturing

Identifying defective products.

Security Systems

Detecting unauthorized activity.

Retail Analytics

Monitoring customer movement in stores.

Facial Recognition

Facial recognition identifies or verifies individuals using facial features.

Common Facial Recognition Capabilities

Face Detection

Determines whether faces exist in an image.

Face Verification

Confirms whether two faces belong to the same person.

Face Identification

Identifies a person from a database of known individuals.

Common Use Cases for Facial Recognition

Smartphone Authentication

Unlocking phones using facial recognition.

Building Security

Controlling physical access.

Attendance Systems

Tracking employee attendance.

Airport Security

Identity verification systems.

Optical Character Recognition (OCR)

OCR extracts text from images, scanned documents, or photographs.

OCR converts visual text into machine-readable text.

OCR Example

Input

A scanned invoice image.

Output

Extracted text including:

  • Invoice number
  • Dates
  • Totals

Common OCR Use Cases

Invoice Processing

Automating financial workflows.

Document Digitization

Converting paper documents into searchable digital text.

Receipt Scanning

Extracting purchase information.

Accessibility

Reading text aloud for visually impaired users.

Image Tagging and Image Analysis

Image analysis systems can automatically generate descriptions or tags for images.

Example Tags

An image may receive tags such as:

  • Beach
  • Ocean
  • Sunset
  • Person

Common Use Cases

Photo Organization

Automatically categorizing photos.

Content Moderation

Identifying inappropriate images.

Search Optimization

Improving image search systems.

Video Analysis

Computer vision can also process video streams.

Common Video Analysis Tasks

  • Motion detection
  • Activity recognition
  • Traffic monitoring
  • Surveillance analysis

What Are Image-Generation Models?

Image-generation models create new images using AI.

These models learn visual patterns from training data and generate entirely new content.

Image-generation AI is part of generative AI.

How Image-Generation Models Work

Image-generation systems are trained on large image datasets.

The models learn relationships between:

  • Objects
  • Colors
  • Styles
  • Shapes
  • Text descriptions

Many systems use deep learning architectures such as:

  • Diffusion models
  • Generative Adversarial Networks (GANs)

Text-to-Image Generation

Text-to-image models generate images from written prompts.

Example

Prompt

“A futuristic city at sunset”

Output

An AI-generated image matching the description.

Common Use Cases for Image Generation

Marketing and Advertising

Creating promotional graphics.

Entertainment and Gaming

Generating concept art.

Design Assistance

Creating mockups or creative inspiration.

Education

Generating visual learning content.

Accessibility

Creating visual representations from text descriptions.

Image Editing and Enhancement

Some AI models can edit or enhance existing images.

Common Capabilities

  • Background removal
  • Image restoration
  • Colorization
  • Resolution enhancement
  • Style transfer

Deepfakes and Synthetic Media

AI-generated images and videos can create highly realistic synthetic content.

This technology can be useful but also creates ethical concerns.

Responsible AI Considerations

Computer vision and image-generation systems raise important Responsible AI considerations.

Organizations should consider:

  • Privacy
  • Consent
  • Bias
  • Security
  • Transparency
  • Misuse prevention

Bias in Vision Models

Computer vision systems may perform differently across demographic groups if training data is unbalanced.

Example risks include:

  • Facial recognition inaccuracies
  • Biased image classification
  • Unequal detection accuracy

Ethical Concerns with Image Generation

Potential concerns include:

  • Deepfakes
  • Misinformation
  • Copyright concerns
  • Identity misuse
  • Harmful content generation

Organizations should implement safeguards and moderation systems.

Azure AI Vision Services

Azure AI Vision Services provide prebuilt computer vision capabilities including:

  • Image analysis
  • OCR
  • Face detection
  • Object detection
  • Video analysis

Azure OpenAI and Image Generation

Azure OpenAI Service supports generative AI capabilities, including image-generation models.

These services help organizations build AI-powered creative applications.

Computer Vision vs. Image Generation

CapabilityPurpose
Computer VisionAnalyze and understand images
Image GenerationCreate new images

Real-World Examples

Scenario 1: Self-Driving Car

Goal

Detect vehicles and pedestrians.

Capability Used

Object detection

Scenario 2: Receipt Scanning App

Goal

Extract text from receipts.

Capability Used

OCR

Scenario 3: Social Media Photo Organization

Goal

Automatically tag uploaded photos.

Capability Used

Image analysis and tagging

Scenario 4: AI Art Generator

Goal

Create artwork from text prompts.

Capability Used

Image generation

Scenario 5: Smartphone Face Unlock

Goal

Verify user identity.

Capability Used

Facial recognition

Important AI-901 Exam Tips

For the exam, remember these key points:

  • Computer vision analyzes images and video.
  • Image classification labels entire images.
  • Object detection identifies and locates objects.
  • OCR extracts text from images.
  • Facial recognition identifies or verifies individuals.
  • Image-generation models create new images.
  • Text-to-image systems generate visuals from prompts.
  • Computer vision and generative AI are different workloads.
  • Responsible AI principles are important in vision systems.

Quick Knowledge Check

Question 1

What is the purpose of OCR?

Answer

To extract text from images or scanned documents.

Question 2

What is the difference between image classification and object detection?

Answer

Image classification labels an entire image, while object detection identifies and locates multiple objects within an image.

Question 3

What do image-generation models do?

Answer

They create new images using AI.

Question 4

Which AI capability is commonly used for smartphone face unlock?

Answer

Facial recognition.

Practice Exam Questions

Question 1

What is the PRIMARY purpose of computer vision?

A. Converting speech into text
B. Analyzing and understanding images and video
C. Predicting stock prices
D. Generating database queries

Correct Answer

B. Analyzing and understanding images and video

Explanation

Computer vision enables AI systems to interpret and analyze visual content such as images and video.

Why the Other Answers Are Incorrect

A. Converting speech into text

This is speech recognition.

C. Predicting stock prices

This is typically a regression task.

D. Generating database queries

This is unrelated to computer vision.

Question 2

Which computer vision capability identifies the main subject or category of an image?

A. OCR
B. Image classification
C. Speech synthesis
D. Clustering

Correct Answer

B. Image classification

Explanation

Image classification assigns labels or categories to entire images.

Why the Other Answers Are Incorrect

A. OCR

OCR extracts text from images.

C. Speech synthesis

Speech synthesis converts text into spoken audio.

D. Clustering

Clustering groups similar data.

Question 3

A self-driving car needs to identify pedestrians, traffic signs, and vehicles in real time.

Which AI capability is MOST appropriate?

A. Sentiment analysis
B. Object detection
C. Keyword extraction
D. Language detection

Correct Answer

B. Object detection

Explanation

Object detection identifies and locates multiple objects within images or video streams.

Why the Other Answers Are Incorrect

A. Sentiment analysis

Sentiment analysis evaluates emotional tone in text.

C. Keyword extraction

Keyword extraction identifies important phrases in text.

D. Language detection

Language detection identifies written languages.

Question 4

What is the PRIMARY purpose of Optical Character Recognition (OCR)?

A. Translating speech between languages
B. Extracting text from images or scanned documents
C. Detecting faces in photographs
D. Generating new artwork

Correct Answer

B. Extracting text from images or scanned documents

Explanation

OCR converts text within images into machine-readable text.

Why the Other Answers Are Incorrect

A. Translating speech between languages

This is speech translation.

C. Detecting faces in photographs

This is facial recognition or face detection.

D. Generating new artwork

This is an image-generation capability.

Question 5

Which AI capability is commonly used for smartphone face unlock features?

A. Facial recognition
B. Speech recognition
C. Regression
D. Text summarization

Correct Answer

A. Facial recognition

Explanation

Facial recognition systems identify or verify users using facial features.

Why the Other Answers Are Incorrect

B. Speech recognition

Speech recognition processes spoken language.

C. Regression

Regression predicts numeric values.

D. Text summarization

Summarization condenses text.

Question 6

What is the PRIMARY function of image-generation models?

A. Extracting text from images
B. Creating new images using AI
C. Detecting network intrusions
D. Translating written languages

Correct Answer

B. Creating new images using AI

Explanation

Image-generation models produce new visual content based on learned patterns and prompts.

Why the Other Answers Are Incorrect

A. Extracting text from images

This is OCR.

C. Detecting network intrusions

This is unrelated to image generation.

D. Translating written languages

This is an NLP capability.

Question 7

Which example BEST represents a text-to-image generation system?

A. A chatbot answering questions
B. An AI model creating artwork from a written prompt
C. A speech recognition application
D. A recommendation engine

Correct Answer

B. An AI model creating artwork from a written prompt

Explanation

Text-to-image systems generate images based on textual descriptions.

Why the Other Answers Are Incorrect

A. A chatbot answering questions

This is generative text AI.

C. A speech recognition application

Speech recognition converts speech into text.

D. A recommendation engine

Recommendation systems suggest products or content.

Question 8

What is the key difference between image classification and object detection?

A. Image classification processes audio while object detection processes video
B. Image classification labels an entire image, while object detection identifies and locates multiple objects
C. Object detection only works with text
D. There is no difference

Correct Answer

B. Image classification labels an entire image, while object detection identifies and locates multiple objects

Explanation

Image classification provides a label for an entire image, while object detection identifies multiple objects and their locations.

Why the Other Answers Are Incorrect

A. Image classification processes audio while object detection processes video

Both work with visual data.

C. Object detection only works with text

Object detection works with images and video.

D. There is no difference

These are distinct computer vision tasks.

Question 9

Which Responsible AI concern is MOST associated with image-generation systems?

A. Deepfakes and synthetic media misuse
B. Spreadsheet formatting errors
C. SQL indexing problems
D. Network bandwidth allocation

Correct Answer

A. Deepfakes and synthetic media misuse

Explanation

Image-generation AI can create highly realistic synthetic content, raising concerns about misinformation and misuse.

Why the Other Answers Are Incorrect

B. Spreadsheet formatting errors

This is unrelated to AI image generation.

C. SQL indexing problems

This is a database issue.

D. Network bandwidth allocation

This is unrelated to Responsible AI concerns.

Question 10

A retailer wants to automatically categorize product photos into categories such as shoes, shirts, and electronics.

Which AI capability is MOST appropriate?

A. Image classification
B. OCR
C. Speech synthesis
D. Sentiment analysis

Correct Answer

A. Image classification

Explanation

Image classification assigns category labels to images based on visual content.

Why the Other Answers Are Incorrect

B. OCR

OCR extracts text from images.

C. Speech synthesis

Speech synthesis generates spoken audio.

D. Sentiment analysis

Sentiment analysis evaluates emotional tone in text.

Final Thoughts

Computer vision and image-generation AI models are essential components of modern AI systems and important topics for the AI-901 certification exam. Microsoft expects candidates to understand how AI systems analyze visual information and generate new content, along with common business scenarios where these technologies are applied.

These capabilities help organizations build intelligent visual applications using Azure AI services and generative AI technologies.

Go to the AI-901 Exam Prep Hub main page

AI, AI-901, Artificial Intelligence (AI), azure, Microsoft Certification

Identify techniques to extract information from text, images, audio, and videos (AI-901 Exam Prep)

 
This post is a part of the AI-901: Microsoft Azure AI Fundamentals Exam Prep Hub. 
This topic falls under these sections:
Identify AI concepts and capabilities (40–45%)
   --> Identify AI workloads
      --> Identify techniques to extract information from text, images, audio, and videos

Note that there are 10 practice questions (with answers and explanations) for each section to help you solidify your knowledge of the material. Also, there are 2 practice tests with 60 questions each available on the hub below the exam topics section.

Information extraction is one of the most valuable uses of AI and an important topic for the AI-901 certification exam. Organizations generate enormous amounts of unstructured data every day, including documents, emails, images, audio recordings, and videos. AI systems help convert this unstructured data into structured, usable information.

Microsoft expects AI-901 candidates to understand common techniques used to extract information from text, images, audio, and video content.

This topic falls under the “Identify AI workloads” section of the AI-901 exam objectives.

What Is Information Extraction?

Information extraction is the process of identifying and retrieving useful structured information from unstructured or semi-structured data.

AI systems analyze content and extract meaningful data automatically.

Examples of Information Extraction

SourceExtracted Information
DocumentsNames, dates, invoice totals
EmailsCustomer requests, keywords
ImagesObjects, faces, text
AudioSpoken words, speaker identity
VideoActivities, objects, movement

Structured vs. Unstructured Data

Understanding structured and unstructured data is important for this topic.

Structured DataUnstructured Data
TablesEmails
DatabasesImages
SpreadsheetsAudio
Defined formatsVideos
Organized fieldsDocuments

AI techniques help transform unstructured data into structured information.

Information Extraction from Text

AI systems commonly use Natural Language Processing (NLP) to extract information from text.

Common Text Extraction Techniques

For the AI-901 exam, important techniques include:

  • Keyword extraction
  • Named Entity Recognition (NER)
  • Sentiment analysis
  • Summarization
  • Language detection
  • Text classification

Keyword Extraction

Keyword extraction identifies important words or phrases within text.

Example

Extracting phrases like:

  • “shipping delay”
  • “billing issue”
  • “customer satisfaction”

from support tickets.

Named Entity Recognition (NER)

NER identifies entities such as:

  • People
  • Organizations
  • Locations
  • Dates
  • Phone numbers
  • Products

Example

Input

“Microsoft will host an event in Seattle on June 15.”

Extracted Entities

  • Microsoft → Organization
  • Seattle → Location
  • June 15 → Date

Sentiment Analysis

Sentiment analysis identifies emotional tone within text.

Possible Results

  • Positive
  • Negative
  • Neutral

Example

Analyzing customer reviews to determine satisfaction levels.

Summarization

Summarization creates shorter versions of long text.

Example

Generating meeting summaries from lengthy transcripts.

Text Classification

Text classification assigns categories to text.

Example

Automatically labeling emails as:

  • Support
  • Sales
  • Billing

Information Extraction from Images

Computer vision techniques extract information from images.

Common Image Extraction Techniques

Important techniques include:

  • OCR
  • Image classification
  • Object detection
  • Facial recognition
  • Image tagging

Optical Character Recognition (OCR)

OCR extracts text from images and scanned documents.

OCR Example

Input

Scanned invoice image.

Extracted Information

  • Invoice number
  • Total amount
  • Vendor name
  • Dates

Common OCR Use Cases

  • Receipt scanning
  • Invoice processing
  • Document digitization
  • Form extraction

Image Classification

Image classification identifies the overall category of an image.

Example

Identifying whether an image contains:

  • A dog
  • A car
  • A building

Object Detection

Object detection identifies and locates multiple objects within images.

Example

Detecting:

  • Cars
  • Pedestrians
  • Traffic lights

in a street image.

Facial Recognition

Facial recognition identifies or verifies people based on facial features.

Example

Smartphone face unlock systems.

Image Tagging

Image tagging automatically generates descriptive labels.

Example Tags

  • Beach
  • Sunset
  • Ocean
  • Person

Information Extraction from Audio

Speech AI technologies extract information from spoken audio.

Common Audio Extraction Techniques

Important techniques include:

  • Speech recognition
  • Speaker recognition
  • Sentiment analysis in speech
  • Speech translation

Speech Recognition

Speech recognition converts spoken language into text.

Also called:

  • Speech-to-text
  • Automatic Speech Recognition (ASR)

Example

Audio Input

A recorded meeting.

Extracted Information

A written transcript.

Speaker Recognition

Speaker recognition identifies or verifies speakers based on voice characteristics.

Example

Voice authentication systems.

Speech Sentiment Analysis

Some AI systems analyze vocal tone and emotion.

Example

Detecting frustration during customer service calls.

Speech Translation

Speech translation converts spoken language into another language.

Example

Real-time multilingual meeting translation.

Information Extraction from Video

Video analysis combines computer vision and audio processing techniques.

Common Video Extraction Techniques

Important techniques include:

  • Motion detection
  • Object tracking
  • Activity recognition
  • Scene analysis
  • Video transcription

Motion Detection

Motion detection identifies movement within video footage.

Example

Security surveillance systems detecting activity.

Object Tracking

Object tracking follows identified objects across video frames.

Example

Tracking vehicles in traffic monitoring systems.

Activity Recognition

Activity recognition identifies actions occurring in video.

Example

Detecting:

  • Running
  • Falling
  • Fighting
  • Driving

Scene Analysis

Scene analysis identifies environments or contexts in video.

Example

Recognizing:

  • Office scenes
  • Outdoor settings
  • Crowded areas

Video Transcription

Video transcription converts spoken content in videos into text.

Example

Generating subtitles for videos automatically.

Multimodal AI

Some AI systems combine multiple data types together.

This is called multimodal AI.

Example of Multimodal AI

A meeting assistant may process:

  • Audio
  • Video
  • Text chat
  • Shared documents

simultaneously.

Real-World Information Extraction Scenarios

Scenario 1: Invoice Processing System

Goal

Extract invoice information automatically.

Techniques Used

  • OCR
  • Entity extraction

Scenario 2: Customer Support Analysis

Goal

Analyze customer complaints.

Techniques Used

  • Sentiment analysis
  • Keyword extraction

Scenario 3: Smart Security Camera

Goal

Detect suspicious activity.

Techniques Used

  • Object detection
  • Motion detection
  • Facial recognition

Scenario 4: Meeting Intelligence Platform

Goal

Generate searchable meeting notes.

Techniques Used

  • Speech recognition
  • Summarization
  • Speaker recognition

Scenario 5: Video Streaming Platform

Goal

Generate subtitles automatically.

Techniques Used

  • Speech recognition
  • Video transcription

Azure AI Services for Information Extraction

Azure AI Services provide tools for extracting information from multiple data types.

Common services include:

  • Azure AI Language
  • Azure AI Speech
  • Azure AI Vision
  • Azure AI Document Intelligence

These services allow organizations to build AI solutions without training models from scratch.

Responsible AI Considerations

Information extraction systems should follow Responsible AI principles.

Important considerations include:

  • Privacy
  • Consent
  • Data security
  • Transparency
  • Bias reduction
  • Compliance

Sensitive personal information may be present in extracted data.

Challenges in Information Extraction

AI systems may face challenges such as:

  • Poor image quality
  • Background noise
  • Ambiguous language
  • Multiple speakers
  • Handwritten text
  • Video quality issues

Performance depends heavily on data quality.

Important AI-901 Exam Tips

For the exam, remember these key points:

  • NLP extracts information from text.
  • OCR extracts text from images.
  • Speech recognition converts speech into text.
  • Object detection identifies and locates objects in images or video.
  • Video analysis can detect activities and movement.
  • Information extraction converts unstructured data into structured information.
  • Multimodal AI combines multiple data types.
  • Azure AI services provide prebuilt information extraction capabilities.

Quick Knowledge Check

Question 1

Which technique extracts text from scanned documents?

Answer

OCR.

Question 2

What does speech recognition do?

Answer

Converts spoken language into text.

Question 3

Which technique identifies objects within images?

Answer

Object detection.

Question 4

What is multimodal AI?

Answer

AI systems that process multiple types of data together, such as text, audio, and images.

Practice Exam Questions

Question 1

Which AI technique is used to extract text from scanned documents or images?

A. Sentiment analysis
B. Optical Character Recognition (OCR)
C. Object detection
D. Speech synthesis

Correct Answer

B. Optical Character Recognition (OCR)

Explanation

OCR extracts machine-readable text from images, scanned documents, and photographs.

Why the Other Answers Are Incorrect

A. Sentiment analysis

Sentiment analysis identifies emotional tone in text.

C. Object detection

Object detection identifies objects within images.

D. Speech synthesis

Speech synthesis converts text into spoken audio.

Question 2

A company wants to convert recorded customer support calls into written transcripts.

Which AI capability should be used?

A. Speech recognition
B. Facial recognition
C. Image classification
D. Regression

Correct Answer

A. Speech recognition

Explanation

Speech recognition converts spoken language into written text.

Why the Other Answers Are Incorrect

B. Facial recognition

Facial recognition analyzes faces in images.

C. Image classification

Image classification categorizes images.

D. Regression

Regression predicts numeric values.

Question 3

Which AI technique identifies and locates multiple objects within an image?

A. OCR
B. Object detection
C. Summarization
D. Clustering

Correct Answer

B. Object detection

Explanation

Object detection identifies objects and their positions within images or video frames.

Why the Other Answers Are Incorrect

A. OCR

OCR extracts text from images.

C. Summarization

Summarization condenses text.

D. Clustering

Clustering groups similar data points.

Question 4

A business wants to automatically determine whether customer reviews are positive or negative.

Which AI technique is MOST appropriate?

A. Sentiment analysis
B. OCR
C. Facial recognition
D. Image tagging

Correct Answer

A. Sentiment analysis

Explanation

Sentiment analysis evaluates emotional tone and opinions in text.

Why the Other Answers Are Incorrect

B. OCR

OCR extracts text from images.

C. Facial recognition

Facial recognition identifies people from images.

D. Image tagging

Image tagging labels image content.

Question 5

Which AI capability is commonly used to identify names, locations, and organizations within text?

A. Named Entity Recognition (NER)
B. Speech synthesis
C. Object tracking
D. Regression analysis

Correct Answer

A. Named Entity Recognition (NER)

Explanation

NER extracts entities such as people, organizations, dates, and locations from text.

Why the Other Answers Are Incorrect

B. Speech synthesis

Speech synthesis generates spoken audio.

C. Object tracking

Object tracking follows objects in video.

D. Regression analysis

Regression predicts numeric values.

Question 6

A smart security camera tracks moving vehicles across multiple video frames.

Which AI technique is being used?

A. Text classification
B. Object tracking
C. Summarization
D. Speech translation

Correct Answer

B. Object tracking

Explanation

Object tracking follows identified objects as they move through video footage.

Why the Other Answers Are Incorrect

A. Text classification

Text classification categorizes written text.

C. Summarization

Summarization condenses text.

D. Speech translation

Speech translation converts spoken language between languages.

Question 7

Which term describes AI systems that process multiple data types such as text, images, and audio together?

A. Regression AI
B. Multimodal AI
C. Clustering AI
D. Rule-based AI

Correct Answer

B. Multimodal AI

Explanation

Multimodal AI combines and processes multiple forms of data simultaneously.

Why the Other Answers Are Incorrect

A. Regression AI

Regression predicts numeric values.

C. Clustering AI

Clustering groups similar items.

D. Rule-based AI

Rule-based systems follow predefined logic rules.

Question 8

Which AI capability would MOST likely be used to generate automatic subtitles for videos?

A. Speech recognition
B. Image classification
C. Facial recognition
D. Recommendation systems

Correct Answer

A. Speech recognition

Explanation

Speech recognition converts spoken words in videos into text subtitles.

Why the Other Answers Are Incorrect

B. Image classification

Image classification categorizes images.

C. Facial recognition

Facial recognition identifies people in images.

D. Recommendation systems

Recommendation systems suggest content or products.

Question 9

A retailer wants AI to automatically identify products such as shoes, shirts, and electronics in uploaded images.

Which AI capability should be used?

A. Object detection
B. Sentiment analysis
C. Speech synthesis
D. Language translation

Correct Answer

A. Object detection

Explanation

Object detection identifies multiple objects within images and can locate them visually.

Why the Other Answers Are Incorrect

B. Sentiment analysis

Sentiment analysis evaluates text emotion.

C. Speech synthesis

Speech synthesis converts text into speech.

D. Language translation

Language translation converts text or speech between languages.

Question 10

What is the PRIMARY goal of information extraction AI systems?

A. Creating video games
B. Converting unstructured data into useful structured information
C. Compressing database files
D. Replacing all human decision-making

Correct Answer

B. Converting unstructured data into useful structured information

Explanation

Information extraction systems analyze unstructured content such as text, images, audio, and video to retrieve meaningful structured data.

Why the Other Answers Are Incorrect

A. Creating video games

This is unrelated to information extraction.

C. Compressing database files

This is a storage task, not AI extraction.

D. Replacing all human decision-making

AI systems are designed to assist and augment human processes, not completely replace all decision-making.

Final Thoughts

Information extraction is one of the most practical and widely used AI workloads covered in the AI-901 certification exam. Microsoft expects candidates to understand how AI systems extract useful insights from text, images, audio, and videos using NLP, speech AI, computer vision, and multimodal AI technologies.

These capabilities help organizations automate workflows, analyze large volumes of data, and build intelligent applications using Azure AI services.

Go to the AI-901 Exam Prep Hub main page