Agentic AI

In this edition of our monthly roundup, we’re highlighting a few of our biggest updates from Microsoft Ignite 2025 and walking through new capabilities available today.

Explore new Microsoft Copilot Studio updates to shape agent behavior, enforce organizational standards, and support agentic business transformation.

Today, we are making GPT-5.1 available in Microsoft Copilot Studio, alongside OpenAI’s release. This is available as an experimental model for U.S. customers in early release cycle Power Platform environments. The GPT-5.1 series brings improved adaptability in thinking time in both chat and reasoning.

In this edition of our monthly roundup, we’re recapping new features released in Microsoft Copilot Studio in October 2025.

Automated agent testing is now built into Copilot Studio—evaluate performance, improve quality, and scale confidently with Agent Evaluation.

Introducing the Copilot Credit Pre-Purchase Plan (P3)—a flexible way for organizations to purchase Copilot Credits and scale their agent initiatives with confidence.

In this edition of our monthly roundup, we’re recapping new features released in Microsoft Copilot Studio in September 2025.

Claude Sonnet 4.5 – Anthropic’s latest and most capable model – is rolling out today in Microsoft Copilot Studio.

This post has been updated as of 9/29/25.  We announced on September 29th, 2025, that Microsoft Copilot Studio now supports Anthropic’s latest model, Claude Sonnet 4.5. You can learn more here. Starting today, Anthropic models are rolling out alongside OpenAI models in Microsoft Copilot Studio.

Starting today, customers with U.S.-based environments can add computer use to their agents directly as a tool – meaning agents can work directly with websites and applications.

Introducing a powerful new capability in Copilot Studio: automatic, intelligent decision-making steps you can build into multistage approval processes.

For automated workflows that still need human oversight, we’ve launched a powerful new capability that bridges the gap between agent-led efficiency and human insight.