AB-410 Business Application Logic And Automation Practice Test 4 — 15 Questions

Question 1

You are creating a Power Automate cloud flow that must call a Microsoft Copilot Studio agent to retrieve a customer's order status. The agent exposes a topic named 'GetOrderStatus' and expects a JSON payload with an 'orderId' property. You need to invoke the agent from the flow. What should you use?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Use the 'Ask a Copilot' action and provide the agent name and a message containing the orderId.. The 'Ask a Copilot' action is the native way to invoke a Copilot Studio agent from a Power Automate cloud flow. It allows you to pass a message, such as the orderId, and receive the agent's response. This action abstracts the underlying API calls and authentication, making it the most efficient and correct choice for integrating agent capabilities into your flow.

Answer

Create a custom connector that wraps the agent's topic and then call it from the flow.

Answer

Use the 'Send an HTTP request to Copilot Studio' action and provide the agent's endpoint URL and JSON payload.

Answer

Add the 'Run a flow' action and specify the agent's topic name and orderId.

Question 2

You are building a Power Automate cloud flow that must invoke a Microsoft Copilot Studio agent from within a business process. The agent has been published and you need the flow to receive the agent's structured response so that downstream actions can branch on the returned intent. Which action should you add to the flow to call the agent and get its response?

Accepted Answer

Add the 'Send a message to the agent' action from the Microsoft Copilot Studio connector and consume the returned response fields.. Invoking a published Microsoft Copilot Studio agent from a cloud flow is done with the Copilot Studio connector's message action, which returns the agent's reply and any structured output. That returned data is what downstream conditions consume, so the flow can branch on intent. Connectors that call a raw language model or that reverse the call direction cannot surface the agent's authored logic or structured response.

Answer

Add the 'Create a completion' action from the Azure OpenAI connector and pass the agent's prompt.

Answer

Add the 'Run a flow from Copilot' action and map the agent's output fields to dynamic content.

Answer

Add the 'Execute agent and wait' action from the Microsoft Copilot Studio connector and use the returned response fields.

Question 3

You are designing an approval process in Power Automate for expense reports. The process must (1) notify the approver in Microsoft Teams, (2) allow the approver to approve or reject directly from the notification, and (3) capture the approver's comments for auditing. Which TWO actions or configurations should you use? (Choose two.)

Accepted Answer

Configure the approval action's 'Details' field to include the expense report link and amount so the approver has context.. The approval requirement is met by the approval action, which both notifies the approver through the Approvals integration and returns a structured outcome with comments. Populating the Details field ensures the approver has enough context to decide from the notification. Manual card posting or email-with-options approaches lack the structured approval object and comment capture that auditing depends on.

Answer

Enable 'Secure inputs' on the approval action to protect approver comments.

Answer

Add a 'Post message in a chat or channel' action to send the approval card manually to the approver.

Answer

Use the 'Send an email with options' action and parse the reply to determine the decision.

Question 4

You are building a flow that must write a generated document to a SharePoint document library and then send a link to the requester. The document must be stored with a filename that includes the requester's name and the current date, and the link must open the file directly in the browser. Which combination of actions should you use?

Accepted Answer

Use 'Create file' in SharePoint with a filename built from dynamic content, then send the 'Link to item' returned by the action.. Storing a generated document in a SharePoint library and sharing it is done with the SharePoint 'Create file' action, which accepts a composed filename and returns a ready-to-use link. Building the filename from dynamic content and the utcNow() expression meets the naming requirement, and forwarding the returned link lets the requester open the file in the browser without manual URL construction.

Answer

Use 'Add attachment' in Outlook with the generated document, then send the email with the attachment.

Answer

Use 'Create file' in SharePoint with a fixed filename, then send the library's URL concatenated with the filename.

Answer

Use 'Upload file' via the OneDrive connector, then send the OneDrive sharing link.

Question 5

A Power Automate flow fails when calling a custom AI Builder prompt. The error indicates a token limit exceed. How should you resolve this?

Accepted Answer

Chunk the input data before sending to the prompt. A token limit exceeded error means the prompt input (plus context) exceeds the model's maximum token window, so the correct fix is to chunk or truncate the input data before sending it to the AI Builder prompt. This keeps each request within the model's token budget.

Answer

Enable 'Retry' policy with exponential backoff

Answer

Increase the AI Builder model capacity in the Admin Center

Answer

Change the AI model to GPT-4o

Question 6

You need to automate document extraction using AI Builder. The documents are varied invoices. Which model type should you configure?

Accepted Answer

Document Processing. Document Processing (formerly Form Processing) models in AI Builder are purpose-built to extract key-value pairs and tables from structured and semi-structured documents such as invoices, receipts, and purchase orders. They handle layout variation across vendors, which is exactly the 'varied invoices' scenario. Prebuilt invoice models or custom document processing models are the right choice here.

Answer

Object Detection

Answer

Entity Extraction

Answer

Prediction model

Question 7

You are building a Power Automate flow that needs to trigger whenever a row is added to a Dataverse table. Which trigger should you select to ensure the flow runs in near real-time?

Accepted Answer

When a row is added, modified or deleted. The Dataverse trigger 'When a row is added, modified or deleted' is a webhook-based trigger that fires in near real-time when a table row changes, making it the correct choice for event-driven flows. It supports filtering by scope (business unit, user, organization) and by specific columns to reduce unnecessary runs.

Answer

When a row is selected

Answer

List rows

Answer

Recurrence

Question 8

You are designing a Power Automate flow that summarizes incoming emails using AI Hub. Which action should you use to pass the email body to a custom prompt?

Accepted Answer

Create text with GPT using a prompt. The 'Create text with GPT using a prompt' action in AI Hub (formerly AI Builder) is the generative action that accepts a custom prompt and input text — such as an email body — and returns GPT-generated output. It is the only listed action that performs free-form text generation, which is exactly what summarization requires. The other actions perform classification, extraction, or prediction tasks, not generative summarization.

Answer

Analyze sentiment

Answer

Extract information from documents

Answer

Predict outcomes

Question 9

You are using AI Hub to categorize customer support emails. You need to ensure the model processes only emails with high sentiment scores. Where do you configure this logic?

Accepted Answer

Within a Condition control in Power Automate. AI Hub provides prebuilt AI models, but the logic that decides which emails to process based on sentiment scores belongs in the orchestration layer. In Power Automate, a Condition control can evaluate the sentiment score returned by the AI model and branch accordingly, ensuring only high-sentiment emails proceed. This keeps the AI model generic and the business logic external and maintainable.

Answer

Inside the AI Hub model settings

Answer

In the Dataverse business rule

Answer

In the Copilot Studio agent manifest

Question 10

A Power Automate flow is failing because an AI-powered extraction step occasionally returns null values. Which expression should you use to provide a default value of 'Unknown'?

Accepted Answer

coalesce(outputs('AI_Step')?['result'], 'Unknown'). The coalesce() expression returns the first non-null argument, so coalesce(outputs('AI_Step')?['result'], 'Unknown') yields the AI result when present and 'Unknown' when it is null. This is the idiomatic Power Automate way to supply a fallback for null values from an action output.

Answer

if(empty(outputs('AI_Step')), 'Unknown', outputs('AI_Step'))

Answer

default(outputs('AI_Step'), 'Unknown')

Answer

replace(null, 'Unknown')

Question 11

You want to allow your Copilot to look up internal company policies stored in a SharePoint site. What should you add to the Copilot?

Accepted Answer

Generative answers. Generative answers is the Copilot Studio feature that lets a copilot retrieve and summarize information from external knowledge sources such as SharePoint sites, public websites, or uploaded documents. By enabling generative answers and adding the SharePoint site as a knowledge source, the copilot can answer policy questions using internal content without manually authoring topics. This is the correct mechanism for grounding responses in enterprise data.

Answer

A new trigger

Answer

A topic

Answer

An entity

Answer

A Power Automate flow

Question 12

You are creating a custom prompt in AI Hub. You want the model to act as a technical support agent. Where do you define this persona?

Accepted Answer

The System message field. The System message field in AI Hub (and similar prompt-building tools) is where you define the model's persona, role, and behavioral instructions. Setting 'You are a technical support agent' in the system message shapes all subsequent responses. Input variables, output format, and trigger settings control data flow, not persona.

Answer

The trigger settings

Answer

The output format settings

Answer

The input variables section

Question 13

You need to automate a process that extracts data from PDF invoices. Some invoices are multi-page. Which model type should you use?

Accepted Answer

Document processing. Document processing (also called document intelligence) models are purpose-built to extract structured data such as key-value pairs, tables, and text from PDFs and scanned documents, including multi-page invoices. Azure AI Document Intelligence supports prebuilt invoice models that handle multi-page layouts and return structured JSON fields like invoice ID, vendor, and totals. Sentiment analysis, text classification, and object detection do not perform OCR-based field extraction from PDFs.

Answer

Sentiment analysis

Answer

Text classification

Answer

Object detection

Question 14

What is the primary function of an AI Builder 'Prebuilt' model?

Accepted Answer

To provide out-of-the-box AI functionality. AI Builder prebuilt models provide out-of-the-box AI functionality — such as invoice processing, receipt parsing, business card reading, sentiment analysis, and category classification — that you can use immediately without training on your own data. They are trained by Microsoft on large datasets and exposed as ready-to-consume components in Power Apps and Power Automate.

Answer

To store training datasets

Answer

To customize the model's architecture

Answer

To train on your own custom data

Question 15

When designing a Copilot, you want to ensure the agent only responds using specific company documents. How do you ensure this?

Accepted Answer

Configure Generative answers with specific data sources. Configuring Generative answers with specific data sources (such as SharePoint sites, websites, or Dataverse) grounds the Copilot's responses in those documents and restricts it to that knowledge. This is the supported mechanism for scoping an agent to company content. It ensures answers are retrieved from the designated sources rather than the model's general knowledge.

Answer

Use a custom entity

Answer

Add a filter in the prompt

Answer

Delete all default topics