Skip to main content

O6: Copilot Ecosystem

"Copilot" is a brand and product family, not a single product. Microsoft has embedded Copilot across its entire stack โ€” from code editors to security operations. This module maps the ecosystem, explains the shared architecture, and shows how to extend Copilots with custom capabilities. For AI agent patterns that Copilots build on, see O2: AI Agents Deep Dive. For the platform hosting Copilot models, see O4: Azure AI Foundry.

Core Architectureโ€‹

Every Microsoft Copilot shares the same foundational pattern:

โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”
โ”‚ Copilot Experience โ”‚
โ”‚ (Chat UI, inline suggestions, side panel) โ”‚
โ”œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ค
โ”‚ Orchestration Layer โ”‚
โ”‚ (Prompt construction, tool routing, safety) โ”‚
โ”œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ฌโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ฌโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ค
โ”‚ LLM โ”‚ Grounding โ”‚ Product โ”‚
โ”‚ (GPT-4o)โ”‚ Data โ”‚ Integration โ”‚
โ”‚ โ”‚ (Graph, code,โ”‚ (Editor, Office, โ”‚
โ”‚ โ”‚ docs, web) โ”‚ Azure portal) โ”‚
โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ดโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ดโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜

The grounding data is what makes each Copilot different โ€” same model, different context.

The Copilot Stackโ€‹

LayerWhat It DoesComponents
User LayerSurface Copilot in the product UXChat panel, inline completions, side panes
OrchestrationConstruct prompts, route tools, enforce safetySemantic Kernel, Prompt Flow, content filters
Data & IntelligenceGround responses in real dataMicrosoft Graph, Azure AI Search, code index
FoundationCore reasoningAzure OpenAI (GPT-4o, GPT-4o-mini)

Copilot Comparisonโ€‹

CopilotDomainGrounding DataLicensing
M365 CopilotProductivity (Word, Excel, Teams, Outlook)Microsoft Graph (emails, files, calendar, chats)$30/user/month add-on
GitHub CopilotSoftware developmentCode repositories, docs, web$10โ€“39/user/month
Copilot for AzureCloud operationsAzure resources, docs, Monitor logsFree (in Azure Portal)
Copilot StudioCustom copilot builderCustomer data sources (configurable)Per-message pricing
Security CopilotSecurity operationsSentinel, Defender, threat intelligencePer SCU (Security Compute Unit)
Copilot in WindowsDesktop assistantLocal files, web, settingsIncluded with Windows 11
Copilot in Dynamics 365CRM/ERPDataverse, customer recordsIncluded with Dynamics license

:::info Key Insight Every Copilot uses the same underlying LLM (GPT-4o). The differentiation is grounding โ€” M365 Copilot is grounded in your Graph data, GitHub Copilot is grounded in your code, Security Copilot is grounded in threat intelligence. :::

Microsoft 365 Copilotโ€‹

CapabilityWhat It Does
WordDraft documents, summarize, rewrite with tone control
ExcelAnalyze data, generate formulas, create charts from natural language
PowerPointGenerate presentations from Word docs or prompts
OutlookSummarize email threads, draft replies, prioritize inbox
TeamsMeeting summaries, action items, catch-up on missed meetings
Business ChatCross-app queries grounded in all your M365 data

Grounding: Microsoft Graph โ€” your emails, files, calendar, chats, contacts. Copilot sees what you have access to (respects existing permissions and RBAC).

GitHub Copilotโ€‹

GitHub Copilot has evolved from code completion into a full development platform:

FeatureWhat It Does
Code CompletionInline suggestions as you type (single line and multi-line)
Copilot ChatNatural language Q&A about code in the editor
Copilot EditsMulti-file code changes from natural language descriptions
Copilot AgentsAutonomous coding agents for issues and PRs
CLI AssistantTerminal command suggestions and explanations
ExtensionsThird-party tools integrated into Copilot Chat

Grounding: Your open files, repository context, language documentation, web search.

# GitHub Copilot tiers
Individual: $10/month โ€” code completion + chat
Business: $19/month โ€” + org policies, audit logs
Enterprise: $39/month โ€” + knowledge bases, fine-tuning
tip

GitHub Copilot's agent mode (@workspace) can autonomously plan, implement, and test code changes across multiple files. For agent patterns beyond code, see O2: AI Agents Deep Dive.

Copilot for Azureโ€‹

Available directly in the Azure Portal โ€” no additional cost:

CapabilityExample
Resource management"Show me all VMs that are stopped"
Troubleshooting"Why is my App Service returning 503?"
Cost analysis"What are my top 5 cost drivers this month?"
IaC generation"Generate Bicep for a web app with SQL backend"
KQL queries"Write a query to find failed requests in the last hour"

Grounding: Azure Resource Graph, Azure Monitor, Azure documentation.

Copilot Studioโ€‹

Copilot Studio is the low-code platform for building custom copilots:

FeatureDescription
TopicsConversation flows โ€” trigger phrases โ†’ actions โ†’ responses
Generative answersGround responses in your data (SharePoint, websites, files)
PluginsExtend with custom actions (Power Automate, HTTP connectors)
ChannelsDeploy to Teams, web, Slack, custom apps
AnalyticsSession tracking, resolution rates, escalation metrics

Best for: customer support bots, HR assistants, IT help desks โ€” where non-developers need to build and maintain AI experiences.

Security Copilotโ€‹

Purpose-built for security operations:

CapabilityGrounding Source
Incident investigationMicrosoft Sentinel alerts
Threat analysisMicrosoft Defender Threat Intelligence
Vulnerability assessmentDefender for Cloud
Script analysisReverse engineering, malware detection
Report generationCross-product security data

Extensibilityโ€‹

Pluginsโ€‹

Plugins extend Copilot's capabilities with custom actions:

{
"schema_version": "v1",
"name": "contoso-crm",
"description": "Look up customer information in Contoso CRM",
"functions": [{
"name": "get_customer",
"description": "Get customer details by name or ID",
"parameters": {
"customer_id": { "type": "string", "description": "Customer ID or name" }
}
}]
}

Declarative Agentsโ€‹

Custom Copilot personas with specific knowledge, instructions, and tool access โ€” deployed as M365 Copilot extensions:

ComponentPurpose
InstructionsSystem prompt defining persona and behavior
KnowledgeGrounding sources (SharePoint, Graph connectors)
ActionsPlugins the agent can invoke
Starter promptsSuggested conversation starters

Extension Typesโ€‹

TypeWhat It IsBuilt WithDeployed To
PluginSingle action/APIOpenAPI specM365, Copilot Studio
Declarative AgentCustom persona + knowledge + actionsTeams ToolkitM365 Copilot
Copilot Studio AgentFull custom copilotCopilot StudioTeams, web, Slack
GitHub Copilot ExtensionChat participant in VS CodeGitHub App + APIGitHub Copilot Chat

Key Takeawaysโ€‹

  1. Copilot = brand, not product โ€” same LLM, different grounding data per domain
  2. The Copilot Stack: User Layer โ†’ Orchestration โ†’ Data & Intelligence โ†’ Foundation
  3. M365 Copilot is grounded in Microsoft Graph (your org's data)
  4. GitHub Copilot has evolved into a full development platform with agents
  5. Copilot Studio enables non-developers to build custom copilots
  6. Extend any Copilot via plugins (actions), declarative agents (personas), or Copilot Studio (full custom)
  7. Security, cost, and grounding data sources are the key differentiators across the family