Oops! Sorry!!


This site doesn't support Internet Explorer. Please use a modern browser like Chrome, Firefox or Edge.

Azure OpenAI & Copilot — BFT Consulting

AB-730 / AB-731 Quiz

Azure OpenAI, Copilot & Agents — Exam preparation

Select how many questions you want to attempt

All 20 questions are scenario-based, modelled on the actual AB-730 and AB-731 exam style. Each question has four options. After each answer you will receive a full explanation and an exam tip.

5Quick check
10Half session
15Full session
20Complete set
correct

Question-by-question breakdown

BFT Consulting Services
Azure OpenAI → Microsoft 365 Copilot → Agents
Interactive study guide — AB-730 / AB-731 Exam Prep
Click any element below to reveal a detailed explanation with consulting examples and exam tips
Who uses the system
👥

You and Your Users

BFT Consulting staff, client organisations — asking questions, running prompts, delegating tasks to agents in Microsoft 365 apps

👥 You and Your Users

This is the starting point. A user types a prompt into a Microsoft 365 app. The system processes it through Copilot, sends it to Azure OpenAI, gets a response, and returns it — all within seconds.

BFT Example: A consultant types in Teams: "Summarise the last three emails from our client and suggest a response." Copilot reads the emails via Microsoft Graph, sends context to Azure OpenAI, and delivers a draft reply in Outlook.
📝 Exam tip: The user does NOT talk directly to Azure OpenAI. The user talks to Copilot. Copilot talks to Azure OpenAI. This distinction is tested in governance and responsible AI questions.
interact with
Layer 3 — Agents (what users interact with directly)

🤖 Microsoft Prebuilt Agents

Ready-made — no setup needed. Microsoft built these for you.
Analyst
Researcher
Prompt Coach

⚙️ Your Custom Agents

Built in Copilot Studio — trained on your own data and rules.
Custom agents
Automated flows

🤖 Microsoft Prebuilt Agents

Microsoft has already built certain agents and embedded them directly into Microsoft 365 Copilot. They come with your Copilot licence — no configuration needed.

BFT Example: When you open Microsoft 365 BizChat, the Analyst, Researcher, and Prompt Coach agents are available immediately. No IT setup required.
📝 Exam tip (AB-731): Know all three agents by name, purpose, and when to recommend them. The exam distinguishes between prebuilt (ready now) and custom (requires Copilot Studio to build).

📊 Analyst Agent

Connects to your Excel and CSV files and performs data analysis, trend identification, and chart generation using natural language.

BFT Example: Upload a client's employee survey results and ask: "What are the top 3 concerns about AI adoption?" The Analyst reads the file and gives a ranked summary with a chart — no Power BI skills needed.
📝 Exam tip: Analyst = structured data files (Excel, CSV). Researcher = the web. This distinction is frequently tested in scenario questions.

🔍 Researcher Agent

Performs deep, multi-step web research — browsing multiple sources, synthesising findings, and producing a structured research report.

BFT Example: Ask: "Research AI adoption in Nigerian banking and give me a 5-point summary with sources." The Researcher browses, synthesises, and delivers a referenced report — saving 2+ hours of manual work.
📝 Exam tip: Researcher = web/external sources. Analyst = your internal files. Confusing these two is the most common error in agent scenario questions.

✍️ Prompt Coach Agent

Helps users write better prompts by analysing what they wrote and suggesting improvements — making prompts clearer, more specific, and more likely to produce useful output.

BFT Example: A junior consultant types "Tell me about AI." Prompt Coach suggests a much more specific version tailored to their actual need — dramatically improving the Copilot response.
📝 Exam tip (AB-730): Prompt Coach directly relates to prompt engineering — a core AB-730 topic. Knowing this agent helps answer questions about helping employees use Copilot more effectively.

⚙️ Your Custom Agents

Using Microsoft Copilot Studio (low-code, included with certain M365 licences), you build agents trained on your organisation's specific knowledge, rules, and workflows.

BFT Example: Build a "BFT Methodology Agent" trained on all BFT proposal templates and frameworks. A junior consultant asks it to draft an AI readiness assessment — it responds using BFT's own IP.
📝 Exam tip (AB-731): Custom agents = the build decision. The exam tests when to build custom vs use prebuilt. Factors: unique data, specific workflows, competitive differentiation.

⚡ Automated Flows

Agents in Copilot Studio can be connected to Power Automate to take actions automatically — approvals, emails, record updates — without human clicking.

BFT Example: A procurement agent reads incoming purchase requests, checks the budget database, and either auto-approves routine purchases or escalates to a manager — all without manual review.
📝 Exam tip: This is agentic AI — AI that acts, not just answers. AB-731 tests governance and risk implications of agentic AI. Know the term.
powered by
Layer 2 — Platform (connects AI to your Microsoft 365 apps and your data)

🧠 Microsoft 365 Copilot

Word Excel Outlook Teams PowerPoint BizChat
Connects to Microsoft Graph (your emails, calendars, files, meetings) — applies responsible AI safety filters before showing results

🧠 Microsoft 365 Copilot — the platform layer

Copilot is the middle layer. It (1) takes your prompt from whatever app you're using, (2) retrieves relevant context from your M365 data via Microsoft Graph, and (3) sends everything to Azure OpenAI and returns the answer — with safety filters applied.

BFT Example: You ask Copilot in Teams: "Prepare me for my 3pm meeting today." Copilot reads your calendar, finds relevant emails, pulls related files, and gives you a 3-minute briefing — without you opening a single file.
📝 Exam tip (AB-730): Microsoft Graph is the data connector. Copilot only shows data users are already authorised to see — it does not bypass existing permissions.

📄 Copilot in Word

Drafts documents, rewrites sections, changes tone, summarises long documents, and generates content from a brief description or existing file.

BFT Example: Upload a client brief and type: "Draft a consulting proposal. Include an executive summary, methodology section, and pricing framework." Copilot produces a first draft in 30 seconds.

📊 Copilot in Excel

Analyses data, creates formulas, generates charts, highlights trends — all using plain English questions. No formula knowledge required.

BFT Example: Ask: "Which regions had the highest revenue growth in Q3 and what might be driving that?" Copilot analyses and produces a summary with a chart.

📧 Copilot in Outlook

Summarises long email threads, drafts replies, adjusts tone, flags urgent messages, and helps schedule meetings.

BFT Example: A 40-email client thread. Ask: "Summarise this thread and tell me what action is needed from me." You get a 4-line summary and a suggested reply — saving 20 minutes.

💬 Copilot in Teams

Transcribes meetings in real time, summarises discussions, captures action items, and answers questions about what was discussed — even if you joined late.

BFT Example: You join a client workshop 20 minutes late. Ask Copilot: "What has been discussed so far?" Copilot gives you a live summary without disrupting the meeting.
📝 Exam tip (AB-730): Copilot in Teams meetings only works when transcription is enabled. Participants must be informed — this is a responsible AI / privacy point.

📊 Copilot in PowerPoint

Creates full presentations from a prompt or Word document, redesigns slides, adds speaker notes, summarises long decks.

BFT Example: Ask: "Create a 10-slide executive presentation on AI adoption strategy for a Nigerian bank." Copilot produces a complete deck with titles, bullets, and speaker notes.

💡 Microsoft 365 BizChat

The central Copilot chat interface — available in Teams and on the web. This is where the three prebuilt agents live. Combines web search, organisational data, and AI in one place.

BFT Example: Open BizChat and type: "Find the proposal we wrote for our 2022 banking engagement and summarise the recommended approach." BizChat searches SharePoint and delivers a summary.
📝 Exam tip: BizChat / Copilot Chat is the home of agents. The exam may ask where you access the Analyst or Researcher agent — answer: Microsoft 365 BizChat (or Copilot in Teams).
runs on
Layer 1 — Foundation (the AI engine Microsoft runs securely on Azure)

⚙️ Azure OpenAI Service

Microsoft-hosted AI models — enterprise-grade, private, secure. Your data stays in Azure. Never used to train other companies' models.
🗣️
GPT-4o
Language
👁️
Vision
Images + docs
🔎
AI Search
Your data
🎨
DALL-E
Images
🎙️
Speech
Audio to text

⚙️ Azure OpenAI Service

OpenAI built the AI models. Microsoft took those same models and hosted them inside Azure's secure cloud. That is Azure OpenAI Service — the same intelligence as ChatGPT, but within Microsoft's enterprise security boundary.

BFT pitch to a bank client: "When your staff use Copilot, their queries go to Azure OpenAI — not ChatGPT. The AI runs inside Microsoft's secure Azure environment. Your customer data never leaves your regulatory boundary."
📝 Exam tip (AB-731): Three points for the data security question: (1) Azure OpenAI, not public ChatGPT, (2) data not used for model training, (3) existing permissions enforced by Microsoft Graph.

🗣️ GPT-4o — Language Model

The core language capability. Reads text, understands context, generates high-quality responses. Powers most of what Copilot does — drafting, summarising, rewriting, answering questions.

BFT Example: A client uploads a 200-page policy document. Ask: "What are the 5 most important AI compliance requirements?" GPT-4o reads and summarises — saving a lawyer 3 hours.
📝 Exam tip: You don't need technical knowledge of GPT-4o. Know what it does (language tasks) and why the enterprise Azure version is more appropriate than public ChatGPT for organisations.

👁️ Vision — Reading Images and Visual Documents

Allows Azure OpenAI to see and interpret images, charts, screenshots, and scanned documents — not just typed text.

BFT Example: A client pastes a screenshot of a legacy system dashboard. Ask Copilot: "What trends do you see and what might be causing the August dip?" Copilot reads the image and responds — even though the data was never typed out.
📝 Exam tip: Vision is relevant to AB-731 scenarios about integrating AI with legacy systems. If a system can export screenshots or PDFs, Vision can read them — a practical bridge when APIs are unavailable.

🎨 DALL-E — Image Generation

Generates images from text descriptions. Used in Microsoft Designer, Copilot in PowerPoint, and BizChat.

BFT Example: Type: "A professional diagram showing a Nigerian organisation transforming from manual to AI-powered workflows, clean corporate style, blue and white." DALL-E generates a custom image in seconds.
📝 Exam tip: DALL-E is less tested in AB-730/AB-731. Know it exists and what it does — image generation from text — but don't over-invest study time here.

🎙️ Speech — Voice to Text and Text to Voice

Converts spoken audio into text and text back to speech. Powers Teams meeting transcription, real-time captions, and voice-input Copilot features.

BFT Example: A 2-hour client strategy workshop over Teams is transcribed automatically. At the end, Copilot summarises the discussion, lists all action items, and identifies who is responsible — without anyone taking manual notes.
📝 Exam tip (AB-730): Speech = Copilot in Teams meeting management — a tested AB-730 topic. Transcription must be enabled and participants must be informed. That is a privacy and responsible AI point.