At Microsoft Build 2025, the tech giant showcased its bold vision for a future powered by AI agents — digital tools capable of reasoning, remembering, and independently executing tasks on behalf of users and organisations. The message was clear: AI is no longer an assistant; it’s an agent of transformation.
“Fifteen million developers are already using GitHub Copilot,” Microsoft shared, highlighting the rapidly growing adoption of agent-driven tools that streamline coding, reviewing, deploying, and troubleshooting. Microsoft 365 Copilot, already in use by hundreds of thousands, is helping users brainstorm, research, and execute solutions more efficiently.
Enterprise adoption is surging, too. According to Microsoft, over 230,000 organisations — including 90% of Fortune 500 companies — have tapped into Copilot Studio to build customised AI agents and automations. Companies like Fujitsu and NTT DATA are leveraging Azure AI Foundry to build apps that prioritise leads and surface client insights. Meanwhile, Stanford Health Care is exploring AI agents to speed up administrative workflows for tumour board preparations.
Tools of the Trade: New Platforms and Smarter Agents
At the heart of this AI leap is Microsoft’s suite of new and upgraded platforms designed for developers. GitHub Copilot is evolving into a full-fledged AI partner, with Microsoft open-sourcing Copilot Chat in Visual Studio Code. This shift reinforces Microsoft’s commitment to a collaborative and open developer ecosystem.
Windows AI Foundry was also announced as a one-stop shop for training and deploying AI models. With support for open-source large language models and proprietary tools, developers can now build, fine-tune, and deploy seamlessly across cloud and client environments.
In Azure, the AI Foundry Models platform is now hosting Grok 3 and Grok 3 Mini from xAI. Developers will have access to over 1,900 models, alongside new tools like a Model Leaderboard and Model Router that help them choose and optimise models in real time.
“Developers are using AI to stay in the flow of their environment longer and to shift their focus to more strategic tasks,” Microsoft noted.
Building Safely and Openly for the Agentic Web
Microsoft emphasised its commitment to making AI not only powerful but also secure. Azure AI Foundry Agent Service now supports multi-agent orchestration, bringing together tools like Semantic Kernel and AutoGen under one SDK. New observability features offer built-in performance, cost, and safety tracking.
Microsoft Entra Agent ID, now in preview, assigns unique identities to each AI agent, helping enterprises manage their AI fleet securely. Tools like Microsoft Purview Compliance Manager further extend governance and risk control.
The company also lifted the veil on Copilot Tuning, which allows businesses to train agents using their own data, with applications ranging from legal document generation to multi-agent collaboration for complex tasks.
Meanwhile, Microsoft is championing open standards for the agentic web. It announced first-party support for Model Context Protocol (MCP) across its platforms and introduced NLWeb — a new open project designed to let websites create conversational interfaces that interact directly with users’ chosen models and data.
AI Meets Science and the Future
Finally, Microsoft introduced Microsoft Discovery, a platform designed to empower researchers with agentic AI across domains such as drug development and sustainability. This tool aims to compress the discovery timeline and expand the scope of scientific research across industries.
With all eyes on Build 2025, Microsoft’s vision for an open, agent-powered web is more than just talk — it’s being coded into reality.