Copilot Explainers · 3 min read

AI models behind Microsoft 365 Copilot: a plain-English guide for business leaders

James Wilkinson

By James Wilkinson 30 April 2026 Updated 3 October 2026

A plain-English guide to the AI models powering Microsoft 365 Copilot, plus the grounding, Microsoft Graph and setup that decide how useful Copilot really is.

TL;DR
  • The model is only one part of Microsoft 365 Copilot. The real business value comes from how Copilot uses your work context.
  • Copilot grounds responses in Microsoft 365 data the user is already allowed to access, so permissions and content quality matter.
  • Leaders should focus less on model names and more on readiness: clean files, clear ownership, sensible governance and useful workflows.

If you are weighing up Microsoft 365 Copilot, it is tempting to start with the AI models powering Microsoft 365 Copilot. Is it using GPT? Is it using Claude? Is it the latest version? Is it better than the tool someone tried last month?

Those questions are understandable, but they are not the whole story. For most businesses, Copilot’s value is less about a single model name and more about the system around the model: your Microsoft 365 apps, your data, your permissions and the way staff ask for help.

This article is about the leadership question, and it deliberately avoids a list of model names. If you want the current models in each part of Copilot, how each is billed and the admin settings behind them, see our living list of the models behind Copilot, which we keep up to date.

The plain-English version

A large language model is the part that understands and generates language. It can draft, summarise, compare, explain and reason across information.

Microsoft 365 Copilot adds something important on top: work context. When a licensed user asks Copilot a question inside Microsoft 365, Copilot can use information from the user’s Microsoft 365 world, such as documents, emails, meetings and Teams messages they are allowed to access. That is what makes it different from a general chatbot on the open web.

This does not mean Copilot can see everything in the business. It should respect the permissions already in place. But it does mean your permission model suddenly matters much more, because Copilot can make buried information easier to surface.

Why model names can distract leaders

Model capability matters, but model names change quickly. Between late May and the end of September 2026 Microsoft added at least ten new OpenAI and Anthropic models to Copilot, plus Grok in preview. It can update the underlying models, add new options and tune Copilot experiences without most users needing to understand every technical detail. By default a setting called Auto picks the model for each request, so most people never need to choose.

There is one model question leaders do need to own, and it is about money rather than capability. Since September 2026 the everyday models sit in the Copilot licence, while the heaviest frontier models and work handed to Cowork are paid in Copilot Credits. We set out that split in Copilot model choice: what the licence covers and what costs Copilot Credits.

The business question is more stable:

  • Can staff get useful answers from the information your organisation already holds?
  • Can they create better first drafts without losing professional judgement?
  • Can they summarise meetings, emails and documents without exposing information to the wrong people?
  • Can the organisation govern the tool well enough to use it confidently?

If the answer to those questions is no, chasing the newest model will not solve the real problem.

Grounding is where the work begins

Grounding means giving the model relevant context before it generates an answer. In Microsoft 365, that context can come from the user’s files, chats, meetings, emails and other workplace signals.

For example, a manager might ask Copilot to prepare a project update. A generic chatbot can help with structure and wording. Microsoft 365 Copilot can be more useful if it can also draw on recent meeting notes, project documents and email threads that the manager can already access.

That is powerful, but it also exposes weak foundations. If the project folder is outdated, if decisions are buried in private chats or if too many people have access to sensitive files, Copilot will reflect those realities back to the user.

What good readiness looks like

Before focusing on advanced prompts, leaders should ask practical questions about the Microsoft 365 estate:

  • Are key files stored in the right SharePoint sites and Teams?
  • Are old, duplicate and misleading documents being cleaned up?
  • Do permissions match how the business actually works?
  • Are sensitive client or employee documents labelled and protected?
  • Do teams know where final versions live?
  • Are meetings captured in a consistent way?

This is not glamorous work, but it is what turns Copilot from an impressive demo into a useful assistant.

What to tell staff

Staff do not need a lecture on model architecture. They need simple guidance:

  • Use Copilot to create a first draft, not a final answer.
  • Ask it to show assumptions and source context where possible.
  • Check anything that affects a client, colleague, number, deadline or decision.
  • Do not use Copilot to work around permissions.
  • If the answer looks wrong, check the source material before blaming the tool.

That guidance is easier to follow when leaders have already cleaned up the data and agreed what safe use looks like.

The leadership takeaway

Microsoft 365 Copilot is not just a model bolted onto Office. It is a work system that uses AI, Microsoft 365 apps, permissions and organisational context together.

So the right question is not “which model is it using today?” The better question is “is our Microsoft 365 environment ready to give any capable model the right context?”

If the foundations are strong, model improvements become useful. If the foundations are weak, every new capability simply makes the gaps more visible.

Sources checked

Last checked: 3 October 2026.

Related reading

More on Copilot Explainers

Copilot Explainers The Copilot model picker nobody switches: when to leave it on Auto and when to choose GPT or Claude Microsoft Copilot has a model picker most teams never touch. When to leave it on Auto, when to choose a GPT or Claude model, what it costs and why the work stays under your Microsoft agreement. Copilot Explainers What AI models does Microsoft Copilot use? The current list, by surface Which models power Microsoft Copilot (formerly Microsoft 365 Copilot) right now in Chat, Word, Excel, PowerPoint, Cowork and Copilot Studio, how each is billed and the admin and UK data points. Kept up to date. Copilot Explainers Copilot Code and managed runtime: anyone in your firm can build an app now. Who owns it? Copilot Code lets anyone describe an app or automation and get one, running on Copilot Managed Runtime inside your tenant. What it is, what it costs and where Copilot Studio still fits. Copilot Updates GPT-5.6 arrives in Microsoft 365 Copilot: what it changes for UK businesses GPT-5.6 began rolling out across Microsoft 365 Copilot, including Cowork, in July. What it changes about model choice for firms that keep client work inside Microsoft. Copilot Updates Claude Opus 5 in Microsoft 365 Copilot: should UK businesses switch it on? Claude Opus 5 is live in the Microsoft 365 Copilot model selector. When it beats GPT-5.6, why UK tenants have it switched off and what enabling it involves. Copilot Licensing Copilot model choice: what the licence covers and what now costs Copilot Credits Microsoft Copilot (formerly Microsoft 365 Copilot) is now split into everyday AI on the licence and advanced AI on Copilot Credits. Which models sit where, the 1 December change for Copilot Business and what a firm will pay for. How we work The FiveForward Framework Five steps from a free Copilot Readiness Check to agents kept running: Check, Plan, Build, Handover, Run. The install Firmwide A fixed-price install, typically six to eight weeks: Copilot made safe and useful in your Microsoft 365, an agent and a skills pack for every department, handed over with training. Then Run keeps it working. Next step Book a Copilot Readiness Check Free, 30 minutes, online: where Copilot can and cannot see your data, the two jobs to automate first and a rough install price.

Common questions

Questions about AI models powering Microsoft 365 Copilot

Does Microsoft 365 Copilot use one AI model?
No. Microsoft describes Copilot as an orchestration system that can use large language models together with Microsoft 365 apps, Microsoft Graph and other work context. The exact model mix can change over time.
Why does Microsoft Graph matter?
Microsoft Graph is the layer that helps Copilot understand the user's work context, including things like emails, chats, meetings and files the user has permission to access.
What should business leaders check first?
Check whether important content is accurate, findable and permissioned correctly. A stronger model will not fix messy folders, unclear ownership or over-broad access.