Skip to content
EN DE

M365 Copilot: AI Embedded in Your Workflow

L3 Lesson 4 of 5 — Context as Infrastructure
1
2
3
4
5

Claude Projects and Custom GPTs have something in common: You upload context — documents, instructions, rules. Microsoft 365 Copilot An AI assistant integrated directly into Microsoft 365 apps (Word, Excel, PowerPoint, Outlook, Teams) that accesses organizational data through the Microsoft Graph — emails, documents, calendars, chats. flips the principle: The context is already there. Your emails, documents, calendars, Teams chats — Copilot has access to your existing work life through the Microsoft Graph Microsoft's data layer that connects information across all Microsoft 365 services — emails, calendars, files, Teams chats, SharePoint documents. Copilot uses the Graph to ground responses in your work context. .

No uploading, no separate interface. AI where you already work.

M365 Copilot uses a three-layer model:

LayerWhat It ProvidesExample
Microsoft GraphYour organizational dataEmails, files, calendars, Teams chats
Web GroundingCurrent information from the webFact-checking, supplementation
Your PromptYour specific instruction”Summarize the last 5 emails about Project X”

The crucial difference: You don’t configure a system prompt. The integration itself is the context. Copilot knows who wrote to you, what you’re working on, and what meetings are coming up — because it sees the same data you do.

What Copilot can do: Generate drafts from bullet points, rewrite text, summarize, adjust tone.

Typical use: “Write a first draft based on the project brief in /SharePoint/Project-Alpha/Brief.docx”

Context source: The current document + referenced files from SharePoint/OneDrive.

What Copilot can do: Suggest formulas, analyze data, identify trends, create visualizations.

Typical use: “Which product category had the steepest decline in Q3?”

Context source: The current spreadsheet. Copilot in Excel remains the most limited — complex data analysis often requires manual follow-up.

What Copilot can do: Generate presentations from documents, suggest designs, restructure slides.

Typical use: “Create a presentation based on the quarterly strategy Word document”

Context source: Referenced Word or PDF files.

What Copilot can do: Summarize emails, draft replies, distill long threads.

Typical use: “Summarize this email thread and suggest a reply”

Context source: Email history, calendar, contacts.

What Copilot can do: Create meeting summaries, extract action items, search chat histories.

Typical use: “What was decided about timelines in the last project meeting?”

Context source: Meeting transcripts, chat histories.

What M365 Copilot Does Well — and What It Doesn’t

Section titled “What M365 Copilot Does Well — and What It Doesn’t”
  • Seamless integration: No tool-switching — AI is where you already work
  • Live context: No document uploads, no setup — your work context is automatically available
  • Enterprise-grade security: Data stays within the organization, compliance-ready
  • Cross-app context: Copilot can connect information from emails, documents, and chats
  • Price: $30/user/month for Business — a significant investment that needs to pay off
  • Lock-in: Only useful within the Microsoft ecosystem. If you don’t work with M365, Copilot isn’t an option.
  • Quality varies: Strong in Word and Outlook, still limited in Excel. Not every app benefits equally.
  • No custom tuning: You can’t fine-tune Copilot to your specific needs the way you can with custom instructions
  • Admin dependency: IT must enable and configure Copilot — you can’t just start using it yourself
  • Privacy awareness required: Copilot accesses all data you have access to. That includes documents you may have forgotten about.
AspectClaude ProjectsCustom GPTsM365 Copilot
Context sourceUploaded filesUploaded filesLive data (Graph)
Where you workClaude.aiChatGPTInside M365 apps
SetupUser creates projectUser creates GPTAdmin deploys
Context controlYou decide what goes inYou decide what goes inOrganization + automatic
StrengthFlexible workspaceShareable, API actionsSeamless workflow
WeaknessOnly on Claude.aiKnowledge extractableExpensive, lock-in
Ideal forProject work with documentsSpecialized tools to shareTeams in M365 environments

The question isn’t “which tool is best” — it’s “which fits your work context”:

  • You work primarily in Microsoft 365? → Copilot is the most natural way to bring AI into your workflow. Check whether the price justifies the value.
  • You need a flexible workspace for project work? → Claude Projects give you the most control over context and quality.
  • You want to build and share specialized tools? → Custom GPTs are the most accessible format for that.
  • You use multiple tools? → That’s normal, and often the smartest approach. Copilot for daily workflow, Claude Projects for deep work, Custom GPTs for specialized tasks.

Take stock: Where does the context you need daily actually live? Emails, SharePoint, local files, notes? Which of the three tools (Projects, GPTs, Copilot) would have the easiest access?

Take three recurring tasks from your work week. Assign each to the best-fit tool: Which would benefit from Copilot? Which from a Claude Project? Which from a Custom GPT?

If you have access to M365 Copilot: Test it consciously for one week. Log every use: What saved time? What was disappointing? Calculate: Is $30/month worth it for you?

M365 Copilot shows where things are heading: AI that doesn’t live in a separate interface but is embedded in the tools you already use. Context isn’t configured — it emerges from your work.

But regardless of the tool, one skill remains central: Writing good instructions. Whether you define custom instructions in a Claude Project, configure a Custom GPT, or simply prompt better — the principles are the same. In the next lesson, you’ll learn the anatomy of a strong system prompt — the synthesis of L3.

Part of AI Learning — free courses from prompt to production. Jan on LinkedIn