Microsoft Just Made AI Smarter By Making It Work Together
Microsoft just did something quietly brilliant. Instead of betting on one AI to rule them all, they're making two AIs work together. And the results are better than either one alone.
Microsoft launched Copilot Cowork through its Frontier Program. Here's what changed: You can now have GPT draft your research while Claude edits it for accuracy. Or flip it around. The new "Council mode" gives you two separate reports plus a summary showing where the AIs agree and where they don't.
The numbers back it up. Microsoft tested this approach on the DRACO benchmark (a standard test for AI research quality) and saw a 13.8% improvement over standalone tools from OpenAI, Google, Perplexity, and Anthropic.
Why does this matter if you're not a researcher? Because we've been asking the wrong question. The AI wars have been about which model is best. Microsoft just proved the real answer is "both."
Think about it. When you're making a big decision at work, do you ask one person and call it done? Or do you get a second opinion? Microsoft built that instinct into AI.
Here's how it works in practice. Say you're researching competitors for a pitch deck. GPT can move fast and gather a ton of information quickly. But Claude tends to be more careful about fact-checking and finding gaps. Together, they catch each other's blind spots.
You can try this today if you have a Microsoft 365 Copilot license. Open the Researcher agent, and you'll see the new Cowork option. Pick your task, and both AIs get to work. You'll get:
- A full report from GPT
- A critique and edit from Claude (or vice versa)
- A summary showing where they agree and where they diverge
That last part is the secret weapon. When two smart AIs disagree, that's where you need to pay attention. It's a built-in BS detector.
Microsoft isn't stopping here. They plan to reverse the workflow in future updates so Claude can draft and GPT can critique. They're also expanding the Critique feature beyond just research tasks.
This is bigger than one feature update. It's a shift in how we should think about using AI at work.
Instead of hunting for the "perfect" AI tool, the smarter play is to use multiple AIs the same way you'd assemble a team. Different strengths, different perspectives, better outcomes.
The single-AI era is over. The collaboration era just started.
Try it: If you have Microsoft 365 Copilot, open the Researcher agent and look for the Cowork option. If you don't, you can mimic this approach manually: ask ChatGPT to draft something, then paste the result into Claude and ask it to critique and improve it.
Who should use this: Anyone doing research-heavy work. Market analysis, competitive intelligence, literature reviews, due diligence, strategic planning. Anywhere you need accuracy and depth, not just speed.
The catch: You need a Microsoft 365 Copilot license. That runs about $30/month on top of your Microsoft 365 subscription. Not cheap, but if research is a core part of your job, the time savings pay for it fast.
The bigger idea here isn't just about Copilot. It's about rethinking how you use AI. Stop looking for one tool to do everything. Start building workflows where different AIs check each other's work.
Your next best idea might come from making your AIs argue with each other.