Microsoft has rolled out significant updates to its Microsoft 365 Copilot AI assistant, introducing features that allow the system to use multiple artificial intelligence models simultaneously when answering user queries. This marks a shift from relying on a single AI model to a multi‑model approach that combines strengths from different systems such as OpenAI’s ChatGPT and Anthropic’s Claude. The changes are part of Microsoft’s ongoing effort to enhance Copilot’s capabilities, improve response quality, and reduce inaccuracies often referred to as AI hallucinations.
New ‘Critique’ Feature for Better Answers
The headline feature in this update is called “Critique.” Under this workflow, Copilot’s Researcher agent first uses a primary model, typically from OpenAI’s GPT family, to create an initial answer draft. A secondary model, such as Claude developed by Anthropic, then evaluates the response for quality, completeness, and accuracy before it is delivered to the user. Microsoft plans to extend this workflow in the future so that both models can review each other’s outputs in either direction.
By pairing models in this way, Copilot aims to reduce common issues in AI responses such as factual errors or omissions. According to Microsoft, the Critique workflow improves the quality of deep research answers by combining the reasoning capabilities of different models. Experts say that multi‑model strategies like this can help catch blind spots that one model alone might miss.
Model Council for Side‑by‑Side Comparison
Another notable addition is the “Model Council” feature. This tool allows users to view responses generated by different AI models side by side. Users can then compare the outputs and choose the one that best meets their needs or provides the most accurate information. This level of transparency and choice is aimed at giving professionals and organisations better control over AI‑assisted research and decision‑making.
Model Council is especially useful in scenarios where users require nuanced insights or diverse perspectives on complex topics. By examining multiple answers in parallel, users can identify where AI models agree or diverge, helping to validate information before it is used professionally.
Copilot Cowork and Broader Availability
In addition to these features, Microsoft is expanding access to Copilot Cowork under its early‑access “Frontier” programme. Copilot Cowork is designed for more autonomous collaboration, enabling the AI to help manage workflows and complete multi‑step tasks across Microsoft 365 applications such as Outlook, Teams, Word, and Excel. This move reflects Microsoft’s broader strategy to make Copilot not only a research tool but also a productivity platform that assists with complex work processes.
Impact on AI and Productivity Tools
Microsoft’s new multi‑model approach aligns with a wider industry trend toward integrating multiple large language models into a single system to improve accuracy and reliability. Analysts believe that combining models from different providers helps balance strengths and address weaknesses inherent in individual models. While this can introduce challenges such as higher computational costs or slower response times, many users and organisations see these trade‑offs as worthwhile for higher‑quality outputs and deeper insights.
As AI continues to evolve, tools like Copilot that leverage diverse AI models could play a central role in workplace productivity, research, and decision support. With these latest enhancements, Microsoft is positioning Copilot as a more powerful and reliable assistant for businesses and professionals seeking intelligent automation and accurate answers.