**Microsoft is bringing Anthropic’s AI into Office 365 apps while keeping its long-standing partnership with OpenAI. The decision signals a shift toward a multi-model strategy, allowing Microsoft to mix strengths from different vendors instead of relying on a single provider.**

Microsoft is preparing to change how artificial intelligence powers its Office 365 products. Reports [confirm](https://www.theinformation.com/articles/microsoft-buy-ai-anthropic-shift-openai) the company will begin using Anthropic’s Claude models for certain features in Word, Excel, PowerPoint, and Outlook, while still leaning heavily on OpenAI. 

![Microsoft Adds Anthropic AI Into Office 365](https://www.stanventures.com/news/wp-content/uploads/2025/09/ChatGPT-Image-Sep-10-2025-12_47_34-PM.avif)

The unusual aspect is that Microsoft will pay Amazon Web Services for access to Anthropic, despite AWS being a competitor in the cloud business.

 

> Throwing shade at OpenAI -> Sources: Microsoft will use Anthropic’s models for some AI features in Office 365 apps, after finding Claude Sonnet 4 outperforms GPT-5 in some tasks
> “The move will blend Anthropic and OpenAI technology in the apps, after years in which Microsoft… [pic.twitter.com/drVxxyty6r](https://t.co/drVxxyty6r)
> — Glenn Gabe (@glenngabe) [September 9, 2025](https://twitter.com/glenngabe/status/1965501943335293027?ref_src=twsrc%5Etfw)

 

## What Microsoft is Changing

Inside Microsoft, engineers tested Anthropic’s Claude Sonnet 4 against OpenAI’s models. 

According to people familiar with the process, Claude excelled in producing polished presentation slides and handling spreadsheet automation. That persuaded Microsoft to adopt a “best model for the job” approach. 

Some features will remain tied to OpenAI, while others will run through Anthropic. Over time, Microsoft will integrate models it develops in-house.

## A Broader Trend in AI

This decision doesn’t mean Microsoft is backing away from OpenAI. 

OpenAI remains central for advanced models, and Microsoft has invested billions in its growth. 

What’s happening is that tech companies are combining models from multiple suppliers to get the best performance in different areas. 

Anthropic’s models have been getting strong reviews for reasoning and creative tasks, but the company also faces lawsuits over how its training data was sourced. That’s a risk factor Microsoft and its customers will have to weigh.

## What Businesses Need to Know

If you manage IT or procurement, expect more vendor variety inside Office. That means governance and security teams will need to know which features run on which model, and where the data is actually processed. 

Microsoft has not announced price increases, but paying AWS for Anthropic access means costs are shifting under the surface.

Enterprises should prepare for sharper questions about where their subscription dollars go and how their data is handled.

## Steps Businesses Should Take Now

Organisations that rely on Office should start preparing early. Here are some practical moves to consider:

1. Ask Microsoft for clear documentation on which models power specific features and where those models are hosted.
2. Update internal policies to reflect that some Office functions will now run through AWS, in addition to Azure.
3. Run small pilot programs before rolling out features widely, so you can measure productivity gains and identify weaknesses.
4. Train employees on both the strengths and limits of these AI tools, reducing the risk of overreliance or blind trust.
5. Review contracts and service agreements carefully to ensure compliance, security, and data protection remain intact.

## Key Takeaways

- Microsoft is adding Anthropic’s Claude to Office 365 but keeping OpenAI in place.
- Claude Sonnet 4 impressed Microsoft with strong slide and spreadsheet output.
- Access will flow through AWS, with Microsoft paying Amazon for usage.
- Anthropic is growing fast but faces legal pressure around training data.
- IT leaders should double-check governance, contracts, and pilot results.