Microsoft's Copilot Cowork introduces more advanced AI research tools designed to identify gaps in your work.

      Digital Trends may receive a commission for purchases made through links on our site. Why should you trust us?

      Your work can now receive a second opinion with Microsoft’s new Researcher features.

      Earlier this month, Microsoft introduced Copilot Cowork, a tool based on Anthropic’s Claude Cowork. The company is now offering early access to Copilot Cowork via its Frontier program, along with updates to its Researcher tool that will assist you in planning, analyzing, and making work-related decisions.

      So, what capabilities does Microsoft’s Copilot Cowork offer?

      Copilot Cowork is an AI tool designed to manage long, multi-step tasks within Microsoft 365. It can help you strategize tasks, break down objectives, and collaborate with you like a coworker across various documents and workflows.

      You specify the outcome you desire, and it formulates a plan and executes the task while displaying its progress. You can intervene and adjust its course at any moment. It is capable of managing everything from one-time requests to recurring workflows like monthly budget assessments.

      New AI upgrades to Copilot’s Researcher tool

      Microsoft is enhancing Researcher, its comprehensive research feature within Copilot, with two significant additions.

      The first addition is Critique, a configuration where two AI models collaborate on the same task. OpenAI’s GPT produces the initial response, and Anthropic’s Claude evaluates it for accuracy and quality before it reaches you.

      As reported by Reuters, Microsoft intends to make this interaction bi-directional in the future, meaning Claude’s drafts might also be reviewed by GPT.

      Microsoft stated that this feature has improved Researcher’s score by 13.8% on the DRACO benchmark, an industry standard for assessing the accuracy and quality of in-depth research.

      The second new feature, Council, allows you to gather responses from various AI models and compare them side by side. This enables you to quickly identify areas of agreement, differences, and the unique contributions each model provides to your inquiry.

      Microsoft indicates that all these developments are part of Wave 3 of Microsoft 365 Copilot, aimed at transitioning AI from a tool for experimentation to one that actively completes your work for you.

      Manisha enjoys covering technology that integrates into daily life, encompassing smartphones and apps to gaming and streaming.

      The Asus Morph 96 Wireless offers a customizable keyboard experience without the DIY complications. It features a gasket-mount design, foam damping, and a hot-swap PCB, allowing you to experiment with switches and sound without building a keyboard from scratch.

      Asus is attempting to enter the custom keyboard market without fully committing, and the ROG Strix Morph 96 Wireless is its latest effort. First revealed at CES 2026, the board is aimed at enthusiasts, combining customization with ease of use.

      AI has outperformed chess grandmasters but struggles to adjust to modern video games.

      Contemporary video games are revealing what AI still struggles to accomplish. Despite the buzz surrounding AI’s achievements in chess, Go, and even coding, a significant weakness persists: AI often underperforms in unfamiliar video games. A recent paper from NYU argues that these notable milestones have created a misleading impression of how close machines are to achieving real general intelligence.

      Stanford researchers caution against relying on AI chatbots for personal guidance.

      Researchers found that users tend to prefer agreeable bots, even when such responses render them less empathetic and more morally rigid. Stanford researchers warn that using AI chatbots for personal advice may have unintended consequences. The issue isn’t solely about accuracy; it’s also about these systems’ responses in complicated, real-world situations. A recent study disclosed that AI models frequently align with users, even when they are mistaken, thereby reinforcing poor decisions rather than challenging them. This trend not only influences the advice given but also alters individuals' perceptions of their actions. Participants who engaged with excessively agreeable chatbots became increasingly convinced of their correctness and less inclined to empathize or rectify the situation.