AI-200 · Domain 3 of 4

AI-200 Domain 3: Connect to and consume Azure services

This domain covers the connective tissue of AI apps: event- and message-based solutions with Azure Service Bus and Event Grid, and serverless APIs and processing with Azure Functions triggers and bindings.

Exam weight
20–25%
Official functional groups
2
Estimated questions estimate
9–14
Suggested study time estimate
15–25 hrs

About 9 min read Reviewed October 2026

Overview

Connect to and consume Azure services covers how AI components communicate and run serverless logic. It has two objective groups. Developing event- and message-based solutions covers queuing and processing back-end operations with Azure Service Bus, including dead-letter queues, messages, topics and subscriptions, and event-driven workflows with Azure Event Grid, including filters, custom events and retries. Developing and implementing Azure Functions covers building serverless APIs with triggers and bindings, and configuring and deploying function apps.

At 20–25%, this group tests the messaging, eventing and serverless patterns that connect AI pipelines. The exam rewards distinguishing messaging (Service Bus) from eventing (Event Grid), knowing how dead-letter queues and topics and subscriptions work, and using Functions triggers and bindings to glue services together. These patterns drive the event-based scaling and processing seen throughout the exam.

How it's tested

AI-200 questions in this group describe a messaging, eventing or serverless need and ask which service and configuration fits. Look for clues in the wording: reliable delivery, ordering or dead-lettering point to Service Bus, while reacting to a resource event points to Event Grid, and a trigger plus binding points to Functions.

  • Messaging or eventing

    A communication need is described, and you choose Service Bus for messaging or Event Grid for eventing.

  • Handle delivery

    A reliability need is described, and you choose dead-letter queues, retries, or topics and subscriptions.

  • Wire the Function

    A serverless processing need is described, and you choose the right trigger and bindings.

Reasoning it through

A back-end must process orders reliably, retry on failure, and move messages that repeatedly fail to a separate location for inspection, with ordered, transactional handling. Event Grid is lightweight event routing without a built-in dead-letter queue for this pattern. Azure Service Bus provides queues with dead-letter queues, retries and transactional, ordered processing, which matches the reliable back-end messaging requirement, so Service Bus is the service.

Functional groups 2

Official functional groups from the Microsoft outline, in outline order.

3.1 Develop event- and message-based AI solutionsAzure Service Bus messaging and Azure Event Grid eventing.
What it means
This covers queuing and processing back-end operations with Azure Service Bus, including dead-letter queue handling, messages, topics and subscriptions, and implementing event-driven workflows with Azure Event Grid, including filters, custom events and retries.
What to understand
Service Bus is enterprise messaging with queues for point-to-point and topics with subscriptions for publish-subscribe, plus dead-letter queues for messages that cannot be processed. Event Grid is lightweight event routing that delivers events to subscribers with filtering and retries. Messaging carries work items reliably; eventing announces that something happened.
How it shows up
A communication or reliability need is described, and you must choose Service Bus or Event Grid and the right feature.
Common confusion
Confusing Service Bus with Event Grid; Service Bus is reliable messaging with queues and dead-letter handling, while Event Grid is lightweight event routing.
3.2 Develop and implement Azure FunctionsServerless APIs, triggers and bindings, and function app deployment.
What it means
This covers building serverless APIs with Azure Functions by implementing triggers and bindings, and configuring and deploying function apps.
What to understand
A trigger starts a function, such as an HTTP request, a queue message, a timer or an event. Bindings connect the function to data sources and sinks declaratively, as input or output, reducing boilerplate. Function apps host functions and are configured with settings and scaling. Functions glue services together serverlessly.
How it shows up
A serverless processing need is described, and you must choose the trigger and bindings or the function app configuration.
Common confusion
Confusing a trigger with a binding; a trigger starts the function, while bindings connect it to data as input or output.

High-yield concepts

The terms, rules and formulas this domain turns into questions most often.

  • Azure Service Bus

    Enterprise messaging with queues and topics.

    Remember
    Reliable, ordered, transactional messaging.
    Tested as
    Messaging questions.
    Trap
    Confusing it with Event Grid.

  • Queues vs. topics

    Point-to-point versus publish-subscribe.

    Remember
    Topics have multiple subscriptions.
    Tested as
    Messaging questions.
    Trap
    Using a queue where fan-out is needed.

  • Dead-letter queue

    Holds messages that cannot be processed.

    Remember
    It isolates poison messages for inspection.
    Tested as
    Reliability questions.
    Trap
    Losing failed messages silently.

  • Azure Event Grid

    Lightweight event routing to subscribers.

    Remember
    It announces events with filtering and retries.
    Tested as
    Eventing questions.
    Trap
    Confusing it with messaging.

  • Event filters

    Rules that route only matching events.

    Remember
    They reduce noise to subscribers.
    Tested as
    Eventing questions.
    Trap
    Delivering all events unfiltered.

  • Retries

    Automatic re-delivery of failed events.

    Remember
    They improve delivery reliability.
    Tested as
    Reliability questions.
    Trap
    Assuming single-delivery guarantees.

  • Function trigger

    What starts a function.

    Remember
    HTTP, queue, timer or event.
    Tested as
    Serverless questions.
    Trap
    Confusing it with a binding.

  • Function bindings

    Declarative connections to data as input or output.

    Remember
    They reduce integration boilerplate.
    Tested as
    Serverless questions.
    Trap
    Writing manual data access instead.

  • Function app

    The host for functions with settings and scaling.

    Remember
    It groups and configures functions.
    Tested as
    Deployment questions.
    Trap
    Misconfiguring the app scaling.

  • Serverless integration

    Using Functions to glue services together.

    Remember
    No servers to manage, event-driven.
    Tested as
    Integration questions.
    Trap
    Running always-on compute instead.

Side by side

Concepts candidates mix up, compared on the point that decides the answer.

Service Bus vs. Event Grid

ConceptMeaningKey differenceCommon trap
Service Bus Reliable messaging with queues and topics Work items, ordering, dead-letter Using it for lightweight event broadcast
Event Grid Lightweight event routing Announcements, filtering, retries Using it for reliable transactional work

Queue vs. topic in Service Bus

ConceptMeaningKey differenceCommon trap
Queue Point-to-point delivery One consumer group Using it where fan-out is needed
Topic with subscriptions Publish-subscribe Many subscribers Confusing with a queue

Trigger vs. binding in Functions

ConceptMeaningKey differenceCommon trap
Trigger Starts the function One per function Confusing with a binding
Binding Connects to data in or out Reduces boilerplate Writing manual data access

Exam traps

Each trap is an answer that looks right. Learn the clue that should change your mind.

  1. Event Grid for reliable work

    What candidates pick
    Use Event Grid to process a reliable, ordered, transactional work queue.
    Why it looks right
    Event Grid delivers events.
    Why it's wrong
    Service Bus provides reliable, ordered messaging with dead-letter queues; Event Grid is lightweight event routing.

    Clue to watch for The need is reliable, ordered processing with dead-lettering.

  2. Queue for fan-out

    What candidates pick
    Use a single Service Bus queue to deliver a message to several independent subscribers.
    Why it looks right
    Queues carry messages.
    Why it's wrong
    A queue delivers to one consumer group; topics with subscriptions fan out to many subscribers.

    Clue to watch for One message must reach multiple independent subscribers.

  3. Drop failed messages

    What candidates pick
    Let messages that repeatedly fail processing disappear.
    Why it looks right
    It avoids blocking the queue.
    Why it's wrong
    A dead-letter queue isolates poison messages for inspection rather than losing them.

    Clue to watch for Failed messages must be retained for investigation.

  4. Trigger is binding

    What candidates pick
    Use a binding to start a function on a new queue message.
    Why it looks right
    Both connect to data.
    Why it's wrong
    A trigger starts the function; bindings connect it to data as input or output.

    Clue to watch for The need is to start the function versus read or write data.

  5. Always-on for glue logic

    What candidates pick
    Run always-on compute to connect two services with small logic.
    Why it looks right
    It is a familiar model.
    Why it's wrong
    Azure Functions runs the glue logic serverlessly, triggered by events, without always-on compute.

    Clue to watch for The logic is short and event-driven.

Answer framework

For any Domain 3 scenario, use these steps to find the answer AI-200 expects.

  1. Decide whether the need is reliable messaging of work items or lightweight routing of events.
  2. For messaging, use Service Bus with queues for point-to-point or topics and subscriptions for fan-out, and dead-letter queues for failures.
  3. For eventing, use Event Grid with filters and retries to route events to subscribers efficiently.
  4. For serverless logic, choose the right Function trigger to start the function and bindings to connect it to data as input or output.
  5. Deploy and configure the function app with appropriate settings and scaling, and prefer serverless glue over always-on compute. Use bindings rather than SDK code for simple input and output.

Practice question

A back-end must process financial transactions reliably and in order, retry transient failures, and send messages that cannot be processed after several attempts to a separate location for later inspection. Which Azure service BEST meets these requirements?

Want more like this? The AI-200 practice test has questions for every domain.

Checklist

0 of 10

Tick each one you can do without notes. Your ticks stay in this browser.

Study strategy

Approach this domain as the integration layer and keep the two communication models distinct: Service Bus is reliable enterprise messaging for work items, with queues, topics and subscriptions, and dead-letter queues; Event Grid is lightweight event routing with filters and retries. Drill when each applies, since confusing them is a common trap.

Next, master Azure Functions: the difference between a trigger (what starts the function) and bindings (declarative data connections in and out), the common trigger types, and function app deployment. Functions are the serverless glue that connects the messaging, eventing and data services across the exam.

Build a small event-driven pipeline, a queue feeding a Function that writes to a data store, in a free subscription to make these concrete. Afterwards, move to Domain 4, securing, monitoring and troubleshooting Azure solutions. Then send a malformed message on purpose and watch it reach the dead-letter queue, so the retry and dead-letter behavior is clear.

FAQ

How much of AI-200 is Domain 3?

Connect to and consume Azure services is 20–25% of the exam, covering messaging, eventing and serverless integration.

What is the difference between Service Bus and Event Grid?

Service Bus is reliable enterprise messaging for work items, with queues, topics, subscriptions and dead-letter queues. Event Grid is lightweight event routing that announces events to subscribers with filtering and retries.

What is a dead-letter queue?

A sub-queue that holds messages which cannot be processed after retries, so poison messages are isolated for inspection rather than lost or blocking the main queue.

What is the difference between a Function trigger and a binding?

A trigger is what starts the function, such as an HTTP request or a queue message. Bindings declaratively connect the function to data sources and sinks as input or output.

When should I use topics instead of queues?

When one message must reach multiple independent subscribers. A queue delivers to a single consumer group, while a topic with subscriptions fans out to many.

Why use Azure Functions for integration?

They run short, event-driven logic serverlessly, triggered by events and connected to data through bindings, avoiding always-on compute to glue services together.

How does Domain 3 connect to the rest of the exam?

The messaging and events here drive the container scaling of Domain 1, move data in and out of the AI data services of Domain 2, and are secured and monitored in Domain 4.

Other Cloud Computing certifications

Where the same subject shows up in other exams. Cards marked "Same subject" open the matching domain.

Test your Domain 3 knowledge

You have covered Service Bus, Event Grid and Azure Functions. Test whether you can choose the right messaging or serverless approach under exam conditions.

Take the AI-200 practice test Check your AI-200 readiness

Keep going