AB-620 Study Guide

AB-620 measures whether you can do the day-to-day work of a developer or advanced builder in Microsoft Copilot Studio: designing and configuring agents with topics, agent flows and advanced responses, integrating them with enterprise knowledge, tools, other agents and Azure services, and testing, deploying and managing them through their lifecycle.

The exam has three domains. Integrating and extending agents carries the most weight at 40-45%, planning and configuring agent solutions carries 30-35%, and testing and managing agents carries 20-25%. The largest domain is about connecting agents to everything around them (enterprise knowledge, tools, other agents and Azure services), and the same integration choices appear as planning work in the first domain.

Plan and configure agent solutions — 30-35%

Planning: integration with enterprise systems, identity strategy, channels and deployment, responsible AI strategy, security and governance considerations, reusable agent components, and designing agents for internal or external audiences.

Agent flows: creating agent flows and human-in-the-loop agent flows, configuring actions and connectors, monitoring agent flows, input and output parameters, and error handling.

Topics: adding agent flows and tools to a topic, response formatting, advanced responses with custom prompts, custom knowledge sources, and API and Send HTTP requests, the generative answers node, adaptive cards, and managing variables.

Integrate and extend agents in Copilot Studio — 40-45%

Enterprise knowledge: Copilot connectors, Microsoft Power Platform connectors and Azure AI Search as knowledge sources.

Tools: configuring and monitoring computer use, MCP tools, a tool built on an existing custom connector, and REST APIs.

Multi-agent collaboration: designing multi-agent solutions and integrating a Foundry agent, an existing agent, a Fabric data agent, and agents that use the Agent2Agent (A2A) protocol.

Azure: generative answers that use Azure AI Search with Foundry, custom prompts that use the Foundry model catalog, and monitoring agents with Application Insights.

Test and manage agents — 20-25%

Evaluation: creating a test set, choosing an evaluation method, and reviewing test results.

Application lifecycle management (ALM): creating a solution, adding existing agents to it, creating and using environment variables, and implementing and extending Microsoft Power Platform Pipelines.

The full bullet-level list lives in the official study guide. Treat it as your checklist:

The Official AB-620 Study Guide

How This Maps to CertiAce Practice Modules

The CertiAce question bank is organized by the five learning paths of the official AB-620 course on Microsoft Learn, not by the three skills-measured domains, and it adds two case studies that combine topics across the domains. Several modules draw on more than one domain, so use the mapping below and the topics behind your missed questions to identify the skills-measured areas to revisit:

AB-620 has no prerequisites: the certification page lists none. It is also a stepping stone: earning Microsoft Certified: AI Agent Builder Associate (exam AB-620) satisfies the certification prerequisite for Microsoft Certified: Agentic AI Business Solutions Architect Expert (exam AB-100). That expert certification requires passing AB-100 plus at least one of thirteen associate certifications: Dynamics 365 Supply Chain Management Functional Consultant Associate (MB-330), Power Platform Developer Associate (AB-400), Dynamics 365 Customer Service Functional Consultant Associate (MB-230), Dynamics 365 Finance Functional Consultant Associate (MB-310), Dynamics 365: Finance and Operations Apps Developer Associate (MB-500), Dynamics 365 Business Central Functional Consultant Associate (MB-800), Dynamics 365 Business Central Developer Associate (MB-820), Machine Learning Operations Engineer Associate (AI-300), Azure AI Apps and Agents Developer Associate (AI-103), AI Agent Builder Associate (AB-620), Dynamics 365 Sales AI Consultant Associate (AB-210), Intelligent Applications Builder Associate (AB-410), or Dynamics 365 Contact Center AI Engineer Associate (AB-250).

If generative AI concepts are new to you, AI-901 first is a gentle on-ramp; if you already build agents in Copilot Studio, go straight to AB-620.

The audience profile describes a professional developer or advanced builder who already configures agents in Copilot Studio and knows the Power Platform around it. Check yourself against this table. Anything unfamiliar is where your preparation should start:

How long you need depends on where you start. Treat these estimates as planning guidance and adjust them to your starting knowledge and weekly study hours:

Step 1: Read the Official Study Guide

Skim the full skills-measured list once, and mark every bullet you could not explain to a colleague. That marked-up list is your personal syllabus: everything else in this plan exists to clear it. The page carries no "skills measured as of" date and no change log, so check it again shortly before your exam. Most questions cover generally available features, but commonly used preview features may also appear. Expect some older names in study material: the study guide's own resource list still says Azure AI Foundry next to the skills list's Microsoft Foundry.

The Official Study Guide

Step 2: Schedule Your Exam

Choose a realistic target date after reviewing the skills list and your available study time. A date on the calendar turns studying into a countdown, so pick one using the prep-time table above and plan backwards from it. The exam is available in thirteen languages. If the exam is not available in your native language, check the English as a Second Language accommodation and arrange any approved extra time before booking.

Exam accommodations

Certification and Exam Details Page

Step 3: Work Through the Official Learning Paths

Complete the five learning paths behind course AB-620T00-A on Microsoft Learn. They are the same five modules you will find in CertiAce practice, and each path title adds "(GitHub Copilot)". That suffix matters. Copilot Studio now builds agents on three harnesses, the runtimes that decide how an agent works: the standard harness for rule-based agents built from topics, the GitHub Copilot harness for reasoning-heavy, multistep work (a Copilot Studio runtime, not the GitHub Copilot service), and the Customer Conversation harness for customer service over voice and chat. The learning paths teach the GitHub Copilot harness, while the skills list names several features that exist only on the standard harness: topics, the generative answers node, agent flows, the prompt tool, computer use and Azure AI Search knowledge. On the GitHub Copilot harness, hooks replace topics, agentic skills replace the prompt tool, Azure AI Search knowledge is not supported and computer use is not yet available. Use the paths for what the two harnesses share (knowledge, tools, connected agents, evaluation, publishing and ALM), and learn the standard-harness features from the Copilot Studio documentation. Take notes on concepts you cannot explain in simple terms, and flag anything that needs hands-on practice. If you prefer instructor-led training, course AB-620T00-A runs over three days.

Official Learning Path Course Page

Harnesses in Copilot Studio

Step 4: Get Hands-On Practice

AB-620 rewards building. Many questions ask which node, tool, knowledge source, connection, setting or test method meets a requirement, and those choices stick once you have configured them yourself. A free Copilot Studio trial lets you create agents and test them in the test chat, which supports many of the core exercises; sign up with a work or school account, and extend the trial by 30 days when it expires if you need more time. The trial license can't publish agents, so practicing channels and deployment needs a Copilot Studio license that allows publishing. Build on the standard harness, because that is where topics, agent flows and the other skills-list features live. Make sure you personally:

• Build a topic that calls a tool, sends an HTTP request, shows an adaptive card, and stores values in variables
• Add a generative answers node with a custom knowledge source, and a prompt tool with your own instructions
• Create an agent flow with the When an agent calls the flow trigger, input and output parameters, a human-in-the-loop step and error handling, then call it from a topic and review its runs
• Add a Copilot connector, a Power Platform connector and Azure AI Search as knowledge, and compare how each one is set up and authenticated
• Add an MCP server, a custom connector and a REST API as tools, switch a tool between end user and maker-provided credentials, and configure computer use
• Add a child agent and a connected agent, and if you can, connect a Foundry agent, a Fabric data agent or an A2A agent
• Create a test set, run it with more than one evaluation method, and review the results
• Put an agent in a solution, add an environment variable, and move the solution to a second environment with Power Platform Pipelines

The Azure and Fabric objectives need access to those services as well: an Azure subscription for Azure AI Search, Microsoft Foundry and Application Insights, and Microsoft Fabric for a Fabric data agent. The official AB-620T00 lab exercises are free to follow and step by step, but they assume a tenant where you can create a Developer environment, and some tasks consume Copilot Credits. They build their agents on the GitHub Copilot harness, but their environment, solution, MCP, connector, connected-agent and evaluation exercises cover skills-list topics too. Learn from the documentation what you cannot try.

Get access to Copilot Studio

Official AB-620T00 Lab Exercises

Step 5: Benchmark Your Knowledge

Use CertiAce to benchmark your readiness module by module. The practice modules map to the exam domains as shown above, so use the module mapping and the topics behind your missed questions to identify the skills-measured areas to revisit. Microsoft also offers a free official practice assessment for AB-620 on AI Skills Navigator (sign in to launch it). Aim for consistent performance across every module, not one lucky high score; pay particular attention to questions that separate similar options (a Send HTTP request node versus a REST API tool, a Copilot connector versus a Power Platform connector as knowledge, a child agent versus a connected agent), and if a topic is unstable, go back to learning plus hands-on practice.

CertiAce AB-620 Exam Practice

Official Practice Assessment on AI Skills Navigator

Step 6: Take the Exam

The day before, review only your weak topics (no brand-new material). On exam day, read each question for what it is truly asking and eliminate wrong options first. Many AB-620 questions turn on one stated detail, so find it before you read the options: the orchestration mode the agent uses, whose credentials a tool runs with, whether something must be published first, and constraints such as least privilege, reuse across agents, or minimal development effort.

Worth knowing before you sit down:

• Microsoft Learn is available during the exam in a split screen. Use it to check details you genuinely need: the timer keeps running, and the exam is deliberately too long to look up every answer. Browsing is limited to Microsoft Learn itself (no Q&A, practice assessments, or profile), and personal notes and other websites are unavailable.
• You can take unscheduled breaks, but the clock keeps running and you cannot return to questions you already saw.
• Case studies lock when you leave them: finish each one before moving on, because you cannot revisit its questions afterwards.
• Labs may appear: AB-620's 120-minute allotment is the one Microsoft uses for role-based exams that may contain labs. The introduction screen tells you whether yours does; if it does, budget extra time for them.
• Watch the harness: features and settings differ between the standard and GitHub Copilot harnesses, so check which one a question describes before you answer.
• Try the exam sandbox beforehand so the question formats and interface hold no surprises: Exam Sandbox