2 weeks ago
2
With the new Copilot Tuning feature, users can train agents to perform specific tasks on behalf of their organization using natural language, no coding required.
The great personalization of AI is underway.
On Monday, during Build, its annual developers conference, Microsoft made some announcements that helped bring into focus the company's strategic vision for the future of its tailor-made AI systems.
Also: How Microsoft Outlook's new AI-powered features reduce tedium, increase security
Copilot Tuning
Microsoft introduced Copilot Tuning, a new feature that enables enterprise customers to train custom models and agents using internal company data and processes to perform specific tasks, no coding required.
"For example," explained Jared Spataro, chief marketer of Microsoft's AI at Work division, in a company blog post, "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."
According to Spataro, Microsoft will not use proprietary customer data to train its foundation models. The feature will be available starting in June as part of the Microsoft 365 Copilot Tuning Early Adopter Program.
Multi-Agent Orchestration
The company also debuted Multi-Agent Orchestration, which is available in public preview. This technology synchronizes individual agents into a single network, allowing them to share information and work with one another to complete tasks. Such inter-agent collaboration is designed to make it even easier for employees to offload tasks to AI.
Finally, there is a suite of new tools that give developers more control over the fine-tuning of agents built in Copilot Studio:
Also: I retested Microsoft Copilot's AI coding skills in 2025 and now it's got serious game
- Azure AI Foundry Models: Customers can bring their own AI models into Copilot Studio with the Azure AI Foundry Models integration feature.
- Entra Agent ID: This new tool automatically assigns new agents with an identity, or organization-specific security clearance and oversight controls.
- Purview Information Protection strengthens data security across agents.
- Microsoft 365 Agents Toolkit enables users to build agents using whichever AI stack they prefer. The toolkit includes a software development kit (a kind of virtual testing ground for agents) and the Teams AI Library, in which agents can be built specifically to operate within Teams, Microsoft's video conferencing and collaboration software.
- Microsoft 365 Copilot APIs allow developers to embed data from Copilot Chat into their organization's internal apps and digital infrastructure.
- Solution Workspace provides a single, integrated digital environment where developers can build apps via natural language prompts.
- Agent Feed acts as a direct interface through which developers can keep an eye on agents and their performance across tasks.
Custom AI for all
Not so long ago, tech developers only built models that worked exactly the same for everybody everywhere. Now, there's a growing focus on customization, specialization, and personalization. Every individual and organization now has access to their own custom-trained AI tools tailored to their particular data and needs.
Microsoft has been especially keen on this approach. The company has, for example, been a major proponent of so-called small language models -- miniature systems that, unlike their larger counterparts, are relatively cheap to train and designed to operate within the closed data ecosystems of specific businesses.
Also: The best AI chatbots: ChatGPT, Copilot, and notable alternatives
The company has also been pouring R&D money into AI agents, systems designed to communicate in natural language and autonomously perform tasks on behalf of human users. In late 2023, the company introduced Copilot Studio, a platform on which customers can build their own custom agents (or "Copilots," as Microsoft calls them).
In small ways, then, each of these new features points to what's become a core part of Microsoft's strategy for commercializing AI: building and deploying tools designed not for everyone but for specific individuals and organizations. The company is banking on this tailored approach to set it apart in the ongoing AI arms race.
"Today's announcements further our ambition to give developers the tools they need to empower every employee with a Copilot," Spataro wrote.
Want more stories about AI? Sign up for Innovation, our weekly newsletter.
English (US) ·