Microsoft 365 Copilot’s new Model Council feature in Researcher lets you run GPT and Claude simultaneously on the same task. I tried it on a Thursday afternoon. Here’s what I saw.

It started as a small experiment. I’d heard about Model Council a new feature inside Microsoft 365 Copilot’s Researcher agent that lets you run multiple AI models in parallel on the same prompt. GPT and Claude, side by side, both answering.


So I gave them a task I do every morning anyway: build my daily briefing from my emails, messages, and calendar. I expected one of them to stumble. Neither did.

What they produced
Both models delivered a complete daily briefing. Not a list of links, not a vague summary. A structured, contextual briefing with meetings, key updates, and relevant context pulled from actual sources.


Claude started with the calendar. It immediately picked up meeting context and looked for prep material per appointment. People and timing first. GPT started broader: gathering relevant sources, building context, then working toward the specific agenda items.
Same input. Same sources. Different instincts about where to begin.


Why this matters more than it looks
For two years we’ve been debating which AI model is best. GPT versus Claude versus Gemini. Benchmarks. Gut feelings. LinkedIn threads that go nowhere.


Model Council sidesteps that debate entirely. It doesn’t ask you to pick. It runs them together and lets you see how they think differently about the same problem.


“This isn’t about which model wins. It’s about what becomes possible when you stop picking one and start orchestrating many.”


That’s not a small shift. That’s a fundamental change in how we think about AI in the workplace.

What I actually noticed
Both models completed the task without hand holding. Claude prioritised calendar and meeting context immediately. GPT prioritised source gathering first. And the differences were visible from the very first sentence.

In practice, that means something real: depending on the task, one model’s instinct will be more useful than the other’s. Running them together means you don’t have to guess in advance which one to use.


The Frontier Firm isn’t coming. It’s here.
I talk a lot about Frontier Firms organisations that don’t just use AI, but fundamentally redesign how work happens around it. Model Council is a small but concrete example of what that looks like in practice.


Not one AI assistant you prompt. A council of models reasoning over your actual work in parallel, inside the tools you already use.