Microsoft has put Critique into the M365 Copilot in-depth research process. The core is not to change to a larger model, but to involve multiple models in generation, review, and revision together. The changes seen by users will be closer to a research process of "drafting first, then reviewing, and then completing a draft".
- The focus is on "using multiple models together"
The most important thing to grasp this time is not the word in-depth research, but the collaboration of multiple models. According to Microsoft, Critique puts different models into the same round of tasks to generate better answers and reports, rather than letting a single model complete the entire content alone.
Second, it changes the report generation method
In the past, such tools were more like spitting out results at one time, but now they are more like going through a round of internal mutual review. One model is responsible for drafting and the other is responsible for picking mistakes, supplementing and organizing structure, so the final deliverable will look more like a "processed" research report than just the original answer.
- How will this ability appear in front of users?
Microsoft put Critique into Researcher and tied it closer to Auto mode. For users, they do not necessarily need to manually specify which model to play. The system will first organize the research process according to the multi-model route.
- Which tasks is more suitable for this update?
What is most suitable for it is not a one-sentence question and answer, but a task that requires checking information, comparing solutions, and writing long documents. For example, content such as industry research, competitive product analysis, and project reports rely more on "write it first and check it again", so Critique's changes will be easier to perceive than ordinary chat.
- What we need to see now is the open scope
This update provides a more accurate understanding that Researcher is upgrading its in-depth research capabilities. This does not mean that all scenarios in the entire M365 Copilot have been switched to multi-model collaboration by default. The actual availability range depends on whether the tenant has entered the Frontier and whether the corresponding entrance has been opened.
