AB-730 Domains Explained With Weightings

The short version

AB-730 has four domains. Fundamentals carries the most weight (25-30%), and the other three share the rest at 20-25% each. The skills below come from Microsoft’s study guide, “Skills measured as of October 20, 2026”. The traps and practice ideas are our own advice, not Microsoft’s list.

DomainWeighting
Understand generative AI fundamentals25-30%
Manage prompts and chats by using Microsoft 365 Copilot20-25%
Manage business content and collaboration by using Microsoft 365 Copilot20-25%
Drive business outcomes by using agents20-25%

1. Understand generative AI fundamentals (25-30%)

What it covers. Microsoft splits it in two. First, Copilot capabilities across Microsoft 365: how Copilot keeps your organization’s information private and secure, how context (Work IQ, web data, the app you are in) changes responses, when to use Copilot Chat, agents or Copilot Cowork, and how features differ between experiences. Second, responsible AI and data protection: inaccuracies, prompt injection, over-reliance, verification steps such as citation checks and human review, risks to sensitive data, and how data protection restricts what a prompt can return.

In practice. Imagine a colleague asks why Copilot did not show a document. Often the answer is permissions: Microsoft says Copilot only accesses content you are already authorized to see.

Typical traps. Mixing up what Copilot Chat can ground on (mostly web and what you provide) with the fuller licensed experience that uses Work IQ. Treating a fluent answer as a correct one. Forgetting that the right check depends on the task: a citation check for a research claim, human review for a message to a customer.

Practice. Write down, for five tasks you do, which Copilot experience fits and what you would verify.

2. Manage prompts and chats (20-25%)

What it covers. Creating effective prompts, including model selection and picking the right resources to reference. Saving, scheduling and sharing prompts. Refining a prompt from the response. How Copilot uses memory and instructions. Then chats: finding, renaming and deleting them, and organizing and sharing information with Notebooks.

In practice. You run the same weekly summary, so you save the prompt and schedule it instead of retyping it. You collect files, chats and notes for a project in a Notebook.

Typical traps. Questions often hinge on one detail, such as a quick answer versus a deeper-reasoning request. In Copilot Chat’s model selector, Microsoft describes Quick response as tuned for routine questions and Think deeper as taking time to plan, gather context and check its work. Another trap is mixing up what a single prompt does with what Copilot’s memory and instructions do; the skills list names both, so learn them as separate things.

Practice. Take one real task, write three versions of the prompt, and compare outputs. Then save one and schedule it.

3. Manage business content and collaboration (20-25%)

What it covers. Two halves. Content: connecting sources beyond Microsoft 365 with connectors, generating or analyzing content from business context including Work IQ, producing images, visuals and charts, working across Word, Excel and PowerPoint, deriving insights across Work IQ, and judging whether the output meets the business goal. Collaboration: preparing for a meeting, using the Facilitator agent, real-time catch-up, Intelligent recap, and Copilot Pages.

In practice. You join a call late and ask Copilot for a catch-up. Afterward you open the recap. During a workshop, you use Facilitator for notes and agenda.

Typical traps. Facilitator versus Copilot in Teams. Microsoft says your prompts to Copilot in Teams are private to you, while a prompt to Facilitator is visible to everyone in the meeting, like a message in the chat. Another trap is accepting a draft without checking it against the goal: the skills list says to interpret and validate AI content.

Practice. Run one real meeting with each feature on, then compare what each produced.

4. Drive business outcomes by using agents (20-25%)

What it covers. Prebuilt Microsoft 365 Copilot agents: spotting scenarios where one improves a workflow, choosing between a prebuilt and a custom agent, using an agent from different Copilot experiences, and filtering chats by agent. Then Microsoft Copilot Cowork: how it carries out multi-step tasks, how it uses skills, security considerations and known limits, prebuilt versus custom skills, delegating work, reviewing its plan, progress and results, and managing scheduled tasks or prompts.

In practice. Microsoft’s Cowork FAQ describes Copilot Chat as helping within a single session, while Cowork takes actions across apps and files. It asks for your approval before sensitive actions such as sending an email or posting in Teams. It can ask you a question with suggested choices, and you can type your own answer or skip it.

Typical traps. Choosing Cowork for a quick one-off draft that Chat handles. Forgetting known limits that Microsoft lists, such as Cowork not deleting files in OneDrive or SharePoint and not reading encrypted files. This domain is new in the October 20 outline, so older material may be missing.

Practice. Write a list of three tasks that are quick (Chat), three that are research-heavy, and three that are multi-step (Cowork), and say why.

Next step

Test yourself with the sample questions. Our AB-730 practice tests give every domain a share of each set, with an explanation for each answer.

Frequently asked questions

Which domain should I start with?

Fundamentals. It has the biggest weight, and the rest builds on its ideas about context, permissions and verification.

Do I need to build agents?

Not under the October 20, 2026 outline. Microsoft lists creating and managing agents as removed, but you still need to choose between a prebuilt and a custom agent.