Microsoft has introduced a few major updates to its Microsoft 365 Copilot. One of the key additions is a multi-model AI feature known as ‘Critique’, which allows one AI model to generate responses while another reviews them for accuracy and quality, helping reduce errors and hallucinations. The company has also launched a ‘Model Council’ tool that shows responses from multiple AI models side by side, allowing users to compare and verify outputs. At the same time, Microsoft is rolling out Copilot Cowork to early-access users through its Frontier program, enabling AI to plan and execute multi-step workplace tasks across apps.
At the center of Microsoft’s latest Copilot upgrade is a new system called Critique, which fundamentally changes how AI-generated responses are produced. Instead of relying on a single model to generate answers, the company now uses a multi-step process in which one model creates a response, and another independently reviews it. For this feature, the Satya Nadella-led company has teamed up with AI giants like OpenAI and Anthropic. Under this system, a model from OpenAI typically produces the initial output, while a model from Anthropic is used to review, critique, and refine the response.
The reviewing model evaluates the response for factual accuracy, logical consistency, and completeness before it is delivered to the user. It can suggest corrections, identify gaps in reasoning, and refine the output to make it more trustworthy and polished. Even the software giant is experimenting with reversing these roles, allowing different models to generate and critique interchangeably depending on the task.
Meanwhile, complementing the Critique feature, the tech titan also introduced a new capability called Model Council, aimed at bringing multiple AI perspectives into a single workflow. Instead of delivering one final answer, the system presents responses from different models side by side, allowing users to directly compare how each interprets the same query. The feature is particularly useful in scenarios that require deeper analysis and strategic thinking, where a single answer may not fully capture the complexity of a problem.
Along with these upgrades, the firm is also expanding its automation push through Copilot Cowork. The feature is designed for long-running, multi-step work across Microsoft 365 applications. And now, the firm has made Copilot Cowork available to early-access customers through its Frontier program, allowing select users to test and provide feedback on the system before a wider rollout. These developments come at a time of intensifying competition in the AI space. Major players like Google DeepMind and Anthropic are racing to improve not only the intelligence of their models but also their reliability and practical utility.
The Tech Portal is published by Blue Box Media Private Limited. Our investors have no influence over our reporting. Read our full Ownership and Funding Disclosure →
