The short version
AB-100 has three domains: Plan, Design, and Deploy. Until November 16, 2026 they weigh 25-30%, 25-30%, and 40-45%. From that date, when the English exam is updated, they weigh 35-40%, 30-35%, and 20-25%. Plan is strategy and economics, Design is solution architecture, and Deploy is everything that keeps a solution safe and running. The topics below come from the study guide titled “Skills measured as of October 14, 2026”; Microsoft has since published a guide for the November 16 outline, and the weightings in the headings show both versions. The “traps” and examples are our own explanation, not Microsoft statements.
Domain 1: Plan AI-powered business solutions (25-30%, then 35-40%)
Microsoft groups this domain into three areas.
Analyze requirements. You judge where agents help with task automation, data analytics, and decisions. You review grounding data for accuracy, relevance, timeliness, cleanliness, and availability, and you organize data so other AI systems can use it.
Design the overall AI strategy. This is the largest part. It includes the AI adoption process from the Cloud Adoption Framework for Azure, multi-agent designs across Microsoft 365 Copilot, Copilot Studio, and Microsoft Foundry, use cases for prebuilt agents, rules and constraints for AI components, prompt library guidelines, small language models, and the elements of an AI Center of Excellence.
Evaluate costs and benefits. ROI criteria including total cost of ownership, an ROI analysis, the decision to build, buy, or extend, and a model router that sends each request to the most suitable model.
Typical traps. Reaching for an agent when plain automation is enough. Ignoring data quality and then blaming the model. Choosing a custom model when prompting or a prebuilt option meets the need.
What to practice. Write a one-paragraph recommendation for a case, naming the deciding constraint (cost, data location, existing licenses, skills). Practice explaining why you did not choose the other two options.
Domain 2: Design AI-powered business solutions (25-30%, then 30-35%)
Design AI and agents. The longest list. It covers Copilot in Dynamics 365 apps for customer experience and service (business terms, customizations, connectors for Sales), agents for Contact Center channels, task, autonomous, and prompt-and-response agents, Foundry Tools for a requirement, code-first generative pages, Copilot Studio topics with fallback, data processing and grounding, AI components in a canvas app, and the Power Platform Well-Architected Framework. You also decide between standard natural language processing, conversational language understanding, and generative orchestration in Copilot Studio, and design agent flows and prompt actions.
Design extensibility. Custom models in Microsoft Foundry, agents in Microsoft 365 Copilot, extensibility in Copilot Studio including MCP, computer use for apps and websites, agent behaviors such as reasoning and voice mode, and agents in Teams and SharePoint.
Orchestrate prebuilt agents and apps. AI features in Dynamics 365 finance and supply chain and in customer experience and service, Microsoft 365 agents, Copilot for Sales and Service configuration, Power Platform AI features, and knowledge sources for finance and operations chat and in-app help.
Typical traps. Mixing up look-alike features, such as a knowledge source (content the agent reads) versus a tool or connector (something the agent does). Picking a preview-only approach when the scenario demands generally available features. Forgetting that the scenario already owns a Dynamics 365 license.
What to practice. For each agent type, say in one sentence when it starts, who it acts for, and what controls it. Click through Copilot Studio topics, knowledge, tools, and settings once.
Domain 3: Deploy AI-powered business solutions (40-45%, then 20-25%)
Analyze, monitor, and tune. Tools and process for monitoring agents, analyzing backlog and user feedback, AI-based tools for finding issues, performance metrics, and telemetry for performance and model tuning.
Manage testing. Test process and metrics for agents, validation criteria for custom models, prompt best practices, end-to-end scenarios across several Dynamics 365 apps, and using Copilot to build test cases.
Design the ALM process. Separate ALM designs for data used in AI, Copilot Studio agents with connectors and actions, Microsoft Foundry Agent Service, custom models, and AI in Dynamics 365 finance and supply chain and in customer experience and service.
Design responsible AI, security, and governance. Security for agents and models, governance, vulnerabilities and mitigations including prompt manipulation, responsible AI review, data residency and movement, access controls on grounding data and tuning, and audit trails for changes.
Typical traps. Reading a symptom (poor answers) and fixing the wrong layer (instructions instead of knowledge sources). Treating security as one control instead of layers: identity, data access, content filtering, and audit. Overlooking that environments, solutions, and pipelines carry configuration between stages.
What to practice. Draw a deployment path from development to production for one agent and mark where testing, approval, and monitoring happen. List what you would log to prove who changed a model.
Putting the weightings to use
Until November 16, 2026, Deploy carries about one and a half times the weight of either other domain (40-45% against 25-30%). After the update the balance flips: Plan is the biggest domain (35-40%) and Deploy the smallest (20-25%). Pick the split for your exam date, then adjust toward your weak area.
More guides for AB-100
- AB-100 Exam Format, Scoring and Retakes
- Is the AB-100 Exam Hard?
- AB-100 Prerequisites and Skills Check
- AB-100 Renewal and Retakes Explained
- Is AB-100 Worth It? An Honest Look
- AB-100 Cheat Sheet: Key Facts to Review
- AB-100 vs AB-620: Which Exam First?
Next step
Try the sample questions to see how these domains show up in scenarios, or build your schedule with the study plan. Our AB-100 practice tests keep the same three-domain split.
Frequently asked questions
Which domain is the biggest?
Until November 16, 2026, Deploy (40-45%). After the update, Plan (35-40%).
Do I need to know product screens?
You need to know what each feature does and where decisions are made. Memorizing every menu is not the aim.
Are preview features tested?
Microsoft says most questions cover generally available features, but preview features may appear if they are commonly used.