5.6 Microsoft Copilot and AI-Powered Applications

Key Takeaways

  • 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: June 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

ApplicationWhat Copilot DoesExample
OutlookDraft emails, summarize threads, suggest replies"Draft a polite reply declining this meeting"
WordWrite, rewrite, summarize, and format documents"Summarize this 20-page report into key findings"
ExcelAnalyze data, build formulas, generate charts, spot trends"What are the top 5 products by revenue?"
PowerPointCreate presentations from prompts or documents"Create a deck from this Word document"
TeamsSummarize meetings, capture action items, answer questions"What decisions were made in today's meeting?"
OneNoteSummarize notes, generate to-do lists, draft content"Create a to-do list from my meeting notes"

How Copilot Works (The Architecture)

  1. User prompt — you type a request in natural language.
  2. Grounding — Copilot searches your organization's data (emails, documents, chats, calendars) for relevant context using Microsoft Graph.
  3. Augmented prompt — your request + retrieved context + responsible-AI instructions are combined.
  4. LLM processing — a GPT model generates a response grounded in your data.
  5. Post-processing — content filters check the response and citations are added.
  6. 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

ProductWhat Copilot Does
WindowsSystem tasks, file search, settings help, content creation
EdgeSummarize web pages, generate content, answer questions about a page
GitHubCode completion, code review, documentation, bug detection
Dynamics 365CRM insights, sales recommendations, customer summaries
Power PlatformBuild 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:

FeatureDescription
Model CatalogBrowse and deploy models from OpenAI, Meta, Mistral, Cohere, and more
Prompt FlowVisual builder for AI application workflows
EvaluationTools to assess quality, groundedness, and safety
Content SafetyConfigure content filters and guardrails
DeploymentsManage model deployments with version control
Data and IndexesConnect data sources for RAG and fine-tuning
Agent BuilderBuild 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:

PatternDescriptionAzure Services
Smart chatbotNatural-language customer service botAzure OpenAI + Azure AI Search + Bot Service
Document processingAutomated data extraction from formsDocument Intelligence + Azure AI Search
Content moderationReview user-generated contentAzure AI Content Safety
Search and discoveryIntelligent search over org knowledgeAzure AI Search + Azure OpenAI
Real-time translationMultilingual communication platformAzure AI Speech + Azure AI Translator
Visual inspectionAutomated quality control from imagesAzure AI Vision + Custom Vision

The AI Application Stack (conceptual layers)

End-user application (web/mobile/chatbot/Copilot) -> orchestration layer (Prompt Flow, Semantic Kernel) -> 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 CopilotCustom application (Azure AI Foundry / Azure OpenAI)
Who builds itMicrosoft (it is a finished product)Your developers
How you get itBuy/enable a Copilot licenseProvision and code Azure AI services
Data groundingAutomatic via Microsoft GraphYou build the RAG pipeline (Azure AI Search)
CustomizationLimited to product settings and plugins/agentsFull control over models, prompts, data, and flow
Best whenYou want productivity in M365 apps fastYou 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.

Test Your Knowledge

How does Microsoft Copilot in Microsoft 365 access your organizational data to provide relevant responses?

A
B
C
D
Test Your Knowledge

What is Azure AI Foundry?

A
B
C
D
Test Your Knowledge

Which Copilot capability allows it to summarize a meeting you missed and list the action items?

A
B
C
D
Test Your Knowledge

A company wants a customer-service chatbot that answers naturally and is grounded in its own product manuals. Which combination of Azure services fits this pattern best?

A
B
C
D
Test Your Knowledge
Matching

Match each Microsoft Copilot application to what it can do:

Match each item on the left with the correct item on the right

1
Copilot in Outlook
2
Copilot in Excel
3
Copilot in PowerPoint
4
Copilot in Teams