Exam Prep: Scenario Capstone
Cross-domain scenarios, exam strategy, and key takeaways. Put everything together and test yourself with realistic AB-730 exam-style questions.
What to expect on exam day
AB-730 is a practical exam. It doesn’t ask you to recite definitions — it puts you in real business scenarios and asks: “What would you do?”
You’ll see questions like: “Ava needs to prepare a client report using data from Excel and insights from a Teams meeting. What’s the best approach?” The answer requires knowing WHICH Copilot features to use and WHEN.
The secret: Most questions have one answer that’s clearly better than the others. The wrong answers are there because they represent common mistakes real users make.
Exam format
| Detail | Info |
|---|---|
| Duration | 45 minutes |
| Passing score | 700 / 1000 |
| Question types | Multiple choice, multi-select, scenario-based |
| Domain weights | D1: 25-30%, D2: 35-40%, D3: 25-30% |
| Cost | $99 USD |
| Level | Fundamentals |
Key patterns the exam tests
Pattern 1: Right tool for the right task
Many questions give you a scenario and ask which Copilot feature or app to use:
| If the task is… | The answer is usually… |
|---|---|
| Search across multiple apps | Copilot Chat |
| Summarise a specific document | Copilot in Word |
| Analyse spreadsheet data | Copilot in Excel |
| Recap a missed meeting | Copilot in Teams |
| Create a repeating workflow | Scheduled prompt |
| Give consistent team answers | Build an agent |
| One-off question | Copilot Chat |
Pattern 2: Right mitigation for the right risk
| Risk | Mitigation |
|---|---|
| Fabrication | Verify citations, check source documents |
| Prompt injection | Review outputs from untrusted content |
| Over-reliance | Human review, don’t skip judgment |
| Sensitive data exposure | Check outputs before sharing externally |
Pattern 3: Right prompt strategy
Questions often compare a vague prompt vs a specific one. The specific prompt (with goal, context, source, expectations) is almost always the answer.
Cross-domain scenarios
Scenario 1: Ava's quarterly campaign review
Ava at BrightLoop needs to:
- Analyse Q3 campaign data from an Excel spreadsheet
- Create a presentation for the client
- Draft a follow-up email to the team with action items
- Save the analysis prompt for reuse next quarter
Domain 1: She verifies Copilot’s statistical claims against the source spreadsheet (citation check) Domain 2: She saves the analysis prompt to her Prompt Gallery and schedules a monthly version Domain 3: She uses Copilot in PowerPoint to create slides from the Excel data, then Copilot in Outlook to draft the follow-up email
This scenario touches ALL three domains — exactly the kind of question you’ll see on the exam.
Scenario 2: Dana's policy rollout
Dana at Oakfield Healthcare is rolling out a new leave policy:
- She summarises the 20-page policy document into a 1-page employee guide (D3)
- She builds an agent scoped to the new policy for employee questions (D2)
- She sets custom instructions so Copilot always considers Oakfield’s healthcare context (D3)
- She reviews the agent’s outputs for any fabricated policy details (D1)
Exam connection: Knowing that fabrication risk applies even to AGENTS (not just Chat) is a commonly tested concept.
Top 10 exam takeaways
- Copilot uses Microsoft Graph — same permissions as the user
- Your data is not used for AI training — stays in your tenant
- Context = better answers — work files + web + app context
- Chat = general, Agent = specialist — build agents for repeating needs
- Effective prompts have 4 parts — goal, context, source, expectations
- Save, share, and schedule prompts for efficiency
- Agents need knowledge + instructions + starter prompts
- Verify every AI output — citation checks, human review, sensitive data scan
- Transcription required for meeting Copilot — no transcript, no summary
- Memory is user-controlled — view, edit, delete in Settings
🎬 Video walkthrough
🎬 Video coming soon
Exam Prep: Scenario Capstone — AB-730 Module 21
Exam Prep: Scenario Capstone — AB-730 Module 21
~12 minFlashcards
Knowledge Check — Cross-Domain Scenarios
Marcus at Horizon needs to: (1) analyse delivery performance data in Excel, (2) present findings to leadership in PowerPoint, (3) email a summary to the operations team, and (4) discuss findings in a Teams meeting. He wants to reuse this workflow monthly. Which combination of Copilot features should Marcus use?
Jordan shares a sales agent with his team. A new team member, Chloe, uses the agent but it can't answer questions about the latest pricing guide. Jordan can see the pricing guide fine. What is the MOST likely cause?
Ava uses Copilot to draft a client report. The draft includes a statistic about market share that Copilot attributed to a McKinsey study. Ava can't find this study. The report is due tomorrow. What should Ava do?
Congratulations! You’ve completed all 21 modules of the AB-730: AI Business Professional study guide. You now understand generative AI fundamentals, how to craft effective prompts and manage agents, and how to draft and analyse business content with Copilot.
Ready for the exam? Review any modules where you struggled with the quiz questions, then book your exam at Microsoft Learn.