What artificial intelligence models does Microsoft 365 Copilot use?

When we talk about Microsoft 365 Copilot, it's common to think that we're dealing with a single artificial intelligence integrated into Word, Excel, PowerPoint, Outlook, or Teams. However, the reality is much more interesting and far-reaching.

Microsoft 365 Copilot It has evolved into a multimodal artificial intelligence platform, capable of using different models and technologies depending on the task we want to perform.

This means that behind Copilot we don't just find 'an AI inside Microsoft 365.' There's a whole architecture that combines advanced AI models, the corporate data the user has access to, and the work context provided by Microsoft 365.

Modelos de inteligencia artificial de Microsoft 365 Copilot.

From a single model to a multimodel architecture in Microsoft 365 Copilot

The first versions of Microsoft 365 Copilot were closely tied to OpenAI's GPT models. However, Microsoft has gradually been expanding this architecture.

Currently, Microsoft 365 Copilot can work with models from different families and providers. Among them, we mainly find:

  • OpenAI GPT models
  • Claude Models from Anthropic
  • Models and technologies developed by Microsoft itself

The goal is simple: use the most suitable model for each type of work.

Not all tasks need the same level of reasoning. Summarizing an email, analyzing complex information, creating a presentation, or doing a job made up of multiple steps are different problems.

That's why the evolution of Copilot is aimed at selecting and using different AI capabilities depending on the context.

GPT-5.6: the preferred model of Microsoft 365 Copilot

One of the biggest changes has been the addition of OpenAI's GPT-5.6 as the preferred model for Microsoft 365 Copilot.

GPT-5.6 is available within different Microsoft 365 Copilot experiences, including:

  • Microsoft Word
  • Microsoft Excel
  • Microsoft PowerPoint
  • Copilot Chat
  • Copilot Cowork

Its main advantage is its greater ability to handle complex tasks and requests that require several steps of reasoning.

For example, instead of just asking Copilot to summarize a document, we can ask it to analyze information, identify specific aspects, compare them, draw conclusions, and finally create a structured document.

What does GPT-5.6 bring to Microsoft 365 apps?

In Word, it lets you work with more complete drafts, improve the structure of documents, and refine their writing.

In Excel, it improves Copilot's ability to reason about data and tackle more complex analysis.

In PowerPoint, it lets you create more detailed presentation drafts and improve both the content and the structure of the slides.

In Copilot Chat, it's especially useful for complex questions, comparing options, making plans, solving problems, and requests where there's some ambiguity.

But maybe one of the most interesting scenarios is found at Copilot Cowork.

Copilot Cowork: the evolution toward an AI capable of performing tasks

Modelos de inteligencia artificial de Microsoft 365 Copilot.

The evolution of the models is also changing the way we use Copilot.

User → Prompt → Response

With Copilot's new agentic capabilities, the model is starting to be different:

User → Goal → Planning → Execution → Result

Copilot Cowork is designed precisely for this second scenario.

The user describes what they need to get done and the system can plan different steps, use information available in its tools and files, and complete the work from start to finish.

The difference is important. We're no longer just talking about an artificial intelligence that tells us how to do a job, but about an artificial intelligence capable of taking part in actually doing it.

Claude Models from Anthropic in Microsoft 365 Copilot

Another very relevant change is the incorporation of models from Anthropic, the developer of the Claude family.

Microsoft has been incorporating these models into certain Microsoft 365 Copilot experiences, thus expanding the possibilities beyond OpenAI's models.

This confirms one of the most important trends we’re seeing in Microsoft’s strategy: Copilot is evolving into a multimodel platform.

In certain experiences, the system can use different models depending on the work it's doing. Microsoft has explained that Copilot Cowork can use OpenAI models for certain reasoning and research tasks, and Anthropic models for other tasks, including some visual work related to presentations and graphics.

For the business user, this has an important advantage: they don't need to become an expert in artificial intelligence models to get value from them. What really matters is the outcome.

Is Microsoft 365 Copilot just ChatGPT connected to Microsoft 365?

No.

This is probably one of the most common confusions when we explain Microsoft 365 Copilot.

GPT models are a key part of the solution, but Microsoft 365 Copilot is much more than just the language model that generates a response.

Artificial intelligence models + business context + Microsoft 365 applications and services

This is where Work IQ becomes especially important, which Microsoft uses to bring work context to Copilot experiences.

The goal is for artificial intelligence not to work only with general knowledge, but to be able to understand the user's authorized work context: documents, communications, meetings, and other elements related to Microsoft 365, always depending on existing permissions and controls.

This difference is fundamental in a business environment.

Business data: a key piece for Microsoft 365 Copilot

Modelos de inteligencia artificial de Microsoft 365 Copilot.

When an organization starts an artificial intelligence project, it's common for the first question to be: what AI model are we going to use?

However, there's another question that could be even more important: what information will AI be able to access, and how is it organized and protected?

An extremely advanced model adds little business value if the organization's information is fragmented, outdated, duplicated, or lacks a proper permissions structure.

On the other hand, when a company has its information properly organized in services like SharePoint Online, OneDrive, Teams, and Exchange Online, Copilot can provide a much more contextualized experience within the user's work environment.

For this reason, adopting Microsoft 365 Copilot shouldn't be seen solely as an AI implementation project. It also involves reviewing aspects such as:

  • The organization of the information.
  • SharePoint and OneDrive permissions.
  • Access to sensitive information.
  • The life cycle of documents.
  • The identity of the users.
  • The policies for security and information protection.

Before providing artificial intelligence to our data, we must make sure that our data is ready for artificial intelligence.

Does the user have to choose which AI model to use in Copilot?

Not necessarily.

This is precisely one of the advantages of the approach that Microsoft is adopting.

In certain scenarios, the user might have options related to the model or type of reasoning, but in other cases, it's the system itself that can determine which technology is most suitable.

The long-term goal doesn't seem to be that the user has to keep wondering which AI model to pick. The ideal experience would be just to say what they need to achieve and have Copilot figure out what capabilities, models, data, and tools to use to get the result.

AI Models and Agents: The Future of Microsoft 365 Copilot

Modelos de inteligencia artificial de Microsoft 365 Copilot.

The evolution of Microsoft 365 Copilot can't be analyzed just from the perspective of the models either. The next big change is in the artificial intelligence agents.

While a traditional model responds to a request, an agent can have instructions, information, tools, and the ability to carry out different actions aimed at achieving a goal.

This allows you to imagine numerous scenarios within a company:

  • An agent that analyzes corporate documentation.
  • An agent specialized in helping the sales department.
  • An agent who checks internal procedures.
  • An agent who prepares information before a meeting.
  • An agent that gathers information from different sources and puts together a report.
  • An agent that carries out multi-step processes using corporate tools.

Microsoft 365 Copilot, Copilot Studio, and Microsoft's new agentic capabilities are making these scenarios increasingly accessible for businesses.

What does the evolution of Microsoft 365 Copilot mean for businesses?

What does the evolution of Microsoft 365 Copilot mean for businesses?

The real power of the platform comes from the combination of advanced AI models, Microsoft 365, Work IQ, and business context, security and governance, and agents.

We're moving from asking 'what can artificial intelligence answer?' to wondering 'what work can it help us do within our company?'

Conclusion: Microsoft 365 Copilot evolves into a multimodal AI

Microsoft 365 Copilot is rapidly evolving from an assistant mainly based on GPT models into a multimodal and agent-based AI business platform.

The incorporation of GPT-5.6, the use of Anthropic's Claude models, and the evolution of experiences like Copilot Cowork clearly show this direction.

But for companies, the model used is just one part of the equation.

The real potential shows up when we combine advanced artificial intelligence models with Microsoft 365 data, proper identity and permission management, and a solid information security and governance strategy.

Is our company ready to take advantage of everything these models can do with our corporate knowledge?

Do you want to get your company ready to take advantage of Microsoft 365 Copilot?

ABD Consultoría

At ABD Consulting and IT Solutions We help companies get the most out of the Microsoft 365 ecosystem and incorporate artificial intelligence into their work processes.

As a Microsoft partner, we support you throughout the entire process: from preparing and organizing corporate data to setting up permissions, information security, and adopting tools like Microsoft 365 Copilot and Copilot Studio.

Because to take full advantage of the potential of artificial intelligence, it's not enough to just choose the right model. It's also necessary to have the right infrastructure, well-organized information, and a security strategy that ensures responsible use.

Do you want to know how to apply Microsoft Copilot in your company? Contact our team and we'll help you identify the opportunities that AI offers for your business.

Table of Contents

Follow us on LinkedIn
Subscribe to the Newsletter




    Labels