AB-730 Cheat Sheet 2026: Quick Reference
Every domain condensed to a table, plus the Agent Store vs. Copilot Studio decision and prompt reference syntax the exam tests directly.
Exam Snapshot
We built this from Microsoft's published AB-730 skills outline, not a generic template — every reference link below points to the current Microsoft Learn page, verified before this went live.
1. Understand Generative AI Fundamentals (25-30%)
| Concept | Feature / Area | Quick note |
|---|---|---|
| Data privacy | Copilot & tenant data | Copilot keeps org data private and secure by default |
| Context sources | Work files, web data, app context | These directly shape and ground Copilot's response |
| Chat vs. agent | Copilot experience types | Chat = conversational; agent = task-specific, persistent |
| Custom agent use case | Agent creation | Build one when a task needs its own dedicated knowledge or instructions |
| App-by-app differences | Copilot in Word, Excel, Teams, etc. | Feature set is not identical across Microsoft 365 apps |
| Common AI risks | Fabrications, prompt injection, over-reliance | Know all three risk categories by name |
| Verification steps | Citation checks, human review | Match the check to the task's actual risk level |
| Sensitive data | Data protection policies | Some policies restrict what Copilot is allowed to return |
2. Manage Prompts and Conversations by Using AI (35-40%)
| Concept | Feature / Area | Quick note |
|---|---|---|
| Effective prompting | Prompt construction | Reference specific resources, not vague asks |
| Prompt resource refs | @file / @meeting / @person | See the syntax table below |
| Save / schedule / share | Copilot prompt management | Three distinct lifecycle actions — test each separately |
| Conversation management | Find, delete, rename a chat | Basic chat-history operations |
| Notebook | Add a conversation to a notebook | Groups related chats and content together |
| Agent Store vs. new agent | Agent Store / Copilot Studio | See the decision table below |
| Agent from template | Copilot Studio | Fastest path to a working agent |
| Agent knowledge & settings | Instructions, capabilities, suggested prompts | Configure before sharing with a team |
| Share an agent | Team sharing | Distinct action from sharing a single prompt |
3. Draft and Analyze Business Content by Using AI (25-30%)
| Concept | Feature / Area | Quick note |
|---|---|---|
| New document from a prompt | Word / Copilot | Draft-from-scratch pattern |
| Document from an existing doc | Word / Copilot | Regenerate or transform an existing file |
| Management summary | Copilot | Condenses a document for a leadership audience |
| Moving data between apps | Cross-app insights | e.g. moving Excel data and insights into a Word doc |
| Copilot for meetings | Microsoft Teams | Meeting-specific Copilot capability |
| Copilot Pages | Collaboration surface | Shared, persistent AI-assisted content |
| Copilot memory & instructions | Personalization | Copilot retains context and instructions across sessions |
Agent Store vs. Copilot Studio: Quick Decision
This single distinction accounts for a disproportionate share of missed Domain 2 questions:
| Scenario | Use this |
|---|---|
| You need a ready-made agent for a common task (research, analysis) | Agent Store — install a prebuilt agent, no development required |
| You need an agent scoped to your own org's data or knowledge | Copilot Studio — build from a template, then add your own knowledge |
| You want a one-off custom instruction for a repeated personal task | Save a prompt — a full agent is overkill for a single pattern |
| You need to share a working agent with your whole team | Copilot Studio agent, plus the explicit team-sharing step |
Common Mistake: Building an Agent Before Checking the Agent Store
A recurring pattern on the exam — and in real deployments — is jumping straight into Copilot Studio to build a custom agent for something the Agent Store already covers. Say your team needs a research assistant that pulls from multiple sources: before configuring instructions and knowledge sources from scratch, check whether a prebuilt agent in the Agent Store already does the job. Installing an existing agent takes minutes and comes pre-tested; building one in Copilot Studio means you own the configuration, testing, and maintenance going forward. The distinction matters on AB-730 because "can this be done with Copilot" and "should I build a new agent for it" are two different questions — the second only applies once the Agent Store option is ruled out.
Prompt Reference Syntax
| Syntax | References | Use when |
|---|---|---|
| @file | A specific document | Grounding the prompt in one exact file, not a general search |
| @meeting | A specific meeting transcript | Pulling content or decisions from one specific meeting |
| @person | A specific person's email/context | Including someone's own communications as context |
Quick Terms & Acronyms
| M365 | Microsoft 365 — the productivity suite Copilot operates inside |
| GA | General Availability — most exam questions assume GA features, not preview |
| Agent | A dedicated, reusable AI assistant with its own knowledge and instructions, vs. a one-off chat |
| Notebook | A container that groups related conversations and content together |
| Copilot Page | A shared, persistent surface for AI-assisted collaborative content |
Before You Book: AB-730 Questions
How is this cheat sheet different from the AB-730 study guide?
The study guide explains the reasoning behind each domain in narrative form with a 3-week plan. This page strips the narrative for tables you can scan in the last hour before the exam.
Does AB-730 require Copilot Studio development experience?
No coding is required. AB-730 tests configuration-level agent work — templates, knowledge, instructions, and capabilities — not custom development.
What's the fastest way to review right before the exam?
Scan the three domain tables plus the Agent Store vs. Copilot Studio decision table — that one distinction accounts for a disproportionate share of missed Domain 2 questions.
Can I print this AB-730 cheat sheet?
Yes — every table here is plain HTML, so a browser print or "print to PDF" renders cleanly.
MSCertQuiz sells practice-exam access for AB-730 and other Microsoft certifications; this cheat sheet is written by the same team that builds those questions.
Where to Go Next
The 3-week plan and narrative behind this quick reference.
25 free questions with detailed explanations, by domain.
How this Associate exam differs from the Fundamentals-level AB-900.
Full exam details and the complete 500-question practice bank.
Reviewed the cheat sheet? Now drill it.
Start with 40 free AB-730 questions covering every domain above.
Start Free AB-730 Practice