For the past few years, Microsoft Copilot meant one thing: OpenAI. GPT under the hood, Microsoft on the label. That is changing faster than most people realize, and the shift has real implications for every business running inside the Microsoft 365 ecosystem.
Microsoft spent years building Copilot on a single foundation: OpenAI. GPT under the hood, Microsoft on the label, and an exclusive partnership that gave them a head start on every competitor. That arrangement made sense in 2023 when OpenAI was the clear leader and Microsoft needed to move fast. It does not make as much sense in 2026, and Microsoft has been quietly dismantling it piece by piece.
In September 2025, Microsoft added Anthropic's Claude Sonnet 4 and Claude Opus 4.1 to Microsoft 365 Copilot. Initially an opt-in feature requiring a separate agreement with Anthropic, it was upgraded in January 2026 when Anthropic became a Microsoft subprocessor. That change means Claude now runs under Microsoft's own data protection terms, is enabled by default for most commercial tenants in the US, and requires no additional procurement or configuration from your IT team. It is just there, available inside the tools your employees already use.
Claude is currently live inside the Researcher agent for complex multi-step research tasks, Copilot Studio for building and managing custom AI agents, and as a preview in Agent Mode for Excel and PowerPoint. xAI's Grok is also available inside Copilot Studio now, meaning Microsoft's platform supports models from three separate AI companies simultaneously.
The signal most people missed came from GitHub Copilot. Microsoft's coding tool with tens of millions of users now primarily defaults to Claude Sonnet 4 for automatic model selection in Visual Studio Code. Coding was supposed to be OpenAI's strongest territory. The fact that Microsoft is routing its developer base to Claude for automatic selection is a louder statement than anything in a press release.
On the OpenAI side, the partnership did not collapse. It got restructured. Microsoft and OpenAI signed a new definitive agreement in October 2025 after months of visible tension. OpenAI completed its conversion to a public benefit corporation, with Microsoft holding a 27% stake valued at roughly $135 billion. Microsoft locked in 20% of OpenAI's total revenue through 2032 and extended its IP rights through the same period. The financial entanglement is deeper than ever.
But the exclusivity is gone. OpenAI can now use other cloud providers. Microsoft can independently pursue AI development with other partners. OpenAI committed to purchasing $250 billion in Azure services, but that is a commercial arrangement, not a lock-in. Both companies are hedging at the same time, which is exactly what you would expect from two organizations that have grown large enough to have diverging interests.
What Microsoft is building underneath all of this is not really an AI product anymore. It is an AI infrastructure layer. The strategy is to always have the best model available for any given task, all delivered through one interface your team already knows. OpenAI for general Copilot Chat. Claude where deep reasoning, extended document work, or coding gives it an edge. Other models as the market evolves. Microsoft does not need to win the model race. It needs to be the platform that routes to whoever is winning at any given moment, and it is building exactly that.
The $650 billion that Alphabet, Amazon, Meta, and Microsoft are collectively spending on AI infrastructure in 2026 is the physical foundation that makes this possible. Data centers, AI chips, networking equipment, power infrastructure. Microsoft's piece of that investment is what keeps Copilot improving at the pace it has been, and what makes it realistic for them to add, swap, and upgrade models without disrupting the end user experience at all.
For businesses already inside Microsoft 365, none of this requires a new decision. The improvements arrive through the tools you already pay for. The Copilot your team opened this morning is materially more capable than the one they used twelve months ago, and that trajectory is not slowing down.
The Wyecliff Perspective
The Copilot your team is using today is not the same product it was twelve months ago. It is materially more capable, particularly for the kind of work that happens inside Word, Excel, and PowerPoint. Contract review. Financial modeling. Report generation. Presentation drafting. These are not experimental AI use cases anymore.
The Microsoft-OpenAI relationship turning rocky is actually good news for buyers. When Microsoft needed OpenAI exclusively, there was no competitive pressure driving improvement. Now that Microsoft is routing workloads to Claude where Claude performs better, you benefit from that competition without changing a single tool or subscription.
We work with a lot of mid-market companies that are already paying for Microsoft 365 Copilot but using maybe 20% of what it can do.
The gap between what they are using and what is available has never been wider. Find out how to close that gap.
One Thing To Try This Week
Open Microsoft Copilot and try one of these prompts against your Outlook inbox or calendar. You do not need to set anything up. Just paste it in and see what comes back.
For your inbox: "Search all my emails from the last week and create an action list of anything I may have missed or still need to respond to."
For your calendar: "Give me a rundown of my meetings this week. For each one, summarize who I am meeting with and flag anything I should prepare or follow up on."
To go deeper: "Find all emails from [client name] in the last 30 days and summarize where things stand."
The point is not that these prompts are perfect. The point is that Copilot is already connected to your inbox, your calendar, and your files. Most people are using it like a search bar when it is closer to an assistant that has already read everything and is waiting to be asked.
Start with one. See how much time it saves. Then think about how many people on your team are still doing that same task manually every Monday morning.
What we are seeing more and more is companies using Copilot as their first real step into AI. Not a massive custom build, not a six figure implementation project, just getting their employees actually using the tools they already have in ways that make their day easier. It is a low commitment entry point that delivers real value fast, and it is how a lot of our client relationships start.
The interesting thing that happens once people start using it is that the requests get bigger. Someone figures out how to clean up their inbox on Monday morning, and by the following month they are asking whether AI can automate their proposal process or flag contract risks before they sign. The vision expands because the comfort level expands. People stop asking "what can AI do" in the abstract and start asking "what else can it do for me specifically."
Copilot gets companies through the door. The more advanced problems that surface once employees are actually using AI day to day, those are the ones that turn into custom builds, workflow automation, and the kind of solutions that deliver significant return on investment. We have seen it happen with clients across industries, and the pattern is consistent. Start simple, build confidence, uncover the real problems worth solving.
If your team is not using Copilot yet, that is the first conversation worth having. If they are already using it and you are starting to see bigger opportunities emerge, that is the second one.
Either way, we would like to hear what you are running into.
Learn more here: wyecliff.ai
