At its annual Build conference in 2025, Microsoft introduced a comprehensive strategy to position itself at the center of what it calls the “open agentic web”. The company unveiled over 50 AI tools and platforms designed to help developers create autonomous systems that can make decisions and complete tasks with limited human intervention.
What are AI Agents?
According to Microsoft, we have entered the “era of AI agents”. These agents go far beyond today’s AI assistants that primarily respond to human questions and commands. Agents actively initiate tasks, make decisions independently, coordinate with other AI systems, and complete complex workflows with minimal human supervision. This marks a fundamental shift in how AI systems operate and interact with users and other technologies. AI agents are described as semi- or fully autonomous pieces of AI software that can perform certain tasks for users, including moving data between different apps or booking tickets. Some can even communicate with each other, creating a web of agents that can complete more complex tasks.
Here’s a look at some key takeaways from Microsoft’s Build 2025 announcements:
GitHub Evolution: GitHub Copilot is evolving from code suggestion to a coding agent that can autonomously solve programming tasks. This agent can operate as a member of a software development team, autonomously refactoring code, improving test coverage, fixing defects, and even implementing new features. For complex tasks, it can collaborate with other agents across all stages of the software lifecycle. Microsoft is also open-sourcing GitHub Copilot Chat in Visual Studio Code, with AI-powered capabilities from the Copilot extensions becoming part of the VS Code open-source repository. Agentic DevOps, where agents automate and optimize software development, will be layered into GitHub Copilot and Microsoft Azure.
Multi-Agent Systems: Microsoft emphasizes the importance of multi-agent systems for enabling complex business workflows and process automation. Splitting tasks across multiple agents improves maintainability, makes building solutions easier, and significantly enhances reliability. Debugging and drilling down into these multiple agents is key. The Azure AI Foundry Agent Service, now generally available, allows developers to build enterprise-grade AI agents with support for multi-agent workflows and open protocols like Agent2Agent (A2A) and Model Context Protocol (MCP). Copilot Studio will get support for multi-agent systems to delegate tasks.
Local AI Capabilities: Microsoft is making a significant push toward local, on-device AI. Windows AI Foundry, an evolution of Windows Copilot Runtime, provides a unified platform for local AI development on Windows. It allows running AI models, tools, and agents directly on devices like Windows 11 or MacOS. Leveraging ONNX Runtime, Foundry Local is designed for scenarios where users can save on internet data usage, prioritize privacy, and reduce costs. Developers can bring their own models and deploy them across various platforms. Client-side AI has advanced remarkably fast.
Security and Identity Management: To address the critical needs of enterprise security, governance, and compliance, Microsoft introduced new capabilities. Microsoft Entra Agent ID, now in preview, assigns unique identities in an Entra directory to agents created in Microsoft Copilot Studio or Azure AI Foundry, helping organizations securely manage agents from the start and avoid “agent sprawl” that could lead to blind spots. Microsoft is also integrating its Purview data security and compliance controls with its AI platforms. Copilot Studio offers a managed infrastructure framework with lifecycle management and governance/observability capabilities built in.
Open Protocols and Standards: Advancing open standards that enable agent interoperability across different platforms and services is central to Microsoft’s vision. Microsoft offers broad support for the Model Context Protocol (MCP) and Agent2Agent (A2A). NLWeb is a new open project akin to HTML for the agentic web, allowing websites to provide conversational interfaces. MCP will be broadly supported across Microsoft’s developer stack, including on Windows 11 (in early preview). Teams will also support A2A protocol, allowing agents to exchange messages, data, and credentials without intermediaries.
Memory and Context: A critical missing element in agents is memory. Microsoft is introducing several memory-related technologies, including structured RAG (Retrieval-Augmented Generation), which helps AI systems more precisely recall information from large volumes of data. Contextual awareness is crucial for creating agents that understand the user well and anticipate their needs.
Scientific Discovery Platform: Microsoft Discovery is a platform designed to accelerate scientific research and development. It was used to discover a non-PFAS immersion coolant for data centers in just 200 hours – a process that traditionally takes years or even a decade. Using this framework, 367,000 potential candidates were screened in 200 hours. This represents a dramatic acceleration of traditional R&D timelines.
Microsoft sees the world transitioning from a web of information to an “agentic web”, where machines don’t just respond to commands but anticipate needs, make decisions, and fundamentally reshape how work gets done. Microsoft’s approach—combining cloud and edge AI, open standards with proprietary technologies, and developer tools with business applications—positions the company as a central player in the emerging agentic ecosystem. Some 230,000 organizations are already using Copilot Studio to develop custom AI agents, and Microsoft expects businesses to deploy around 1.3 billion AI agents by 2028. The immediate impact for enterprises may be most visible in increased automation of complex workflows, more intelligent responses to business events, and the ability to build custom agents.
No comment