For years, picking an AI assistant meant picking a side: Copilot with OpenAI’s GPT, or Anthropic’s Claude, or Google’s Gemini. Each one stayed in its own lane. Microsoft just broke that logic. As of March 30, 2026, its Copilot assistant no longer relies on a single brain: it now puts GPT and Claude to work together on the same question, for more reliable research.

The change lands in the Researcher agent inside Microsoft 365 Copilot, through two modes called Critique and Council. The idea is simple: instead of trusting one model’s word, Copilot has a second model check the first one’s work. For a small business in Trois-Rivières or the Mauricie region that’s starting to lean on AI for reports, bids, or market research, that detail matters more than it sounds.

1. What Microsoft actually announced

On March 30, 2026, Microsoft added the ability to combine multiple AI models to its Researcher agent (Copilot’s deep-research mode), instead of relying on just one. This is what’s known as a multi-model approach. In practice, two new features arrived.

The first, Critique, splits the research into two steps. One model plans the work, gathers information, and writes a draft. A second model then reviews that draft to check its accuracy, completeness, and how solid the cited sources are. In practice, Microsoft often has GPT write and lets Claude play the reviewer, a bit like an employee who drafts a document and a seasoned colleague who checks it before it goes out.

The second, Council, asks GPT and Claude the same question in parallel. Each one produces its own full report independently. A third model, acting as referee, reads both versions and puts together a summary: where the two AIs agree, where they differ, and what each one found on its own. You end up with two points of view instead of one.

2. Why two AIs beat one

The biggest flaw in AI assistants is what’s called a confident error: the model invents a figure, a date, or a source, and presents it with total assurance. For a business, one wrong statistic in a bid or a management report can be costly to its credibility.

Having a second model check the answer cuts that risk down. Microsoft measured the gain using a benchmark called DRACO (100 complex research tasks across 10 domains, originally designed by researchers at Perplexity). According to Microsoft, Critique mode scored about 7 points higher on this test, roughly 14% better than a competing deep-research tool. The gains showed up in analysis depth, presentation quality, and factual accuracy alike. These numbers come from Microsoft, so it’s worth testing against your own use, but the trend is clear.

For a small business, the value isn’t just theoretical. A few concrete examples where double-checking matters:

• Writing a market analysis before a product launch.
• Comparing vendors or software solutions for a project.
• Preparing a regulatory filing or a document for a key client.
• Summarizing a long contract or a tender without missing a clause.

3. What this actually changes for your business in Quebec

Good news: there’s nothing to install. If your business already has a Microsoft 365 Copilot license, these features show up in the Researcher agent. In the model selector, the “Auto” option turns on the verification logic, and “Model Council” launches the side-by-side comparison between the two AIs. Access still depends on your plan and on the permissions your IT administrator turns on (the features first rolled out to organizations in Microsoft’s Frontier program, ahead of a wider release).

This is exactly the kind of setting where a bit of guidance helps. Turning on the right option, deciding who in the company gets access, and setting rules around what data goes into the AI: these things take planning. Our team can handle that for you as part of our managed IT services, with your obligations in mind, including Quebec’s Law 25.

One note on cost, without getting into pricing since it keeps changing: running two AIs uses more resources than running one. Microsoft states, as a general guide, that Critique mode costs about 20% more than a single model, and Council mode can cost close to 2.5 times more. The logic makes sense: save the double-checking for research that’s actually worth it, not for every small question.

4. The real signal behind this announcement

Beyond the feature itself, this news says something important about AI in 2026. The big players are stepping away from betting everything on one model. Microsoft, long tied to OpenAI, is now building its rival Claude into its flagship product. The message for businesses: no AI is perfect, and reliability comes from comparing points of view.

For a small business, the lesson is easy to apply. Don’t depend on a single tool, a single vendor, or a single habit. The same caution applies to your backups, your access controls, and your security. An AI that checks its own work is progress, but it doesn’t replace a person reviewing what goes out to a client. Always keep a final human check on documents that matter.

• Use double-checking (Critique or Council) for high-stakes research.
• Keep simple questions on a single model; it’s faster and cheaper.
• Always have a person review before an official send-out.
• Set clear rules around the confidential data your teams feed into AI.

Frequently asked questions

Does Copilot really use Claude and GPT at the same time?

Yes, within the Researcher agent in Microsoft 365 Copilot. Critique mode has one model write the answer and the other review it, and Council mode queries both in parallel and compares their answers. It launched on March 30, 2026.

Does my business need a special license?

You need a Microsoft 365 Copilot license. The features appear in Researcher’s model selector. Depending on your plan and your IT administrator’s settings, access may roll out gradually. An IT provider can check and turn this on for you.

Does this eliminate AI errors?

No, but it reduces them. Having a second model check the answer improves accuracy, according to Microsoft’s own testing. A human review is still recommended before any important document goes out, especially to a client.

Get the most out of AI, without the pitfalls

Multi-model AI is a solid step forward, but it only delivers real value when it’s set up and managed properly. At OKTO Solutions, based in Trois-Rivières, we help small businesses across the Mauricie region and Quebec deploy Microsoft 365 Copilot securely, train their teams, and stay in control of their data. Check out our managed IT services or reach out through our contact page to talk it over.