Skip to main content
CertRampLearning

Microsoft · Free practice questions

AB-620 Practice Questions

15 original AB-620 practice questions across the three skill areas of the Microsoft study guide, weighted roughly like the real exam. Each one comes with the answer, a short explanation and the reason every other option is wrong.

Written for the study guide last updated 16 October 2026 (no "skills measured as of" date published) · Updated October 27, 2026 · Übungsfragen auf Deutsch

These are original practice questions written by CertRamp to the official Microsoft exam outline. They are not real exam questions and do not come from any exam. Try each question first, then open the answer to see the explanation for every option.

Domain 1

Plan and configure agent solutions

Plan identity strategyFoundationDomain: Plan and configure agent solutions

Question 1. An internal HR agent will be published only through the built-in Teams and Microsoft 365 channel. Employees must be signed in, and you want Microsoft Entra ID authentication set up without manual app registration. Which authentication option should you choose?

  1. Option A: No authentication
  2. Option B: Authenticate manually
  3. Option C: Authenticate with Microsoft
  4. Option D: An API key stored in a global variable
Show answer and explanation

Correct answer: C. Authenticate with Microsoft

Copilot Studio offers no authentication, Authenticate with Microsoft and Authenticate manually. Pick the simplest option that meets the channel and identity requirements; manual authentication is for other channels or custom identity providers.

  • A (incorrect)Anyone with the link could use the agent, and you would not know who the user is.
  • B (incorrect)Manual authentication is for channels and identity providers that need custom configuration; it is more work than required here.
  • C (correct)Correct. Microsoft documents this option for the Teams and Microsoft 365 channel; it sets up Entra ID authentication automatically.
  • D (incorrect)API keys authenticate calls to APIs; they do not sign users in to an agent.

Reference: Microsoft Learn: Configure user authentication in Copilot Studio (opens in a new tab)

Create a human-in-the-loop agent flowIntermediateDomain: Plan and configure agent solutions

Question 2. An agent flow processes supplier invoices. Invoices above a threshold must first be assessed by an AI model and then approved by a finance manager before payment continues. Which action fits best?

  1. Option A: Run a multistage approval with an AI stage followed by a manual stage
  2. Option B: A Send HTTP request node in a topic
  3. Option C: A generative answers node grounded on the invoice policy
  4. Option D: A scheduled refresh of the knowledge source
Show answer and explanation

Correct answer: A. Run a multistage approval with an AI stage followed by a manual stage

Human-in-the-loop flows pause for a person's decision. A multistage approval can route an AI decision to a human reviewer, and approvers can respond in Teams, Outlook or the Power Automate portal. Multistage and AI approvals are currently in preview.

  • A (correct)Correct. Multistage approvals in agent flows combine AI stages, manual stages and conditions, and the flow waits until the approval completes.
  • B (incorrect)This calls an API from a topic; it provides no approval process or waiting for a human decision.
  • C (incorrect)Generative answers produce responses from knowledge; they do not collect a manager's approval.
  • D (incorrect)Refreshing knowledge has nothing to do with approving an invoice.

Reference: Microsoft Learn: Multistage and AI approvals in agent flows (opens in a new tab)

Configure advanced agent responses with API and Send HTTP requestsIntermediateDomain: Plan and configure agent solutions

Question 3. In a topic you call an external order API with the Send HTTP request node. You want later nodes to use fields from the JSON response with IntelliSense in the Power Fx editor. What should you do?

  1. Option A: Set the body to Raw content
  2. Option B: Store the response as a string and parse it with a custom prompt
  3. Option C: Add an Authorization header
  4. Option D: Generate the response schema from sample JSON and save the response to a variable
Show answer and explanation

Correct answer: D. Generate the response schema from sample JSON and save the response to a variable

The Send HTTP request node supports GET, POST, PATCH, PUT and DELETE, custom headers and JSON or raw bodies. Providing a sample response turns the result into a structured variable instead of plain text.

  • A (incorrect)The body setting controls what you send, not how the response is typed.
  • B (incorrect)This adds an unnecessary model call and loses the typed structure.
  • C (incorrect)Headers may be needed for authentication, but they do not type the response.
  • D (correct)Correct. Get schema from sample JSON creates a typed Power Fx variable that later nodes can use.

Reference: Microsoft Learn: Make HTTP requests in Copilot Studio (opens in a new tab)

Configure adaptive cardsFoundationDomain: Plan and configure agent solutions

Question 4. A topic must collect a travel date, a destination and a cost centre in one step, with a single Submit button, and store each value for later nodes. What should you add?

  1. Option A: Three separate Message nodes
  2. Option B: An Adaptive Card node with three input fields and a submit action
  3. Option C: A generative answers node
  4. Option D: A global variable with a default value
Show answer and explanation

Correct answer: B. An Adaptive Card node with three input fields and a submit action

Use the Adaptive Card node for interactive cards that expect a submission. For cards that only display information, use Message or Question nodes; check the generated output variables with Edit Schema.

  • A (incorrect)Message nodes display information; they do not collect input.
  • B (correct)Correct. The Adaptive Card node shows an interactive card and creates output variables for its inputs, which are filled when the user submits.
  • C (incorrect)Generative answers respond from knowledge sources; they do not present a form.
  • D (incorrect)A variable stores values but does not ask the user for them.

Reference: Microsoft Learn: Ask with Adaptive Cards (opens in a new tab)

Manage variablesIntermediateDomain: Plan and configure agent solutions

Question 5. Several topics need the customer's account number once it has been captured in a sign-in topic. In addition, the API base URL must differ between development, test and production environments without editing the agent. Which combination fits?

  1. Option A: A global variable for the account number and a Power Platform environment variable for the API base URL
  2. Option B: Topic variables for both values
  3. Option C: A system variable for the account number
  4. Option D: Hard-coding the API base URL in every HTTP node
Show answer and explanation

Correct answer: A. A global variable for the account number and a Power Platform environment variable for the API base URL

Copilot Studio has topic, global, system and environment variables. Environment variables are set in Power Platform and travel with solutions, so each environment can hold its own value.

  • A (correct)Correct. Global variables are available in all topics; environment variables hold per-environment values and are read-only in Copilot Studio.
  • B (incorrect)Topic variables are only available in the topic that creates them.
  • C (incorrect)System variables are built in and hold context about the conversation or user; they are not meant for storing your own values such as an account number.
  • D (incorrect)This requires editing the agent after each deployment, which the requirement rules out.

Reference: Microsoft Learn: Work with variables in Copilot Studio (opens in a new tab)

Domain 2

Integrate and extend agents in Copilot Studio

Configure MCP toolsIntermediateDomain: Integrate and extend agents in Copilot Studio

Question 6. Your company runs an MCP server that exposes tools and prompt templates for its ticketing system. You connect it to a Copilot Studio agent. Which statement is correct?

  1. Option A: Every tool must be re-imported manually whenever the server adds a tool
  2. Option B: MCP works with classic orchestration only
  3. Option C: The agent can use all three MCP primitives, including the prompt templates
  4. Option D: The agent can use the server's tools and resources, and generative orchestration must be turned on
Show answer and explanation

Correct answer: D. The agent can use the server's tools and resources, and generative orchestration must be turned on

MCP defines resources, tools and prompts. Copilot Studio uses tools and resources; the server supplies names, descriptions, inputs and outputs, so the orchestrator can decide when to call each tool.

  • A (incorrect)Tools published by the server become available automatically, and changes on the server are reflected dynamically.
  • B (incorrect)Generative orchestration must be turned on to use MCP.
  • C (incorrect)Copilot Studio currently supports MCP tools and resources, not prompts.
  • D (correct)Correct. Copilot Studio supports MCP tools and resources and needs generative orchestration to use them.

Reference: Microsoft Learn: Extend your agent with Model Context Protocol (opens in a new tab)

Add REST APIs to an agentIntermediateDomain: Integrate and extend agents in Copilot Studio

Question 7. You want to add a partner's REST API as a tool to an agent. The partner provides an OpenAPI 3.0 specification. What happens when you upload it?

  1. Option A: The upload is rejected; you must build a custom connector first
  2. Option B: Copilot Studio translates it to an OpenAPI v2 specification automatically
  3. Option C: Copilot Studio creates an MCP server from it
  4. Option D: Only APIs without authentication can be added
Show answer and explanation

Correct answer: B. Copilot Studio translates it to an OpenAPI v2 specification automatically

To add a REST API tool you need the OpenAPI specification and the authentication details. The v2 requirement comes from how Power Platform processes API specifications.

  • A (incorrect)A custom connector is not required for a REST API tool, and v3 specifications are accepted.
  • B (correct)Correct. REST API tools are created from OpenAPI v2; a v3 specification is translated automatically.
  • C (incorrect)Uploading a specification creates a REST API tool, not an MCP server.
  • D (incorrect)REST API tools support no authentication, API key and OAuth 2.0.

Reference: Microsoft Learn: Add REST API tools to agents (opens in a new tab)

Configure and monitor computer use for an agentIntermediateDomain: Integrate and extend agents in Copilot Studio

Question 8. An agent must enter claims into a legacy Windows desktop application that has no API or connector. What should you configure?

  1. Option A: A Send HTTP request node pointing at the application's executable
  2. Option B: Azure AI Search as a knowledge source
  3. Option C: The computer use tool on a Windows machine set up for it, with generative orchestration turned on
  4. Option D: A Fabric data agent
Show answer and explanation

Correct answer: C. The computer use tool on a Windows machine set up for it, with generative orchestration turned on

Computer use runs on a Windows machine you set up and select in the tool. During testing you see a step-by-step log and a preview of the machine; in conversations it shares reasoning messages and screenshots.

  • A (incorrect)HTTP requests call web APIs; the application has none.
  • B (incorrect)Knowledge sources ground answers; they cannot operate an application.
  • C (correct)Correct. Computer use operates websites and desktop apps with a virtual mouse and keyboard when no API exists.
  • D (incorrect)Fabric data agents answer questions over Fabric data; they do not operate desktop applications.

Reference: Microsoft Learn: Computer use in Copilot Studio (opens in a new tab)

Create a multi-agent solution by using A2A protocolIntermediateDomain: Integrate and extend agents in Copilot Studio

Question 10. A logistics partner exposes a shipment-tracking agent built on another platform through the Agent2Agent (A2A) protocol. Your Copilot Studio agent should delegate tracking questions to it. What should you do?

  1. Option A: Rebuild the partner agent as a child agent in your Copilot Studio agent
  2. Option B: Copy the partner's answers into a knowledge file
  3. Option C: Ask the partner to publish its agent to your Teams tenant
  4. Option D: Add it as a connected agent that is available over the A2A protocol
Show answer and explanation

Correct answer: D. Add it as a connected agent that is available over the A2A protocol

Copilot Studio can add child agents, other Copilot Studio agents in the environment, A2A agents, and (in preview) Microsoft Foundry agents, Fabric data agents and Microsoft 365 Agents SDK agents.

  • A (incorrect)Rebuilding duplicates the partner's work; it can be connected directly.
  • B (incorrect)Static files cannot answer live tracking questions.
  • C (incorrect)Publishing to Teams makes it available to people, not as a connected agent of yours.
  • D (correct)Correct. Copilot Studio can add other agents that are available over A2A to a multi-agent solution.

Reference: Microsoft Learn: Add other agents in Copilot Studio (opens in a new tab)

Integrate a Fabric data agentAdvancedDomain: Integrate and extend agents in Copilot Studio

Question 11. You connect a Fabric data agent to your main Copilot Studio agent so that users can ask sales questions over a lakehouse. During design, which limitation must you plan for?

  1. Option A: The main agent must use classic topics only
  2. Option B: Fabric data agents cannot be the target of a topic's Redirect node and do not currently work when the main agent is deployed to Microsoft Copilot
  3. Option C: The Fabric data agent must be rebuilt as a child agent
  4. Option D: Fabric data agents can only answer questions in English
Show answer and explanation

Correct answer: B. Fabric data agents cannot be the target of a topic's Redirect node and do not currently work when the main agent is deployed to Microsoft Copilot

Connecting Foundry, Fabric and Microsoft 365 Agents SDK agents is in public preview. Fabric data agents also cannot be referenced in instructions, so plan channels and routing around these limits.

  • A (incorrect)There is no such requirement in the documentation for connected agents.
  • B (correct)Correct. Microsoft documents both limitations for Fabric data agents connected in Copilot Studio.
  • C (incorrect)Fabric data agents are connected as they are; they do not need rebuilding.
  • D (incorrect)The documentation does not list such a restriction.

Reference: Microsoft Learn: Add other agents in Copilot Studio (opens in a new tab)

Monitor agents by using Application InsightsFoundationDomain: Integrate and extend agents in Copilot Studio

Question 12. Your operations team wants Copilot Studio conversation telemetry in the same Azure Application Insights resource they already use for other services. What do you do?

  1. Option A: Export the transcript table from Dataverse to Application Insights every night
  2. Option B: Install the Azure Monitor Agent on the users' devices
  3. Option C: Enter the Application Insights connection string in the agent's advanced settings
  4. Option D: Add an MCP server for Application Insights
Show answer and explanation

Correct answer: C. Enter the Application Insights connection string in the agent's advanced settings

In Settings > Advanced you add the connection string and can optionally log message details and node execution events. Test-canvas conversations are flagged with designMode so you can filter them out.

  • A (incorrect)A manual export is not needed; Copilot Studio can send telemetry directly.
  • B (incorrect)Agents on devices collect device logs, not Copilot Studio conversations.
  • C (correct)Correct. The agent then logs telemetry, which you query in the Logs section, for example in the customEvents table.
  • D (incorrect)MCP extends what the agent can do; it is not how telemetry is configured.

Reference: Microsoft Learn: Capture telemetry with Application Insights (opens in a new tab)

Domain 3

Test and manage agents

Create a test setFoundationDomain: Test and manage agents

Question 13. You have 80 real user questions with the answers the business expects, kept in a spreadsheet. You want to evaluate your agent against them. What is the quickest way to create the test set?

  1. Option A: Save them as a CSV file with the headings Question and Expected response and import it
  2. Option B: Type each question into the test chat and wait for errors
  3. Option C: Generate questions from the agent's topics
  4. Option D: Upload the spreadsheet as a knowledge source
Show answer and explanation

Correct answer: A. Save them as a CSV file with the headings Question and Expected response and import it

Test sets can be generated from knowledge or topics, built from the latest test chat, written manually or imported. Expected responses are optional for import but needed for comparison-based test methods.

  • A (correct)Correct. Imported test sets use a CSV or text file with these headings, up to 100 questions.
  • B (incorrect)Manual chatting does not produce a repeatable test set with expected responses.
  • C (incorrect)Generated questions would not be your 80 real questions and expected answers.
  • D (incorrect)That would add the content to the agent's knowledge rather than test it.

Reference: Microsoft Learn: Create a test set (opens in a new tab)

Choose an evaluation methodIntermediateDomain: Test and manage agents

Question 14. Your agent's answers may be worded differently from the expected responses, as long as they say the same thing. Which test method should you choose?

  1. Option A: Exact match
  2. Option B: Keyword match
  3. Option C: Tool use
  4. Option D: Compare meaning
Show answer and explanation

Correct answer: D. Compare meaning

Copilot Studio offers general quality, content safety, compare meaning, tool use, keyword match, text similarity, exact match and custom methods. Methods that compare against an expected value return Invalid when that value is missing.

  • A (incorrect)Exact match fails any answer that is not word-for-word identical.
  • B (incorrect)Keyword match checks for specific words or phrases, not overall meaning.
  • C (incorrect)Tool use checks whether the expected tools or topics were used, not the answer content.
  • D (correct)Correct. Compare meaning evaluates how well the meaning of the answer matches the expected response.

Reference: Microsoft Learn: Agent evaluation in Copilot Studio (opens in a new tab)

Implement and extend Microsoft Power Platform PipelinesAdvancedDomain: Test and manage agents

Question 15. You use Power Platform pipelines to move a solution containing your agent from development to test and production. Which statement is correct?

  1. Option A: Pipelines deploy the solution together with its Dataverse table data
  2. Option B: The development environment must be the pipelines host
  3. Option C: Test and production targets must be managed environments, and Microsoft recommends deploying managed solutions to them
  4. Option D: Pipelines cannot be customised
Show answer and explanation

Correct answer: C. Test and production targets must be managed environments, and Microsoft recommends deploying managed solutions to them

Pipelines deploy solutions plus configuration such as connection references and environment variables. Keep agents in solutions from the start so that ALM, environment variables and pipelines work as intended.

  • A (incorrect)Pipelines and solutions do not contain data stored in Dataverse tables.
  • B (incorrect)Using the same environment for development and the host is not supported.
  • C (correct)Correct. All non-developer environments used in pipelines must be managed environments, and managed solutions are recommended for non-development targets.
  • D (incorrect)Pipelines can be extended with extension hooks, Power Automate and integration with other CI/CD tools.

Reference: Microsoft Learn: Overview of pipelines in Power Platform (opens in a new tab)

Exam information last verified: October 27, 2026

Exam details can change. Always confirm them with the official provider before you book.

Want more practice?

15 questions are a sample. The exam is much longer.

The full CertRamp AB-620 practice exams have 6 tests that grow with you, from a diagnostic start to an exam-level Test 4 and a challenge set beyond the exam, with an explanation for every option.

Who writes the exams

Written by a practitioner, not a question mill

Joshua Ravnjak, founder of CertRamp Learning

Joshua Ravnjak

Senior Consultant, digital transformation and AI

Founder, CertRamp Learning

921 learners and 61 reviews on Udemy (as of November 3, 2026)

I have more than 12 years of professional experience and work as a senior consultant in an IT management consultancy, focusing on digital transformation, AI and requirements engineering in the insurance industry. I have earned many certifications myself, so I know what exam preparation feels like. That is why every CertRamp practice exam follows the same progression: an easy start to see where you stand, then real exam level, then beyond it. You walk into the exam with a buffer and a good feeling.

Selected certifications

  • IREB® CPRE Foundation Level
  • ITIL® 4 Foundation
  • Professional Scrum Master I (PSM I)
  • Professional Scrum Product Owner I (PSPO I)
  • Certified SAFe® 6 Scrum Master
  • Certified AI-Empowered SAFe® 6 Product Owner/Product Manager

Plus further certifications from Microsoft and Anthropic (Claude).

ITIL® is a registered trademark of PeopleCert. IREB® is a registered trademark of IREB e.V. SAFe® is a registered trademark of Scaled Agile, Inc. PSM and PSPO are trademarks of Scrum.org.

Also in print

Cover: CPRE Foundation Level Prüfungsvorbereitung 2026 by Joshua Ravnjak

CPRE Foundation Level Prüfungsvorbereitung 2026

6 Übungsprüfungen mit 270 Fragen und ausführlichen Erklärungen

Paperback, German, October 2026

My CPRE Foundation Level exam-prep book (in German) follows the same progressive approach.

View on Amazon (opens in a new tab)