Study guide

5.6 Microsoft Copilot and AI-Powered Applications

Microsoft Azure AI Fundamentals AI-900

Track your progress as you read

Finished sections get a check mark.

1 Introduction

About the Azure AI Fundamentals (AI-900) Exam Study Strategies and Exam Tips for the AI-900

2 Domain 1: Describe AI Workloads and Considerations (15-20%)

1.1 What Is Artificial Intelligence? 1.2 Responsible AI Principles 1.3 Identifying Common AI Workloads and Use Cases 1.4 Responsible AI Tools and Practices on Azure

3 Domain 2: Fundamental Principles of Machine Learning on Azure (20-25%)

2.1 Core Machine Learning Concepts 2.2 Regression Models 2.3 Classification Models 2.4 Clustering Models 2.5 Deep Learning and Neural Networks 2.6 Azure Machine Learning Service

4 Domain 3: Computer Vision Workloads on Azure (15-20%)

3.1 Computer Vision Concepts and Tasks 3.2 Azure AI Vision Service 3.3 Azure AI Custom Vision and Face Services 3.4 Azure AI Document Intelligence

5 Domain 4: Natural Language Processing Workloads on Azure (15-20%)

4.1 NLP Concepts and Common Tasks 4.2 Azure AI Language Service 4.3 Azure AI Speech Service 4.4 Azure AI Translator Service

6 Domain 5: Generative AI Workloads on Azure (15-20%)

5.1 Generative AI Fundamentals 5.2 Azure OpenAI Service 5.3 Prompt Engineering 5.4 Retrieval-Augmented Generation (RAG) 5.5 Responsible Generative AI 5.6 Microsoft Copilot and AI-Powered Applications

7 Exam Review and Full-Length Practice Questions

6.1 Domain Summary and Key Comparisons 6.2 Practice Questions: AI Workloads and Responsible AI 6.3 Practice Questions: Machine Learning 6.4 Practice Questions: Computer Vision, NLP, and Generative AI

5.6 Microsoft Copilot and AI-Powered Applications

Key Takeaways Explain This Topic AI

Microsoft Copilot is a GPT-powered AI assistant integrated into Microsoft 365, Windows, Edge, GitHub, Dynamics 365, and Power Platform.

In Microsoft 365, Copilot drafts emails in Outlook, writes in Word, analyzes data in Excel, builds decks in PowerPoint, and summarizes meetings in Teams.

Copilot uses a RAG pattern over your data via Microsoft Graph, keeping enterprise data inside the Microsoft 365 boundary.

Azure AI Foundry (formerly Azure AI Studio) is the unified portal for building, evaluating, and deploying custom AI apps, with hubs and projects.

Custom AI solutions combine Azure services — e.g., Azure OpenAI + Azure AI Search for grounded chatbots, or Vision + Custom Vision for visual inspection.

Last updated: October 2026

Quick Answer: Microsoft Copilot is an AI assistant integrated into Microsoft 365, Windows, Edge, and other products. It uses GPT models with RAG over your data to help draft emails, build presentations, analyze data, and summarize meetings. Azure AI Foundry is the portal for building custom AI applications. The AI-900 tests awareness of these capabilities.

Microsoft Copilot

Microsoft Copilot is an AI-powered assistant embedded across Microsoft's product ecosystem. It uses large language models (the GPT family) combined with RAG over your data to provide contextual assistance grounded in your content.

Copilot in Microsoft 365

  • Application What Copilot Does Example
  • Outlook Draft emails, summarize threads, suggest replies "Draft a polite reply declining this meeting"
  • Word Write, rewrite, summarize, and format documents "Summarize this 20-page report into key findings"
  • Excel Analyze data, build formulas, generate charts, spot trends "What are the top 5 products by revenue?"
  • PowerPoint Create presentations from prompts or documents "Create a deck from this Word document"

Teams Summarize meetings, capture action items, answer questions "What decisions were made in today's meeting?"

  • OneNote Summarize notes, generate to-do lists, draft content "Create a to-do list from my meeting notes"
  • How Copilot Works (The Architecture)

User prompt — you type a request in natural language.

Grounding — Copilot searches your organization's data (emails, documents, chats, calendars) for relevant context using Microsoft Graph .

Augmented prompt — your request + retrieved context + responsible-AI instructions are combined.

LLM processing — a GPT model generates a response grounded in your data.

Post-processing — content filters check the response and citations are added.

Response — the grounded, filtered answer appears in the application.

This is a RAG pattern : retrieve (Microsoft Graph) -> augment (build the prompt) -> generate (GPT). Knowing this clarifies why Copilot can answer about your meetings and files while a public chatbot cannot — and why your enterprise data stays inside the Microsoft 365 boundary.

Copilot in Other Products

  • Product What Copilot Does
  • Windows System tasks, file search, settings help, content creation
  • Edge Summarize web pages, generate content, answer questions about a page
  • GitHub Code completion, code review, documentation, bug detection
  • Dynamics 365 CRM insights, sales recommendations, customer summaries
  • Power Platform Build apps and automations from natural-language descriptions
  • Azure AI Foundry

Azure AI Foundry (formerly Azure AI Studio) is the unified portal for building, evaluating, and deploying custom AI applications:

  • Feature Description
  • Model Catalog Browse and deploy models from OpenAI, Meta, Mistral, Cohere, and more

Prompt Flow Visual builder for AI application workflows (retiring April 20, 2027; not recommended for new work)

  • Evaluation Tools to assess quality, groundedness, and safety
  • Content Safety Configure content filters and guardrails
  • Deployments Manage model deployments with version control
  • Data and Indexes Connect data sources for RAG and fine-tuning
  • Agent Builder Build agents that call tools and take actions

Within Foundry, a hub is the top-level container for shared resources (compute, storage, connections), and a project is a workspace within a hub for a specific AI application.

Building AI Applications on Azure

Organizations combine Azure AI services to build custom solutions. Common patterns:

  • Pattern Description Azure Services
  • Smart chatbot Natural-language customer service bot Azure OpenAI + Azure AI Search + Bot Service
  • Document processing Automated data extraction from forms Document Intelligence + Azure AI Search
  • Content moderation Review user-generated content Azure AI Content Safety
  • Search and discovery Intelligent search over org knowledge Azure AI Search + Azure OpenAI
  • Real-time translation Multilingual communication platform Azure AI Speech + Azure AI Translator
  • Visual inspection Automated quality control from images Azure AI Vision + Custom Vision
  • The AI Application Stack (conceptual layers)

End-user application (web/mobile/chatbot/Copilot) -> orchestration layer (Microsoft Agent Framework, or Prompt Flow until its April 20, 2027 retirement) -> AI models (Azure OpenAI, Azure AI services) -> data layer (Azure AI Search, SQL, Cosmos DB) -> infrastructure (subscriptions, RBAC, networking).

Copilot vs. Building Your Own AI Application

A common point of confusion is the difference between using Copilot and building a custom AI application :

  • Microsoft Copilot Custom application (Azure AI Foundry / Azure OpenAI)
  • Who builds it Microsoft (it is a finished product) Your developers
  • How you get it Buy/enable a Copilot license Provision and code Azure AI services
  • Data grounding Automatic via Microsoft Graph You build the RAG pipeline (Azure AI Search)
  • Customization Limited to product settings and plugins/agents Full control over models, prompts, data, and flow
  • Best when You want productivity in M365 apps fast You need a tailored experience for your own scenario

In short, Copilot is a ready-made assistant embedded in products you already use, while Azure AI Foundry is where you build your own assistant from the underlying services. Microsoft also offers Copilot Studio , a low-code tool for building custom copilots and agents that extend Copilot or stand alone — useful when you want more customization than Copilot's defaults but less coding than a full Foundry application.

Knowing where each option sits on the build-vs-buy spectrum answers many "which approach should this organization take?" questions: pick Copilot for fast in-app productivity, Copilot Studio for low-code custom copilots, and Azure AI Foundry plus Azure OpenAI for fully bespoke applications.

On the Exam: Know that Copilot uses RAG over organizational data via Microsoft Graph to give grounded responses, and that Azure AI Foundry is the portal for building custom AI applications. You do not need to build applications — just recognize the concepts and which services solve which problem.