Microsoft Developer Conference: 50 new products fully bet on Agent

On May 19, local time, the four-dayMicrosoft corporation The Build annual developer conference officially kicked off with CEO Satya Nadella delivering a keynote address with a group of executives, and also connecting with OpenAI CEO Sam Altman, NVIDIA CEO Jen-Hsun Huang, and xAI CEO Musk.

This keynote speech, Microsoft officially put Agentic AI in the center of the spotlight, announcing Microsoft's future strategy in the field of AI: fully bet on Agentic AI (Agentic AI), to become the core of the future AI agent ecosystem.

Nadella and other Microsoft executives together painted a blueprint about the future of AI intelligence:AI is more than just a collaboration tool, it's a trusted digital colleague that can think independently, perform tasks, and enable end-to-end integration across multiple technology stacks.

Figure | Copilot agent opening (Source: Microsoft)

"We've entered the age of the AI agent." Frank Shaw, Microsoft's chief communications officer, pronounced, "The number of people using agents has exploded over the past year, doubling the number of daily active users."

What is Agentic AI, as defined by Microsoft CTO Kevin Scott in his keynote:"An intelligent body that can be entrusted with tasks by humans with confidence."

In his view, Agentic AI is no longer limited to simple input-output interactions, but is an AI entity that can autonomously perceive the environment, formulate strategies, disassemble tasks and execute them. The doubling of the number of daily users of such AI tools seems to indicate that the tipping point for Agentic AI to be applied on a large scale has arrived.

Based on this, at the Build conference, Microsoft showed its comprehensive layout in the AI tool chain, cloud services, productivity platform, and developer ecosystem.

Figure | Microsoft-supported Foundry models, DeepSeek among them (Source: Microsoft)

The company has introduced more than 50 innovative releases across its entire product line.From GitHub and Azure to Windows and Microsoft 365, all of these are focused on advancing "AI agent technologies that can solve complex business problems independently or collaboratively".

Together, Microsoft has given it a new name: the open agentic web - a new vision of the Internet powered by AI agents.

Figure | Open Agent Web Vision (Source: Microsoft)

For developers, Microsoft has chosen to put GitHub Copilot at the forefront of its agent strategy: it's no longer just a "pair programmer" but a "peer programmer". programmer.

In the demo, with the new coding agent, developers can delegate tasks directly to Copilot, including fixing bugs, developing new features, refactoring code, maintaining projects, and more, and Copilot automates all the processes.

The GitHub Copilot agent allows developers to focus on other tasks while ensuring that their previously written code works properly. Meanwhile, Microsoft has announced that it will open source GitHub Copilot Chat in VS Code, enabling the developer community to contribute to its evolution.

Microsoft also announced the launch of the Azure SRE agent (Site Reliability Engineering Intelligence), which will be embedded in GitHub Copilot to help developers deal with failures in production environments.

For example, when website anomalies occur, the AI agent can sense and locate the problem at the first time, and complete the repair without human intervention, reducing the burden of engineers oncall late at night.

Nadella said, "This marks a leap in the way development is done. In the past we've been talking about AI as your assistive tool, now you can actually assign tasks to it, like working with a seasoned colleague."

(Source: Microsoft)

For enterprise users, Microsoft has introduced a new feature called Copilot Tuning. This feature allows organizations to quickly inject their own internal data into an AI agent to create enterprise-grade, customized intelligences that "talk, think, and execute.

At the core of this tool is the fine-tuning of data context to make AI a "digital employee" that truly understands the business, processes, and terminology of an organization, for scenarios such as customer service, sales, knowledge management, project management, and more. Microsoft confirmed that the program has been known internally as Tenant Copilot, or Tenant Level Copilot.

Microsoft said that the goal of a series of Copilot updates is to embed AI capabilities into "every layer of the software development process", so that developers can not only use AI to write code, but also use AI to manage the code, deploy services, maintain the system, and ultimately, AI as the software industry's "new infrastructure. The ultimate goal is to make AI the "new infrastructure" of the software industry.

Some other important updates for AI agents relate to AI Foundry, a platform for developing and managing AI applications and agents.

Figure | The newly launched Windows AI Foundry, which makes it easy for developers to deploy AI models and applications locally (Source: Microsoft)

On Azure AI Foundry in the cloud, Microsoft has added several new features and security metrics around AI agents to make it easier for developers to securely manage agents.

And with the newly launched native version of Windows AI Foundry, Microsoft is providing a unified platform for native AI development on Windows. It includes Windows ML, a built-in AI inference runtime, and tools for preparing and optimizing models for deployment on devices. This will make it easy to run AI models, tools, and agents directly on devices.

With the vision of an open agent-based web, Microsoft also announced the NLWeb open project, which aims to simplify the creation of natural language interfaces for websites and make it easy to transform any website into an AI application.

(Source: Microsoft)

Microsoft is positioning NLWeb as the HTML of the agent web space, providing a standardized way for websites and developers to have their content discovered and used by AI agents. As a result, users will be able to query website content directly in natural language, just as they would with an AI assistant or Copilot.

NLWeb takes semi-structured formats such as Schema.org, RSS, and other data already published on the site and combines them with big-model-driven tools to create natural language interfaces that can be used by humans and AI agents.

"Each NLWeb instance is also itself a model context protocol (MCP) server that allows sites to serve their content to agents and other participants in the MCP ecosystem." Microsoft explains, "It can augment this structured data by integrating external knowledge from the underlying big model, such as adding geographic knowledge to restaurant queries, to provide a richer user experience."

Finally, Microsoft announced at the conference a new enterprise agent platform called Microsoft Discovery to accelerate research and development (R&D) for researchers.

Figure | Microsoft Discovery, the new enterprise agent platform (Source: Microsoft)

Using AI models and high-performance computing simulation tools in Microsoft Discovery, Microsoft researchers discovered a new coolant prototype for immersion cooling in data centers in about 200 hours, a process that would have taken months or even years.

"Our goal is to bring the power of AI to scientists and engineers to revolutionize the entire process of discovery - from advanced knowledge reasoning and hypothesis formulation, to experimental simulation and iterative learning." Microsoft said.

In addition to Microsoft's own product launches, CEOs from three of the AI industry's top companies - OpenAI, NVIDIA and xAI - also connected remotely.

OpenAI CEO Sam Altman introduced Codex, the new programming intelligence released a few days ago, "It's one of the biggest technological changes I've ever seen impact programming." In internal testing, OpenAI has found that developers using Codex are significantly more productive and have fundamentally changed the way they work.

Figure | Musk connects with Nadella (Source: Microsoft)

Musk, in his capacity as CEO of xAI, introduced Grok 3 and Grok 3 Mini, two large models that have been connected to Microsoft Azure and can be called directly by developers on Azure. At the same time, Musk also revealed that Grok 3.5 is coming soon.

This series of collaborations reflects the fact that Microsoft is building Azure into a global compute base for AI agents to run on to support a network of intelligences where multiple models coexist and multiple ecosystems converge.

This shows that Microsoft's AI strategic positioning for itself is gradually moving closer to a central role in the emerging AI agent ecosystem.

References:

• https://build.microsoft.com/en-US/home
• https://blogs.microsoft.com/blog/2025/05/19/microsoft-build-2025-the-age-of-ai-agents-and-building-the-open-agentic-web/
• https://www.tomsguide.com/news/live/microsoft-build-2025