Microsoft rolls out Copilot Cowork to its Frontier program for early-stage testing, including a new Researcher Critique tool using Anthropic and OpenAI models
Try Microsoft 365 Copilot — Available on desktop and mobile devices — Microsoft 365 Copilot is your AI for work.
The new Critique and Council-style approach follows Microsoft’s prior rollout of OpenAI-powered Researcher and Analyst agents as digital colleagues. The notable change is the shift from standalone research agents toward workflows that can draw on more than one model.
First-order effects
Frontier participants can test Copilot Cowork and the new Researcher Critique capability before broader availability, giving Microsoft feedback on a multi-model research workflow.
Anthropic and OpenAI models gain roles inside the same Microsoft 365 Copilot feature, while Microsoft remains the product layer that presents and coordinates their output.
Second-order effects
Microsoft will need to evaluate model-routing, output consistency, and user trust when multiple models contribute to one work task; the early-access program makes that evaluation part of product development.
The move raises the bar for workplace AI rivals: a single assistant increasingly has to combine specialized model capabilities rather than rely on one model for every research and critique task.
Third-order effects
If this testing becomes a durable product pattern, enterprise AI assistants may compete less on exclusive access to one foundation model and more on orchestration, grounding, and governance across models.
That would reinforce Microsoft 365 Copilot as an agentic work surface, with model providers supplying components while the workplace platform owns the user workflow and organizational context.
The trend: Enterprise AI is moving toward multi-model, work-contextual assistants that coordinate specialized capabilities inside a single productivity interface.
Introducing Critique, a new multi-model deep research system in M365 Copilot. You can use multiple models together to generate optimal responses and reports. [video]
Microsoft's Researcher AI agent is getting a new ‘Critique’ feature that uses OpenAI's GPT and Anthropic's Claude models together to improve the quality of its responses https://www.thurrott.com/...
Imagine reading this and being bullish model cos? $MSFT “Council is an alternative approach, designed for side-by-side comparison across multiple models. Available when Model Council is selected in the model picker in Researcher, Council runs an Anthropic and OpenAl model
I was eating dinner with my family last night and they all started talking about copilot Now I've never heard anybody in the wild to talk about copilot. My family is in medical and banking. They are constrained to the Microsoft ecosystem because Microsoft has an amazing sales
$MSFT just introduced Critique which is a new multi-model deep research system inside M365 Copilot that lets users combine models to generate stronger reports & responses. Microsoft is increasingly positioning itself as the interface layer for enterprise AI by monetizing [video]
Fascinating watching Microsoft come to the realization that betting the whole company on a single LLM provider is actually a horrible decision Microsoft was the same business that invested billions of dollars in OpenAI and wrapped them up in Azure cloud credits Markets rewarded
openai's biggest investor is now launching products powered by Claude and competing AI labs. microsoft's new research AI is now #1: - claude generates research report then chatgpt fact-checks and improves it (vice-versa). - tested it on 100 real complex research tasks across 10 […
Microsoft has just released a VERY powerful feature Copilot Critique allows one model from Anthropic/OpenAI to generate the research output and another one reviews it. → Model 1 plans, retrieves sources, synthesizes, and drafts → Model 2 evaluates claims, strengthens