What Microsoft Just Changed in Copilot
Microsoft (MSFT) unveiled new upgrades to its Copilot research assistant aimed at making results more reliable while keeping work moving faster. The headline shift is a multi-model workflow that lets Copilot tap more than one AI system for a single task—an approach designed to improve quality control as AI becomes more embedded in everyday office work.
At the center of the update is a new feature called Critique inside Copilot’s Researcher agent. Instead of relying on one model per response, Copilot can now combine two different models in the same flow.
- OpenAI’s GPT generates the draft response.
- Anthropic’s Claude reviews that output for accuracy and quality before it’s shown to the user.
- Microsoft expects the workflow to become bi-directional over time, with GPT also reviewing Claude’s drafts.
The company’s framing is straightforward: multiple models working together should help reduce AI “hallucinations,” where systems generate false information, while also speeding up the end-to-end workflow for users.
Why the Multi-Model Approach Matters
For investors, the significance is less about a single feature and more about how Microsoft is positioning Copilot in a crowded market: not as one monolithic model, but as an orchestrator of best-in-class systems that can check each other’s work.
Nicole Herskowitz, corporate vice president of Microsoft 365 and Copilot, said the ability to use different models from different vendors is attractive, and that Microsoft is pushing it further by letting customers benefit from the models “working together.”
In practical terms, these updates target three common friction points in enterprise AI adoption:
- Trust: automated review is meant to catch mistakes before they reach users.
- Consistency: standardized critique could reduce variability across responses.
- Productivity: fewer manual cross-checks can shorten research cycles.
Model Council Brings Side-by-Side Comparisons
Microsoft is also launching Model Council, a feature that allows users to compare responses from different AI models side-by-side. The idea is to make model differences visible inside the workflow, rather than forcing users to test tools separately.
That matters in organizations where teams are still deciding when to rely on AI, when to verify, and which model fits which job. Side-by-side comparisons can also make it easier to build internal standards around how Copilot outputs are evaluated.
- Compare multiple AI responses in one view via Model Council.
- Use comparisons to select the best answer for a given task.
- Create a tighter feedback loop for how teams validate AI-generated work.
Copilot Cowork Expands Through Frontier Early Access
Alongside the Copilot research upgrades, Microsoft is making its Copilot Cowork agentic AI tool more widely available to customers in its Frontier program, which provides early access to newer AI capabilities.
Microsoft introduced Copilot Cowork in testing earlier this month and positioned it as a response to growing demand for autonomous AI agents—tools designed to take action across tasks with less step-by-step prompting.
The company also noted competitive pressure as it works to improve adoption of Copilot, with rivals including Google (GOOGL) and its Gemini offerings, as well as agent-style products such as Claude Cowork.
Investors will watch how quickly these features move beyond early access, and whether multi-model coordination measurably improves user satisfaction and retention in Microsoft 365 workflows for U.S. investors tracking AI monetization.
WSA Take
Microsoft (MSFT) is leaning into a pragmatic idea: AI adoption rises when outputs are easier to trust, not just faster to generate. The “Critique” workflow effectively builds a second set of eyes into Copilot, using Claude to review GPT before results reach the user. Model Council adds transparency by letting teams compare model outputs directly inside the product. The key next step is execution—how smoothly these tools work at scale and how quickly Microsoft can turn better reliability into sustained Copilot usage.
Disclaimer
WallStAccess is a financial media platform providing market commentary and analysis for informational and educational purposes only. This content does not constitute investment advice, a recommendation, or an offer to buy or sell any securities. Readers should conduct their own research or consult a licensed financial professional before making investment decisions.
