What it does, what changed in 2026, and how it compares to Claude and ChatGPT. A practical guide, not a product brochure.
If your business runs Microsoft 365, there is a good chance CoPilot has already landed in your apps. Whether you are actively using it or ignoring the little icon in your sidebar, here is what it does.
CoPilot is AI built directly into the Microsoft 365 apps you already use. It is not a separate tool. It lives inside:
Draft documents from a prompt, rewrite sections, summarise long docs
Analyse data, generate formulas and charts from plain English
Create decks from a brief, redesign layouts, apply branding
Summarise email threads, draft replies, triage your inbox
Meeting summaries, action items, catch-up recaps
Summarise notes, generate study guides, organise research
The key difference from ChatGPT or Claude: CoPilot can see your emails, calendar, files, and Teams conversations. It is grounded in your organisation's data via Microsoft Graph, not just the open web.
CoPilot has moved from "helpful sidebar" to something closer to an active collaborator. The major updates this year, sourced from the Microsoft 365 Blog and HubSite365:
CoPilot now edits iteratively rather than giving one-shot responses. It can refine a document across multiple steps while showing what it changed.
Users can point CoPilot at specific SharePoint sites and lists, so responses draw from your organisation's structured data rather than guessing.
Choose from speaker summaries, executive summaries, or design your own template. Save templates for reuse across all meetings.
A centralised workspace where you can gather Word docs, Excel files, PDFs, and notes in one place, then ground AI agents on that collection.
As of January 2026, Microsoft reports 15 million paid CoPilot licences across enterprise and business customers worldwide.
These are not competitors. They are different tools for different jobs.
| CoPilot | Claude | ChatGPT | |
|---|---|---|---|
| Best for | Working inside M365 apps with your org's data | Long-form writing, analysis, structured reasoning, coding | General chat, image generation, browsing, plugins |
| Data access | Your emails, files, calendar, Teams via Microsoft Graph | What you paste in or upload | What you paste in, upload, or connect via plugins |
| Where it lives | Inside Word, Excel, Outlook, Teams, PowerPoint | Web app, API, Claude Code | Web app, API, mobile app |
| Strengths | Context across your entire M365 environment without copy-pasting | Longer context windows, nuanced writing, complex reasoning | Broad ecosystem, image generation, GPT store |
| Weakness | Output quality depends heavily on how you prompt it | No direct M365 integration | No direct M365 integration |
| Free tier | CoPilot Chat (basic, web-grounded) | claude.ai (usage limits) | chatgpt.com (usage limits) |
The honest take: most office workers would benefit from learning to prompt CoPilot well first (since it is already in their apps), then using Claude or ChatGPT for tasks that need deeper reasoning, longer documents, or work outside the Microsoft ecosystem.
Most organisations that deploy CoPilot see the same pattern: initial excitement, a few weeks of experimentation, then a drop-off because people do not know how to prompt it effectively. They type vague requests, get vague results, and conclude "AI does not work for my job."
The fix is not more features. It is structured prompting. The same principles work across every AI platform:
Give the AI a role and audience. Provide context about what you are working on. Specify the output format you want. Set constraints and quality criteria. Include examples of what good looks like.
These skills work in CoPilot, Claude, ChatGPT, and whatever comes next. That is why we focus on the skill, not the tool.
If you want to see CoPilot in action before trying it yourself, these are the best walkthroughs we have found.
13 free courses on Claude including Claude 101 and AI Fluency
Pre-built prompts organised by role and app
Before-and-after comparisons across finance, HR, ops, and executive roles
Why the skill behind prompting matters more than the tool you use
We help teams get more from the AI tools they already have. Face to face, with your real workflows.