NON-Microsoft models in Copilot: what changes for your company

Updated on October 6, 2026

Microsoft is quietly removing OpenAI and Anthropic from Copilot and replacing them with its own internal models. If your company uses Excel or Outlook with AI, the engine under the hood has already changed — and no one has communicated it to you.

Is it still worth relying on Copilot after the model change?

It depends on what you use it for. For structured tasks like Excel formulas and email summaries, MAI models hold up well. For high-reasoning activities — complex analysis, persuasive texts, strategic briefs — the answer is still open.

What really happened: did Microsoft change the Copilot engine?

Yes. Microsoft has quietly started a transition: OpenAI and Anthropic AI models are being progressively replaced with proprietary MAI (Microsoft AI) models inside products like Excel and Outlook.

According to reports by The Decoder , tens of thousands of weekly queries are already passing through the new internal models. This move wasn't announced to end users with any dedicated communication.

Mustafa Suleyman, CEO of Microsoft AI, stated explicitly to Bloomberg: «Anthropic is extremely expensive and I think many people are urgently looking for alternatives. We pay a lot of money to Anthropic, so our goal is to reduce and ultimately eliminate that cost.» It is reported by SiliconANGLE .

The direction is clear: reduce reliance on third-party providers and lower the cost per AI interaction. For companies paying for a Copilot subscription, this means the model processing their requests can change without notice.

Are MAI models really on par with GPT-4o and Claude?

On structured tasks yes, for complex reasoning it's still too early to say for sure. MAI models are designed for speed and cost, not for processing depth.

At Build 2026, Microsoft unveiled 7 new AI models. The most advanced is MAI-Thinking-1: 35 billion active parameters, a 256,000 token context window (meaning it can process very long documents at once), and trained entirely on commercially licensed data, without using output from OpenAI or other third-party models. This confirms Windows Central .

OpenAI models — particularly GPT-4o — and Anthropic models like Claude 3 were chosen in the past for their complex reasoning and natural language handling capabilities. LLMs seem to prioritize speed and cost per query over processing depth.

This is not a neutral detail. Those who use Copilot for high cognitive-value activities — qualitative data analysis, persuasive text generation, complex document summaries — might notice differences in the outputs.

How many companies already use Copilot and how widespread is it in Italy?

Copilot surpassed 30 million paid licenses in July 2026. In Italy growth has been fast, but actual adoption among license holders remains low.

According to Italian Industry Microsoft 365 Copilot surpassed 30 million paid licenses in July 2026, with new licenses more than doubling quarter-over-quarter. The number of companies with over 50,000 licenses increased more than 7x year-over-year.

In Italy, always Italian Industry reports a 70% increase in enterprise clients and a doubling of daily users quarter over quarter. Globally, 70% of Fortune 500 companies have already integrated the technology into their processes.

The figure that should worry you the most, however, is another one: only 35.8% of employees with access to the license actively use Copilot, according to Avantiico . Less than 4 workers out of 10 with a license actually use it.

What actually changes for everyday Copilot users?

Anyone who has integrated Copilot into their workflows might notice quality variations in the outputs, even though Microsoft hasn't promised guaranteed equal performance.

The issue is not just technical. Companies pay a Copilot subscription without knowing precisely which model processes their requests. The quiet swap of models raises an issue of AI governance — meaning control and responsibility over the use of AI tools in the company — which IT managers and marketing managers cannot ignore.

For those using Copilot in activities related to advanced Copilot features like agents and coding , the underlying model switch can alter behaviors that previously seemed stable.

Three concrete scenarios: how does the model switch impact an Italian company?

The impact varies a lot depending on how you use Copilot. Here are three common situations and what they mean for your work.

Illustrative examples: the numbers and situations are just to show how it works, they are not real customer data.

Why did Microsoft make this move and what is behind it?

It is both a financial and strategic choice together. Microsoft wants to reduce its dependence on third-party suppliers and lower the marginal cost of every AI interaction, but the reputational risk is real.

Last year, the AI investment race put pressure on the margins of all major players. Microsoft, which has invested billions in OpenAI, now has to balance operating costs with shareholder expectations. Bringing models in-house is a financially rational response: VentureBeat reports a claimed savings of up to 89% on API costs thanks to hardware-level optimizations on the Azure infrastructure.

Reputational risk, however, is real. Copilot was sold as a tool powered by the best models available on the market. If the perception of quality were to drop, companies might reconsider adoption or reduce purchased seats. Competitors — from Google with Gemini to Salesforce with Einstein AI — could leverage this moment to position themselves as more transparent alternatives.

It is also worth following the evolution of the other models in play: Anthropic is building an independent go-to-market strategy which could make it less dependent on Microsoft as a distribution channel.

Why does the Italian AI market make this news even more urgent?

The Italian AI market is growing fast, but the majority of SMEs are still standing still. Those who already have Copilot need to figure out what they are actually using.

According to Artificial Intelligence Observatory of the Polytechnic University of Milan , in 2025 the Italian AI market reached 1.8 billion euros, growing by +50% compared to 2024. 84% of large Italian companies bought Generative AI licenses, with Microsoft Copilot among the most popular solutions.

The picture in SMEs is the opposite. According to the 2025-2026 Research by the Digital Innovation in SMEs Observatory from the Politecnico di Milano, 76% of Italian SMEs have not invested or do not plan to invest in AI, and only 7% have started structured training programs. Those who already have Copilot, often without a clear strategy, are exposed to the risk of paying for a tool they don't use or use incorrectly.

How to get started concretely: what to do in the next 30 days?

Before waiting for official announcements from Microsoft, there are five things you can do right away using internal resources.

What are the most common mistakes companies are making right now?

The main one is not noticing the change. The second is having no comparison data to measure it.

• Do nothing while waiting for official announcements. Microsoft hasn't announced the switch to end users. Anyone waiting for a notification might never get one.
• Renewing licenses without an audit. Auto-renewal is convenient, but right now it's a risk: you're paying for a service that might have different features from the ones you originally chose it for.
• Having no baseline for comparison. If you haven't saved examples of past outputs, you can't measure if the quality has changed. Start doing it now.
• Treating Copilot as a monolithic tool. Copilot is not a single product: it is an interface that changes the underlying model over time. You need to evaluate it just like you would evaluate any service provider.
• Ignoring internal training. With only 35.8% actual adoption among license holders, the issue is often not the model: it's that employees don't know how to use it well. Investing in AI training for companies is the first step.

How to figure out if Copilot is still working well for your company?

You need concrete indicators, not impressions. Here is a reference table to monitor output quality over time.

So, is it worth keeping Copilot or is it time to look at alternatives?

It is worth it if you use Copilot for structured tasks (formulas, summaries, internal searches), you have an enterprise plan with dedicated support, and you are willing to conduct periodic output audits. In this case, the cost savings that Microsoft passes on (at least in part) justify the continuity.

A middle ground You use Copilot for mixed tasks — some structured, some high-reasoning. Benchmark now, before the next renewal. If the results are acceptable, continue. If not, explore plans with selectable models or hybrid solutions.

Not yet if You've built critical automated workflows on Copilot and haven't audited the outputs yet. In this case, the operational risk is high: a silent model change can break logic that seemed stable. Act before the next renewal.

How we see it at SHM

For us, the news of model changes in Copilot is not a crisis: it's a reminder. AI should be integrated to free companies from routine, not to create new opaque dependencies. When the engine under a subscription changes without communication, the problem is not technical — it's governance.

We have worked with companies that use complex AI tools — from managing technical catalogs to international SEO — and in every case, the starting point is the same: understand what you are actually using, not just what you have bought. This applies to Copilot as much as to any other tool.

If you are evaluating how to integrate AI into your processes in a controlled way, our service of strategic AI integration it starts right with an audit of what you already use. For those who want to understand how AI models impact online visibility, we also work on SEO for AI . And for teams that need training to actually use these tools, we offer paths of business process automation training .

You can explore all these topics in our dedicated section on AI and process automation , where we gather hands-on insights for decision-makers, not just for those who want to stay up to date.

It's also worth keeping an eye on how the model landscape evolves: OpenAI is developing models for browsers and computer use which could shake things up again, and security incidents on AI models show that governance is not optional.

Frequently Asked Questions

The real problem isn't which model Microsoft uses: it's that your company is paying for a tool whose specifications it doesn't control.