Build 2025: Microsoft Expands Copilot for Microsoft 365 with Copilot Tuning, Multi-Agent Orchestration
Custom AI training, multi-agent collaboration, and new developer tools are coming to Microsoft 365 Copilot.
Key Takeaways:
- Microsoft is launching Copilot Tuning in June, allowing businesses to customize AI agents using their own data—without coding.
- New developer tools and security features are being added to Copilot Studio for more powerful and secure AI experiences.
- Multi-agent collaboration and enhanced data integration are reshaping how AI assistants work across Teams and Microsoft 365 apps.
Microsoft 365 Copilot is about to get a major productivity boost for commercial users with the upcoming launch of Copilot Tuning. Rolling out in June, Copilot Tuning is a low-code solution that lets businesses train AI models and build custom agents using their company data in Copilot Studio.
“With Copilot Tuning, you can use your company data, workflows, and processes to train models and create agents that perform domain-specific tasks with a high degree of accuracy using Copilot Studio—no coding required. For example, a legal firm can create an agent to reflect its unique voice and expertise—automating document creation and even drafting arguments that blend institutional knowledge with client-specific context to help build the strongest case possible,” Jared Spataro, Chief Marketing Officer, AI at Work.
In June, Microsoft will begin rolling out Copilot Tuning as part of the Microsoft 365 Copilot Tuning Early Adopter Program. However, this feature will only be available for customers with 5,000 or more Microsoft 365 Copilot licenses. Microsoft says that it doesn’t use customer data to train AI models, and its agents and data operate securely within the Microsoft 365 service boundary.
Microsoft Copilot Studio has introduced a new feature in preview called multi-agent orchestration, which is part of a broader shift toward human-agent collaboration. This new capability allows multiple AI agents to exchange data, collaborate on tasks, and divide work based on each agent’s expertise. When a customer asks a question that is related to all three areas, these agents collaborate behind the scenes and then present a unified response.
New pro-code features coming to Microsoft Copilot Studio
Microsoft has also announced new tools to help developers create powerful AI agents. Azure AI Foundry Models integration for prompts and generative answers allows developers to bring their own models to Copilot Studio. This feature is currently available in preview and provides access to more than 1,900 models.
Microsoft Entra Agent ID automatically assigns a secure identity to any agent built using Copilot Studio and Azure AI Foundry. This feature eliminates the need for developers to configure it manually. This gives security administrators full visibility and control over the agent’s access and activity.
In June, Microsoft will extend Purview data loss prevention (DLP) controls to Microsoft 365 Copilot agents to protect sensitive data. This security feature allows IT admins to prevent sensitive content within a labeled document from being pasted into other apps or processed by Copilot for grounding data.
Microsoft will add Purview Information Protection capabilities to Copilot Studio agents that use Dataverse. Organizations can leverage this security feature to automatically classify and protect sensitive data at scale.
The new computer use tool lets agents automate tasks on desktop applications and websites based on natural language commands. Currently, this feature is available via the Frontier program for eligible customers with 500,000+ Copilot Studio messages and an environment in the United States.
Microsoft has also announced the general availability of the new Model Context Protocol (MCP) within Copilot Studio. This feature helps to ensure that agents have governed access to external data and models. With deep reasoning, agents can run complex and multi-faceted business processes.
Copilot Studio gets new tools to give developers the freedom to build their way
The Microsoft 365 Agents Toolkit is a set of tools designed to help developers build AI-powered agents. This toolkit includes the Microsoft 365 Agents SDK, which enables developers to test, debug, and deploy AI agents across Copilot, Teams, the web, and Agent Playground, using smart defaults for Azure. With the enhanced Teams AI library and support for open standards like the Agent2Agent Protocol and MCP, developers can now build collaborative agents optimized for Teams chats, channels, and meetings.
The Microsoft 365 Copilot APIs are a set of enterprise-grade tools that allow developers to create fast, context-aware, secure, and generative AI (GenAI) experiences. One key feature is the Retrieval API, which helps developers fetch relevant Microsoft 365 data to power intelligent chat and search experiences. The Retrieval API is currently available in public preview, with others to follow later this year.
Lastly, Solution Workspace in Power Apps enables developers to collaborate with agentic AI to define business requirements, generate data models, process maps, and design solution architecture. Microsoft has also announced the public preview of a new feed agent for apps. It’s a new hub that can be configured by developers for end users to view, manage, and monitor agents within Power Apps.
Rabia has a master's degree in Software Engineering and she has years of experience writing professionally about Microsoft products and other technologies. Rabia has also written for OnMSFT.com as well as Windows Report. She is always up to date on t...
