Microsoft 365 Copilot now includes Anthropic Claude: what it means for business AI

The original announcement: Available today: Anthropic Claude Opus 4.7 in Microsoft 365 Copilot, on the Microsoft 365 Copilot blog. Microsoft also published a comparison piece worth reading: Microsoft 365 Copilot vs Claude Enterprise. Both are cited throughout this post. Credit for the announcement sits with Microsoft and Anthropic.

What was announced

Microsoft has made Anthropic’s Claude Opus 4.7 available inside Microsoft 365 Copilot. Until now, Copilot has run on OpenAI’s models, configured and operated inside the Microsoft enterprise-AI envelope. The change is that Claude is now one of the model options behind Copilot, available to enterprise and business customers using the platform.

The practical effect: a Copilot user can now ask for the same task to be handled by a different underlying model when the task suits it better. The selection happens inside Copilot. The data never leaves the Microsoft 365 tenant boundary the customer already has in place.

Why this is more than a vendor announcement

Anthropic and OpenAI are the two most capable AI labs in the market today, and they have distinctly different strengths. Claude is widely considered the stronger model for long-context reasoning, thoughtful writing, and complex coding tasks. The OpenAI models in Copilot have been the fastest and most consistent for general knowledge work and tightly integrated app features. Until this announcement, you had to pick one platform or the other, with separate logins, separate data flows, and separate security postures.

Copilot adding Claude means the choice between "OpenAI" and "Anthropic" is no longer a choice between two separate enterprise contracts and two separate security models. It is now a choice between two models inside the same governed environment, with the same tenant boundary, the same permissions, the same sensitivity labels, the same audit trail.

Why the security envelope matters more than the model

Microsoft published a comparison piece making the case that Microsoft 365 Copilot, by virtue of being inside Microsoft 365, offers things a separate Claude Enterprise subscription cannot:

• Native app integration, with the AI inside Outlook, Word, Excel, PowerPoint, Teams and SharePoint, rather than as a separate workspace with its own file copies
• Microsoft Purview policy enforcement, with sensitivity labels, retention rules and DLP applied to AI interactions automatically
• Tenant-bound data handling, with no copies of prompts, inputs or responses leaving the customer’s Microsoft 365 boundary
• Long-running, multi-step agentic work using the platform’s Work IQ layer, rather than chat-only interaction

The comparison is, naturally, written from Microsoft’s position. Claude Enterprise on its own is a strong product. The argument worth taking seriously is not that Claude is less capable than OpenAI. It is that running Claude inside Copilot keeps the model choice without losing the governance, while running Claude Enterprise as a separate vendor relationship means two different policy environments to manage.

For an SMB that is already governing its data inside Microsoft 365, the second option is usually a step backwards on security posture, not forwards.

What it means in practice

For Agile IT clients using Copilot through AgileAI, the change is operationally simple and strategically useful:

• Same licence, more capability. Copilot subscribers gain access to Claude without a separate Anthropic contract or rollout project
• Better task-fit. The model that is best for a long-context summary of a contract is not necessarily the model that is best for drafting a marketing email. Now the user can choose without leaving the workflow
• Same governance. Sensitivity labels, retention policies, audit logging and permissions still apply, regardless of which model is in use behind the scenes
• Less shadow AI risk. One of the strongest reasons staff drift into using consumer Claude or ChatGPT accounts has been a perception that Copilot is "not the same". With Claude inside Copilot, that argument is weaker, and the risk of sensitive data going into unsanctioned AI tools drops

The honest caveat

Two things worth saying clearly:

1. Model availability is not the only thing that matters. Copilot adoption still depends on the foundation work, sensitivity labels, sharing policies, training, and the AI usage policy that tells the team how to actually use the tool. A new model option does not skip those steps.

2. Microsoft’s comparison piece is marketing, written to keep customers inside Microsoft 365. The points it makes about governance and integration are accurate. The way it frames Claude Enterprise is one-sided. For Australian SMBs already running Microsoft 365, the practical conclusion still holds. For businesses with substantial workflows outside the Microsoft world, the picture is less clear, and worth thinking through carefully.

How we think about it inside AgileAI

Agile IT deploys, configures and supports Microsoft 365 Copilot for our clients as a managed capability inside the AgileAI service. The Claude availability is a positive development for that work, not a rethink of it. The shape of a deliberate Copilot rollout, foundation check, pilot group, policy and training, wider rollout, ongoing review, does not change because of a new model option. It just gets a more capable Copilot to roll out to.

For businesses still on the fence about whether to bring AI into the work, the practical answer keeps getting clearer. Microsoft 365 Copilot is now the most capable, best-governed business AI available to SMBs running on Microsoft 365, with both major model families behind it and the same security envelope around all of them.