AB-730
Free AB-730 practice questions
10 questions from our AB-730 practice tests, each with the correct answer and a written explanation. Read them with the answers shown, or switch to quiz mode and answer first.
Question 1 of 10, Understand generative AI fundamentals
A contractor has Microsoft 365 Copilot Chat through her work account but no Microsoft 365 Copilot license. She asks Copilot Chat to summarize her unread Outlook emails. What should she expect?
- ACopilot Chat forwards the emails to a Microsoft reviewer who prepares the summary within one business day
- BCopilot Chat summarizes only emails from internal senders, and external messages require an add-on license
- CCopilot Chat summarizes her mailbox as long as she has opened Outlook on the web at least once during that day
- DCopilot Chat without that license can't ground on her mailbox, so a Microsoft 365 Copilot license is needed Correct answer
Answer and explanation: Copilot Chat is available with an Entra work account and offers web-grounded chat with enterprise data protection, but grounding on the user's own mail, files, and meetings through Work IQ requires a Microsoft 365 Copilot license. Opening Outlook doesn't change this, there's no internal-only mailbox mode, and no human reviewer prepares summaries.
Question 2 of 10, Understand generative AI fundamentals. Choose 2.
A user writes a prompt telling Copilot to ignore its rules and reveal a confidential salary file she can't access. Which statements are correct? (Choose two.)
- ACopilot still can't return content the user doesn't have permission to access Correct answer
- BThis is a direct prompt injection or jailbreak attempt by the user Correct answer
- CCopilot will reveal the file if the request is repeated in several new chats in a row
- DThis is a hallucination risk, because the model is being asked to invent salary data
- EThis is a cross-prompt injection, because the instruction came from an external email
Answer and explanation: A user directly instructing the AI to bypass its rules is a jailbreak, or direct prompt injection, and permissions still prevent access to content the user can't open. Cross-prompt injection comes from external content, repeating the request doesn't defeat permissions, and the scenario is about manipulation rather than fabrication.
Question 3 of 10, Understand generative AI fundamentals
Your legal team asks whether prompts and responses entered in Microsoft 365 Copilot Chat while signed in with a work account are used to train the public foundation models. What is the correct answer?
- ANo. Under enterprise data protection, prompts and responses aren't used to train foundation models Correct answer
- BYes, but only anonymized excerpts are used, and only after a 30-day retention period has expired
- CYes, for web-grounded prompts only, because web searches are processed by a consumer search service
- DNo, unless a user selects the thumbs-up button, which opts that conversation into model training
Answer and explanation: Copilot Chat used with an Entra ID work account is covered by enterprise data protection, and Microsoft commits that prompts and responses aren't used to train foundation models. Feedback ratings don't opt a chat into training, there is no anonymized-excerpt training path, and web grounding doesn't change the training commitment.
Question 4 of 10, Manage prompts and chats by using Microsoft 365 Copilot
A team uses Copilot Notebooks and Copilot Pages. Which statement BEST describes when to use each?
- AUse a notebook only for images and charts; use a page only for text-based meeting summaries
- BUse a notebook to schedule prompts automatically; use a page to manage permissions for chats
- CUse a notebook to gather many references for ongoing Q&A; use a page to co-author one output Correct answer
- DUse a notebook for real-time co-editing with Copilot; use a page to store references for later Q&A sessions
Answer and explanation: Notebooks collect files, notes, and other references so Copilot can answer questions grounded in them over time, while Pages are collaborative canvases for co-creating content from Copilot responses. Neither is limited to images or text only, and neither schedules prompts or manages chat permissions.
Question 5 of 10, Manage prompts and chats by using Microsoft 365 Copilot
A user saved a prompt that summarizes customer feedback. Results have become too long since product lines were added. She wants to keep using it but improve the output. What is the BEST approach?
- ADelete her chat history so the saved prompt starts again from the original product lines
- BShare the prompt with her team so their feedback ratings retrain the model on shorter answers
- CRun it, refine the wording based on the response, and save the improved version for reuse Correct answer
- DTurn on a scheduled run so Copilot learns to shorten results after several weekly executions
Answer and explanation: Prompts are refined iteratively: test, adjust the wording or constraints, and save the improved version. Deleting chat history doesn't change the prompt, scheduled runs don't train the model, and team ratings don't retrain Copilot on the fly.
Question 6 of 10, Manage prompts and chats by using Microsoft 365 Copilot
A project lead wants Copilot to draft a status update based on a specific planning document and last week's email thread with the vendor. What is the BEST way to make sure Copilot uses those items?
- AForward the email to yourself and rename the document so its title begins with the word Copilot
- BCopy both items into a new OneNote page and ask Copilot to wait until indexing has finished
- CDescribe the document's title in detail and let Copilot search for something similar on the web
- DReference the file and the email in the prompt by typing / and selecting them Correct answer
Answer and explanation: Typing / (or using the attach option) in Copilot lets you point it at specific files, emails, meetings, or people so the response is grounded in exactly those sources. Web search won't find internal content, copying to OneNote adds delay without guaranteeing use, and renaming or forwarding items doesn't steer grounding.
Question 7 of 10, Manage business content and collaboration by using Microsoft 365 Copilot
A product manager wants Copilot to answer questions using tickets stored in the company's ServiceNow instance. An administrator has already configured the connection. How does this content become available to Copilot?
- AThrough a browser extension that lets Copilot read whatever ServiceNow page is currently open
- BThrough a Copilot connector that brings the external content into the organization's Work IQ context Correct answer
- CThrough a Power Automate flow that emails each new ticket to the manager so Copilot reads it
- DThrough a nightly export of tickets to an Excel file that the user attaches to every prompt
Answer and explanation: Copilot connectors let admins bring data from external systems such as ServiceNow into Microsoft 365 so Copilot can reason over it, respecting the source permissions. Emailing tickets, browser extensions, or manual Excel exports are workarounds that don't integrate the data into Copilot's work context.
Question 8 of 10, Manage business content and collaboration by using Microsoft 365 Copilot. Choose 2.
A scheduled Teams meeting has the Facilitator agent enabled. Which tasks can Facilitator perform during the meeting? (Choose two.)
- ARecord the meeting automatically even when the organizer has disabled recording
- BPost the agenda from the invite and help keep the discussion on track Correct answer
- CGrant external guests access to private channel files mentioned in the meeting
- DTranslate each speaker's audio into the listener's preferred language in real time
- ETake collaborative AI notes in a shared notes page that participants can edit Correct answer
Answer and explanation: Facilitator reads the agenda from the invite or notes, posts it, helps manage time and topics, and generates collaborative notes that participants can edit. Real-time speech translation is the Interpreter agent's job, and Facilitator neither changes file permissions nor overrides recording policies.
Question 9 of 10, Drive business outcomes by using agents
A team wants Copilot Cowork to follow its own format whenever it produces the weekly status report, including fixed headings and a specific approval checklist. No built-in skill covers this. What should a team member do?
- ACreate a Copilot Notebook with the format and reference it as a connector so Cowork treats it as a built-in skill
- BPaste the format into a saved prompt and set Cowork to load the Prompt Gallery as a skill at the start of each session
- CCreate a custom skill by adding a SKILL.md file with a name, description, and instructions to the Cowork skills folder in OneDrive Correct answer
- DAsk an administrator to upload the checklist as a plugin to the Microsoft 365 App Store, because users cannot extend Cowork
Answer and explanation: Users can create custom Cowork skills by placing a SKILL.md file (YAML frontmatter with name and description, then instructions) in a subfolder of their OneDrive /Documents/Cowork/skills/ folder; Cowork discovers it at the start of each session. Users don't need an admin plugin for this, saved prompts aren't loaded as skills, and notebooks aren't connectors or skills.
Question 10 of 10, Drive business outcomes by using agents
A department head must choose between a prebuilt agent and a custom agent. The team needs answers grounded in its own SharePoint policy library, with a consistent tone and fixed instructions, used daily by 40 staff. Which approach is BEST?
- AUse the Researcher agent and ask each staff member to include the SharePoint URL and tone in every prompt
- BUse Copilot Cowork and schedule a prompt that answers policy questions for each staff member every morning
- CCreate a custom agent with instructions and the policy library as knowledge, then share it with the team Correct answer
- DUse the Analyst agent and upload the policy documents as CSV files so it can run Python over the text
Answer and explanation: When a recurring task needs specific knowledge, fixed instructions, and consistent behavior for many users, a custom agent with that knowledge and instructions is the best fit. Relying on each user to repeat context in Researcher is inconsistent, scheduled Cowork prompts can't answer ad hoc questions, and Analyst is for data analysis rather than policy Q&A.