Microsoft is experimenting with a modified approach to artificial intelligence by introducing a new deep research system called Critique, which focuses on improving the reliability of output results.

Instead of relying on a single model, Critique incorporates multiple models into the workflow. One model takes the lead in generating and structuring responses, while another intervenes before finalization to review, refine, and question the output. The underlying concept is straightforward: treating AI more like a collaborative team rather than a single entity, which Microsoft claims leads to superior outcomes.

In parallel, Microsoft has also launched a system named Council, which operates different models concurrently, including those from OpenAI and Anthropic. Each model produces its own report, after which an independent system compares them, highlighting areas of agreement, disagreement, and unique contributions from each model.