The Story

Microsoft has unveiled new multi-model capabilities in its Copilot Researcher tool, introducing two features called Critique and Council.

As detailed in the official announcement on Microsoft Tech Community, these features allow multiple AI models to collaborate, review, and challenge each other’s outputs in real time.

How the system works

• Critique mode: One model generates a research report, while a second model evaluates it for quality, accuracy, and completeness.
• Dual-model validation: The reviewing model checks sources, reasoning, and evidence grounding before final output.
• Improved reliability: The approach aims to reduce hallucinations and improve trust in AI-generated research.

Council mode explained

• Side-by-side comparison: Multiple models run in parallel on the same task.
• Agreement detection: The system highlights where models agree or disagree.
• Insight surfacing: Unique contributions from each model are surfaced for better decision-making.

Part of a broader shift

The update builds on Microsoft’s growing investment in multi-model AI systems, where different models are orchestrated together instead of relying on a single model.

This comes alongside developments like Copilot Cowork and other agentic tools designed to handle complex, multi-step workflows.

Why it matters

The future of AI usage is increasingly moving toward multi-model orchestration, where systems collaborate, critique, and validate outputs rather than operate independently.

By enabling models to challenge each other, Microsoft is pushing toward more reliable and trustworthy AI systems — a key requirement for enterprise adoption and high-stakes use cases.