Microsoft has unveiled early access to Copilot Cowork through its Frontier program, introducing significant enhancements to its Researcher feature. These updates are designed to streamline planning, analysis, and decision-making workflows, reflecting the company’s commitment to integrating advanced AI capabilities into its suite of tools.
Copilot Cowork Targets Multi-Step Work Tasks
At the heart of this initiative is Copilot Cowork, an innovative agent-based AI system tailored to manage intricate, multi-step tasks within the Microsoft 365 ecosystem. Users can outline their desired outcomes, prompting the tool to formulate a structured plan and execute the task while providing real-time progress updates.
This dynamic system allows for user intervention at any point, enabling adjustments or redirection of the process as needed. It accommodates both one-off tasks and recurring workflows, such as monthly budget reviews, functioning much like a digital collaborator across various documents and work environments.
Notably, Copilot Cowork is built on the Claude Cowork framework developed by Anthropic, showcasing a blend of cutting-edge technology and user-centric design.
Researcher Tool Gains Critique Feature
In tandem with Copilot Cowork, Microsoft has enhanced its Researcher tool by introducing a new Critique feature. This innovative setup employs two AI models working collaboratively on the same task. Initially, GPT generates a response, which is subsequently reviewed by Claude for accuracy and quality before reaching the user.
According to reports from Reuters, Microsoft envisions expanding this system into a two-way process, where GPT may also evaluate responses generated by Claude in the future. Early results indicate that the Critique feature has bolstered the Researcher tool’s performance by 13.8% on the DRACO benchmark, a metric that assesses deep research accuracy and quality.
Model Council Enables Side-by-Side AI Comparisons
Another noteworthy addition to the Researcher tool is the Model Council feature. This functionality empowers users to generate outputs from multiple AI models and compare them side by side. It emphasizes the differences, similarities, and unique contributions of each model, thereby enhancing transparency and user control over AI-generated content.
Part of Microsoft 365 Copilot Wave 3
These advancements are part of Wave 3 of Microsoft 365 Copilot, an initiative aimed at transitioning AI from a mere assistance tool to an active participant in executing work tasks on behalf of users. This strategic shift underscores Microsoft’s vision of a more integrated and efficient workplace powered by AI innovation.
