Microsoft Certified AI Transformation Leader 國際認可證書課程 (1 科 AI 轉型)
課程簡稱：Microsoft AI Transformation Training Course

本中心是 Microsoft 認可的合作夥伴 (Microsoft Certified Partner)。

微軟為您的企業提供 AI 整合式解決方案，從 Microsoft 365 Copilot、智能代理 (Agents) 到強大的 Foundry 功能，令即使沒有任何 AI 技術背景的您，也能輕鬆步入 AI 協作新時代。掌握生成式 AI 商務應用，成為 AI 時代的生產力領導者！

生成式 (Generative) AI 正重塑現代商務工作流程。它已不僅是對話工具，更是能自動執行任務的智能代理 (Agent)，甚至是能幫您自動讀取報表、辨識影像的企業級大腦。

本課程專為應考 Microsoft Certified AI Transformation Leader 認證考試而設，助您在無須編程 (Zero Coding) 的情況下，全面掌握 Copilot 與智能代理 (Agent) 的商務應用能力，並通過考試。

微軟為各行各業提供整合式解決方案，本課程將教您如何準確評估「AI 成本驅動因素 (Cost drivers)」，極大幅度地節省購買軟件使用授權 (Software License) 的成本及避免不必要的 I.T. 硬件基建開支！

AI 技術助您事半功倍

• 全面了解 Microsoft 365 Copilot 在不同應用 (Outlook、Word、Teams、PowerPoint、Excel) 中的 AI 功能體驗，並將您的業務流程與 Copilot 解決方案完美對接。
• 掌握微軟 Foundry 工具 (Azure AI)： 無需寫程式，介紹如何應用 Azure 視覺 (Vision)、語言處理 (Language)、智能文件擷取 (Document Intelligence) 與 Azure AI 搜尋 (AI Search)，解決真實的商業挑戰。
• 理解「前文後理 (Context) 與 RAG (檢索增強生成)」 如何影響 AI 回應品質，讓 AI 運用「可信數據 (Trusted data)」與您公司的專屬檔案來回答問題，排除幻覺。
• 發掘機器學習 (Machine Learning) 的商業價值： 透切了解何時該使用機器學習或生成式 AI。
• 區分「聊天體驗 (Chat)」與「代理體驗 (Agent)」的應用場景，了解何時該親自下指令，何時該讓 AI 自動化執行。
• 負責任 AI 與數據保護： 識別 AI 常見風險 (虛構內容、提示注入攻擊、過度依賴)，學會引用查閱與人工審核等驗證步驟，保護敏感資料。

AI 技術助您企業轉型

• 統一數據資產與現代化基礎設施 (Unified Data Estate & Modernization)： 結合 Azure 雲端與 AI 平台，為企業打造堅實的數位基礎。
• 將 AI 實驗化為可靠驅動力： 緊密結合業務策略 (Business strategy)、科技數據策略與組織文化 (Organization culture)，確保 AI 專案帶來可衡量的投資回報。
• 賦能內部人才： 不需要所有人都是工程師。學會如何賦能一般業務使用者 (Business users) 與領域專家 (Subject matter experts)，讓他們在日常營運中自主使用 AI。
• 透過 AI 翻譯、語言分析與智能線上會議工具，無縫連繫全球客戶、同事和供應商。您可以隨時隨地透過任何裝置，安全地使用企業 AI 能力工作。

AI 技術保護您的資料安全與合規 (Security & Compliance)

• 落實微軟負責任的 AI 框架 (Responsible AI in action)： 學習微軟的 AI 原則與治理系統，為您的企業設計專屬的 AI 政策、流程與安全防護網 (Safeguards)。
• 防範 AI 安全威脅與建構信任： 識別並阻止惡意提示 (Prompt Injection) 及非法存取，確保 AI 的輸出符合道德且合規。
• 防止過度分享： 監控文件權限，防止敏感資料因「過度分享」而被外界錯誤獲取。
• 控制 AI 數據存取： 精準定義 AI 處理特定的商務資訊的權利，阻止 AI 存取商業機密 (採用 Azure 企業級雲端保護)。
• 自動化合規工具： 建立 AI 治理系統，確保 AI 應用符合法規 (Compliance) 與隱私政策 (例如中國個人信息保護法 PIPL、香港個人資料 (私隱) 條例 PCPD、美國 HIPAA 及歐盟 GDPR 等等)。

本課程聚焦於商務人士與零基礎大眾，學習如何以最低成本配置最強 AI 辦公環境。我們將解析 Copilot 的授權選項 (Licensing) 與微軟 Foundry 的訂閱模式，避免您購買不必要的昂貴授權。

本課程教授的技術適用於各大中小型企業，能讓員工發揮出頂級的 AI 人機協作生產力，並能保護各個裝置及雲端上的商務資料，為您的企業提供了一個完善的解決方案。助您在善用 AI 的同時，保護商業機密，並實現可衡量的商業價值 (Measurable business value)。

修畢本課程後，學員便可考取下列 1 張國際認可證書：

Microsoft Certified AI Transformation Leader

本中心為Microsoft指定的考試試場。報考時請致電本中心，登記欲報考之科目考試編號、考試日期及時間 (最快可即日報考)。臨考試前要出示身份證及繳付每科HK$566之考試費。

考試題目由澳洲考試中心傳送到你要應考的電腦，考試時以電腦作答。所有考試題目均為英文，而大多數的考試題目為單項及多項選擇題，其餘則為實戰題。

考試合格後會收到來自Microsoft的作實電郵，並進入該電郵內的連結，登入 Microsoft Credentials Dashboard 下載您的證書。

考試不合格便可重新報考，不限次數。欲知道作答時間、題目總數、合格分數等詳細考試資料，可瀏覽本中心網頁 "各科考試分數資料"。

Exam AB-731: Microsoft Certified AI Transformation Leader

1. The Foundations of generative AI for business leaders
1.1 Generative AI introduction
1.1.1 What is generative AI?
1.1.2 How does generative AI work?
1.1.3 Business applications of generative AI
1.2 Business value of generative AI solutions
1.2.1 Explore Microsoft generative AI solutions
1.2.2 Understand agents in Microsoft generative AI solutions
1.2.3 Agents in Microsoft Copilot
1.2.4 Agents in Microsoft Foundry
1.2.5 Understand scenario-based business value
1.2.6 Automate report generation
1.2.7 Identify and evaluate business value
1.2.8 Map business needs to recommended solution types
1.3 Understanding generative AI models
1.3.1 Types of generative AI models
1.3.2 Choose the right model for your business
1.4 Cost drivers in generative AI
1.4.1 Why generative AI costs matter
1.4.2 Cost drivers in Microsoft Copilot
1.4.3 Cost drivers in Azure AI
1.4.4 Use Copilot and Azure AI together
1.5 Challenges and Opportunities in generative AI
1.5.1 Incorrect or fabricated responses
1.5.2 Reliability
1.5.3 Bias
1.5.4 Other challenges
1.5.5 Scenario: Turning a challenge into an opportunity

2. Building Effective Generative AI solutions in your organization
2.1 Understanding Prompt Engineering
2.1.1 Why does prompt engineering matter?
2.1.2 Explore prompt engineering
2.1.3 Prompt engineering for business users
2.1.4 G-C-S-E prompting in depth
2.1.5 Prompt engineering for developers
2.2 Ground AI using trusted data
2.2.1 What is grounding?
2.2.2 What is retrieval-augmented generation (RAG)?
2.3 Build Trustworthy AI - Data and security considerations
2.3.1 Why data quality matters
2.3.2 Data and responsible AI
2.3.3 Security
2.3.4 Best practices for enterprise adoption
2.4 Business value of Machine Learning (ML)
2.4.1 What is machine learning?
2.4.2 Capabilities of machine learning models
2.4.3 Understand the machine learning lifecycle
2.4.4 Business considerations for machine learning adoption
2.4.5 Build or buy

3. Drive business value with Microsoft Copilot solution
3.1 Copilot solutions
3.1.1 What is Microsoft Copilot?
3.1.2 Overview of Copilot solutions
3.2 Copilot Chat – Your AI Assistant
3.2.1 Access Copilot Chat
3.2.2 Work and Web modes in Copilot Chat
3.2.3 How to chat in three steps
3.2.4 Copilot Memory
3.2.5 Search with Copilot Chat
3.3 Writing effective prompts to optimize Copilot responses
3.3.1 Add references to your prompts
3.3.2 Key benefits for organizations
3.4 More about Copilot Chat experiences
3.4.1 Copilot Chat: fast answers and organizational context
3.4.2 Copilot in Microsoft 365 apps: embedded intelligence for everyday work
3.4.3 Copilot Search: AI-powered enterprise search
3.4.4 Researcher Agent
3.4.5 Analyst Agent
3.5 Mapping business processes to Copilot solutions
3.5.1 Identify high-impact processes
3.5.2 Map processes and use cases to Copilot solutions
3.6 Extend and customize Copilot
3.6.1 Copilot Studio: build and customize without heavy coding
3.6.2 Create a Microsoft 365 Copilot Chat agent - Part A
3.6.3 Use the Describe tab (if available) to begin the agent creation process
3.6.4 Create a Microsoft 365 Copilot Chat agent - Part B
3.6.5 Microsoft Graph: the backbone of enterprise context
3.6.6 Decide when to buy, extend, or build Copilot
3.6.7 Integrate with Azure AI: advanced scenarios
3.7 Responsible AI in Copilot
3.7.1 Safety filters and transparency
3.8 Responsible AI principles
3.8.1 Fairness
3.8.2 Inclusiveness
3.8.3 Reliability and Safety
3.8.4 Privacy and security
3.8.5 Transparency
3.8.6 Accountability
3.8.7 What are the responsible AI features in Microsoft 365 Copilot?
3.8.8 Is my data kept private and secure?
3.8.9 Avoid overreliance
3.8.10 Human oversight and compliance basics
3.8.11 Integrated Microsoft AI solutions reduce risk
3.9 Copilot licensing
3.9.1 Copilot licensing options
3.9.2 Understand Microsoft 365 license types

4. Business value with AI using Microsoft Foundry Tools
4.1 Why Microsoft Foundry matters
4.1.1 Foundry tools introduction
4.1.2 Keep AI secure and compliant: Enterprise-grade security and compliance
4.1.3 Scale without disruption and integrate with what you have: Scalability and integration
4.1.4 Unlock generative AI safely and at scale: Generative AI with guardrails
4.2 Core AI capabilities
4.2.1 Foundry in action
4.3 Connect business scenarios to Foundry solutions
4.3.1 Core Foundry solutions
4.3.2 Explore common scenarios
4.3.3 Automate visual inspections for better quality control
4.3.4 Analyze customer sentiment and deliver multilingual support
4.3.5 Streamline document-heavy workflows for faster processing
4.3.6 Enable quick knowledge retrieval and self-service
4.3.7 Orchestrate generative AI for scalable content creation
4.3.8 Foundry solutions in action
4.4 Choosing the right Foundry model for your goals
4.4.1 Choose a right-sized model
4.4.2 Common model approaches in Foundry
4.4.3 Retrieval-Augmented Generation (RAG)
4.4.4 Fine-tuning or custom models
4.4.5 Compare at a glance
4.4.6 Put model choices into action
4.5 Foundry subscription models
4.5.1 The importance of pricing in AI adoption
4.5.2 Subscription models at a glance
4.5.3 Cost optimization strategies

5. Leverage AI tools and resources for your business
5.1 The Microsoft AI approach
5.2 Copilot and agents at work
5.2.1 Accelerate personal productivity with Copilot
5.3 Key AI capabilities for productivity
5.4 The Microsoft 365 Copilot Adoption site
5.4.1 Case Study: Structured rollout at a university
5.5 Manage your AI agents with Agent365
5.6 Real-world use cases for Copilot and agents
5.6.1 Introduction
5.6.2 Use cases for Microsoft 365 Copilot
5.6.3 Use cases for the Researcher agent in Microsoft 365 Copilot
5.6.4 Use cases for the Analyst agent in Microsoft 365 Copilot
5.6.5 Use cases for other prebuilt agents
5.7 Drive transformation with Azure cloud and AI platforms
5.7.1 Introduction to Azure Cloud
5.7.2 Unify your data platform
5.7.3 Innovate with Azure AI apps and agents
5.8 Secure AI for a strong foundation
5.8.1 Understand AI business risks
5.8.2 Get started with a phased approach
5.8.3 Governing AI
5.8.4 Data governance and security
5.8.5 Build a foundation for effective AI governance
5.8.6 Stay ahead of compliance with regulatory governance
5.8.7 Building a strong foundation for AI compliance
5.8.8 Navigate AI compliance

6. Embrace responsible AI principles and practices
6.1 Prepare for responsible AI
6.1.1 Societal implications of AI
6.1.2 The importance of a responsible approach to AI
6.1.3 Novel threats
6.1.4 Biased and unfair outcomes
6.1.5 Sensitive use cases
6.2 Guiding principles for responsible AI
6.2.1 Fairness
6.2.2 Reliability and safety
6.2.3 Privacy and security
6.2.4 Inclusiveness
6.2.5 Transparency
6.2.6 Accountability
6.3 Design a system for AI governance
6.3.1 Chief Ethics Officer
6.3.2 Ethics office
6.3.3 Ethics committee

7. AI Transformation Leader Strategy Guide
7.1 About this guide
7.2 Preparing with this guide: the three-stage method
7.3 Assessment strategy: the method
7.3.1 Read the constraint, not just the story
7.3.2 Common traps in the wrong choices
7.4 Key Topics Abstraction: Generative AI foundations
7.4.1 What generative AI is
7.4.2 How the models behave
7.5 Prompt engineering and grounding
7.5.1 Prompt engineering
7.5.2 Grounding and retrieval
7.6 Microsoft Foundry tools
7.6.1 The service map
7.6.2 Keeping Machine Learning models healthy
7.7 Governance, adoption and costs
7.7.1 Governance
7.7.2 Adoption
7.7.3 Costs
7.7.4 Shadow AI in the workplace
7.7.5 Inappropriate language in generated content
7.8 Model selection and evaluation in depth
7.8.1 Right-sizing as the default habit
7.8.2 Reasoning versus non-reasoning models
7.9 The three-layer diagnosis: a troubleshooting framework
7.9.1 The knowledge layer
7.9.2 The behavior layer
7.9.3 The control layer
7.10 Data protection and compliance essentials
7.10.1 Controlling what AI can reach
7.10.2 Minimising what AI processes
7.10.3 Governing compliance over time
7.11 Common misconceptions
7.12 Frequently asked questions
7.12.1 "Do I need to memorize the services?"
7.12.2 "How much detail should I give in explanations?"
7.12.3 "Is the newest model always the best choice?"
7.12.4 "What if two options both seem correct?"
7.12.5 "Should I answer from what I know about the products?"
7.12.6 "How do I handle the long scenarios?"
7.12.7 "Are the statement sets really independent?"