Microsoft Enhances Copilot with Multi-Model AI Integration

On March 30, 2026, Microsoft announced significant upgrades to its Copilot Researcher agent, introducing a feature called "Critique." This enhancement allows the integration of multiple AI models, specifically OpenAI's GPT and Anthropic's Claude, to improve the quality and accuracy of responses generated within Microsoft 365. In this workflow, GPT is responsible for drafting responses, while Claude reviews them for accuracy, completeness, and citation integrity before they are presented to users. Microsoft plans to eventually enable a bi-directional process where Claude can also draft responses for GPT to critique.

Microsoft 365 Copilot’s Researcher Agent Now Uses GPT and Claude to Improve AnswersThe New StackMicrosoft pairs GPT with Claude to reduce AI hallucinationsGeekWireMicrosoft 365 Copilot's new agent uses Claude to fact-check GPT's workReuters

Microsoft's strategy reflects a shift from reliance on a single AI model to a more collaborative approach that leverages the strengths of different models. Nicole Herskowitz, corporate vice president of Microsoft 365 and Copilot, emphasized that this multi-model approach aims to enhance user workflows and mitigate issues related to AI hallucinations—instances where AI generates inaccurate information. The integration has reportedly improved the Researcher agent's performance by 13.8% on the DRACO benchmark, an industry standard for deep research quality.

Microsoft 365 Copilot’s Researcher Agent Now Uses GPT and Claude to Improve Answers

Alongside the Researcher agent updates, Microsoft launched Copilot Cowork, a tool designed to automate long-running, multi-step tasks across various Microsoft 365 applications, including Excel and Outlook. This feature allows users to describe desired outcomes, enabling the system to create and execute plans autonomously. Copilot Cowork is currently available through Microsoft's Frontier program, which offers early access to new features for select enterprise customers.

Microsoft Copilot Cowork takes on multi-step AI automationMicrosoft pairs GPT with Claude to reduce AI hallucinationsGeekWireThe New StackReutersMicrosoft 365 Copilot’s Researcher Agent Now Uses GPT and Claude to Improve Answers

Barton Warner, Senior Vice President of Enterprise Technology at Capital Group, highlighted the practical benefits of Copilot Cowork, stating that it effectively connects steps, coordinates tasks, and enhances everyday workflows. This integration of AI models not only aims to improve productivity but also addresses concerns regarding data security, as the Copilot Cowork operates within a controlled environment, ensuring that enterprise data remains secure.

The New Stack

Despite these advancements, Microsoft faces intense competition from other AI providers, including Google's Gemini and Anthropic's Claude. Critics may question whether the value of Microsoft’s services lies in the orchestration of these models or in the underlying enterprise data and trust layer that supports them. As Microsoft diversifies its AI partnerships, it remains to be seen how effectively it can maintain its competitive edge in the rapidly evolving AI landscape.

Microsoft pairs GPT with Claude to reduce AI hallucinations

Microsoft's latest enhancements to its Copilot tools signify a strategic pivot towards a multi-model AI framework, aiming to improve the quality of outputs and streamline workflows for enterprise users. As the company continues to innovate and expand its offerings, the effectiveness of these integrations will be crucial in determining their long-term success in the competitive AI market.

Reuters