Trois-Rivières, serving all of Quebec 450 231-3836 info@oktosolutions.ca
FR

Microsoft brings its own AI models (MAI) to Copilot

Portrait of Antonio Pazzi, president of OKTO Solutions

By ·

President of OKTO Solutions · 6 minute read

At its Build 2026 conference, held from June 2 to 4, Microsoft made an announcement that slipped past most of the general public but speaks directly to businesses: the company unveiled its own family of artificial intelligence models, called MAI (short for Microsoft AI). Until now, nearly all of Microsoft’s AI, from the Copilot assistant in Word to GitHub Copilot for developers, ran on models built by OpenAI, the maker of ChatGPT.

With MAI, Microsoft is starting to build and run its own digital brains. For a small business in Trois-Rivières, the Mauricie or Quebec City that already pays for Microsoft 365 or Copilot, this shift raises one plain question: are my tools going to change, and do I have to do anything about it? Here is what matters, without the jargon.

Quick answer: Microsoft introduced MAI, its own family of AI models (reasoning, code, image, voice), to depend less on OpenAI. These models will gradually power Copilot and Windows. For your business, the change will be invisible at first (same button, different engine underneath), but it is worth knowing what is coming.

1. What did Microsoft announce at Build 2026?

For years, Microsoft counted on OpenAI to run its smart tools. At Build 2026, the tone changed: the company now wants to build a good share of its AI itself. It presented a family of MAI models, each one specialized for a specific job. Based on what was shared on stage and reported by the trade press, that family includes:

  • A reasoning model (presented as MAI-Reason, or MAI-Thinking depending on the source) for analysis, problem solving and the much discussed "agents" that carry out actions on their own.
  • A code model (MAI-Code) to assist developers, meant to power GitHub Copilot.
  • Image, transcription and voice models to generate visuals, transcribe meetings and read text out loud.

One detail stands out: according to Microsoft, its reasoning model was trained without using OpenAI data. In other words, Microsoft is no longer just renting a partner’s AI, it is developing its own recipe from scratch.

2. Why is Microsoft building its own models?

Since 2023, Microsoft and OpenAI have been the most visible pairing in the AI industry. But relying on a single supplier for a technology this strategic carries real risk: price increases, shifting priorities, or simply not owning your own recipe. By developing MAI, Microsoft is trying to take back control of three concrete things.

  • Cost: paying an outside partner for every AI request adds up quickly at scale. Owning the models lets Microsoft keep a tighter grip on the bill, which can eventually show up in subscription pricing.
  • Confidentiality: by keeping the technology in house, Microsoft promises better control over the path your data takes, an important argument for companies bound by strict rules (think of Quebec’s Law 25).
  • Integration: by owning the engine, Microsoft can tune it more precisely for Windows, Microsoft 365 and Azure, without waiting on a third party’s schedule.

This is not a clean break with OpenAI overnight. Several sources point instead to a gradual transition, with a stretch where both technologies live side by side under the hood. Microsoft also remains an investor in OpenAI.

OKTO Solutions technician migrating a client’s data to the cloud

3. What changes in practice for your small business?

If your company already uses Microsoft 365, Copilot or Windows 11, here is what to expect. Note that the dates below are indicative: rollout schedules shift often, and you should always confirm them on Microsoft’s official channels.

  • Microsoft 365 Copilot: according to Microsoft, the new reasoning model should start reaching enterprise customers as early as the summer of 2026 in Word, Excel and PowerPoint. The Copilot button you know stays the same, but the engine behind it could be an MAI model rather than an OpenAI one.
  • GitHub Copilot: for teams that build software, the MAI code model is starting to be folded in, with a phase where it runs alongside the existing technology. Developers in the region who use this tool will see the switch happen bit by bit.
  • Windows 11: some AI features (image generation, transcription, text-to-speech) are starting to show up in tools such as Paint and Narrator, first for people in the Windows Insider program.

For most small businesses, the change will be largely invisible at first: same interface, same shortcuts, different engine in the background. That is exactly why it pays to know about it now rather than find out after the fact. If you want to get the most out of Copilot without getting lost in technical detail, our managed IT services exist for that.

4. Should you worry? Privacy, quality and continuity

Swapping the engine raises fair questions. Here is how to look at it calmly, without panic and without brushing it off.

  • Privacy: in theory, keeping the AI in house gives Microsoft more control over your data. In practice, it changes nothing about your own responsibility: you still need to know what information your employees paste into Copilot, and to set the right protection options.
  • Quality of results: a new model can give slightly different answers. During the transition, keep the habit of rereading and validating what the AI produces, especially for important documents or figures.
  • Continuity: nothing forces you to act in a hurry. Your tools keep working. The smart move is to stay informed and plan ahead, not to overhaul everything overnight.

OKTO Solutions advisor listening carefully to a client’s needs

5. How do you get your business ready for the AI shift?

Whether you already use Copilot or you are still on the fence, a few simple steps put you in good shape for the months ahead.

  • Take inventory of your AI tools: who uses Copilot, ChatGPT, Gemini or anything else in your company, and what for? You can only protect what you know about.
  • Spell out your data rules: put in writing which information can go into an AI tool and which has to stay confidential (client records, financial data, personnel files).
  • Train your team: two or three good habits people actually understand beat a ten-page policy nobody reads.
  • Keep a plan B: never let a critical process rest on a single tool. Technology transitions like MAI are a reminder that you always need a way to keep working if a service changes.

At OKTO Solutions, we walk small businesses in Trois-Rivières, the Mauricie and Quebec City through decisions like these, without selling jargon or fear. The goal: technology that works for you, not the other way around.

OKTO Solutions technician answering the phone with a smile

Frequently asked questions

What are Microsoft’s MAI models?

MAI (Microsoft AI) is the family of artificial intelligence models Microsoft is developing in house, unveiled at the Build 2026 conference. It brings together several specialized models (reasoning, code, image, transcription, voice) meant to run Copilot, Windows and other products.

Will Copilot stop using OpenAI technology?

Not right away, no. Microsoft is planning a gradual transition where its own MAI models and OpenAI technology coexist for a while. The aim is to depend less on a single supplier, not to cut ties all at once.

Do I need to change anything in my business right now?

No, nothing urgent. Your Microsoft 365 and Copilot tools keep working as usual. The right reflex is to stay informed, keep sound habits with your data and check what the AI produces, especially on important documents.

Stay ahead of technology shifts

Microsoft’s move toward its own MAI models shows one thing: the AI world moves fast, and the tools you use every day can change engines without warning. You do not have to follow all of it yourself. Take a look at our IT services for small business or get in touch with the OKTO Solutions team to review how you use Copilot and AI, in plain language.

Reading about AI is one thing. Connecting it to your own data is another: artificial intelligence in business, custom AI application development and our IT services in Quebec City.

A question on this subject, for your own company?

An article explains the principle. A twenty minute call tells you what it changes at your place, with your systems and your constraints.