Microsoft announces over 50 AI tools to build the ‘agentic web’ at Build 2025

# azure # nlp

Microsoft Unveils Vision for the Open Agentic Web at Build Conference

This morning, Microsoft kicked off its annual Build conference with a bold declaration: the dawn of the "open agentic web." In a sweeping array of over 50 announcements, the tech giant outlined a comprehensive strategy to place itself at the heart of this transformative movement. From GitHub and Azure to Windows and Microsoft 365, every product line received updates aimed at advancing AI agent technologies. These agents promise to solve complex business challenges independently or in collaboration, marking a significant leap forward in how artificial intelligence operates. Frank Shaw, Microsoft's Chief Communications Officer, captured the essence of this shift in a blog post: "We've entered the era of AI agents. Thanks to breakthroughs in reasoning and memory, these systems are becoming more capable and efficient. They're opening doors to solving problems in entirely new ways."

Revolutionizing Software Development Through Autonomous AI

AI agents represent much more than enhanced versions of today's assistants. Unlike conventional AI tools that merely react to human prompts, these agents proactively initiate actions, make decisions autonomously, and even collaborate with other systems to complete intricate workflows with minimal oversight. This evolution signifies a seismic shift in how AI interacts with both users and other technologies. During a press conference, Kevin Scott, Microsoft's CTO, likened this transformation to reimagining the relationship between humans and technology. "Reasoning will keep improving," he noted. "But we're on the cusp of some critical developments that must happen soon for agents to handle more sophisticated tasks." One glaring gap, according to Scott, lies in memory. "Agents currently lack the kind of persistent memory that would allow them to truly understand context," he explained. To bridge this gap, Microsoft introduced several memory-centric innovations, such as structured RAG (Retrieval-Augmented Generation), which empowers AI systems to retrieve and apply information from vast datasets more accurately. Steven Bathiche, CVP and technical fellow at Microsoft, elaborated further: "Imagine having separate agents for personal and professional use. The work agent would hold valuable organizational data, offering deeper insights tailored to your role and responsibilities. This contextual awareness is key to creating agents that intuitively cater to your needs."

Github Evolves Beyond Code Suggestions

At the forefront of Microsoft's agentic strategy sits GitHub, its beloved developer platform. With the launch of GitHub Copilot, the platform transcends mere code suggestions to deliver autonomous coding assistance. Now, this coding agent can act as a full-fledged team member, handling tasks like refactoring code, enhancing test coverage, bug fixes, and feature implementation. For more complex projects, GitHub Copilot can collaborate with other agents throughout the entire software development lifecycle. Moreover, Microsoft has open-sourced GitHub Copilot Chat in Visual Studio Code, inviting the developer community to contribute to its growth. This dual approach—driving innovation while embracing open-source principles—underscores Microsoft's commitment to fostering collaborative progress.

Multi-Agent Systems Streamline Enterprise Workflows

For businesses eager to adopt AI agents, Microsoft's Azure AI Foundry offers substantial enhancements. Ray Smith, VP of AI Agents at Microsoft, highlighted the significance of multi-agent systems in an interview with VentureBeat. "Splitting tasks among multiple agents is vital," he emphasized. "Not only does this improve maintainability and simplify solution-building, but it also boosts reliability significantly." Azure AI Foundry Agent Service is now widely available, enabling developers to craft enterprise-grade AI agents equipped with support for multi-agent workflows and open protocols like Agent2Agent (A2A) and Model Context Protocol (MCP). These capabilities empower organizations to coordinate multiple specialized agents to tackle intricate challenges.

From Cloud to Edge: On-Device AI Advances

While cloud-based AI continues to dominate discussions, Microsoft is steering attention toward local, device-level AI. Several initiatives aim to assist developers deploying AI directly onto user devices. Windows AI Foundry, an extension of Windows Copilot Runtime, provides a unified framework for local AI development on Windows. It includes Windows ML, an integrated AI inference runtime, along with tools for optimizing models for on-device execution. "Foundry Local simplifies running AI models, tools, and agents directly on devices, regardless of whether it's Windows 11 or macOS," the company stated. Leveraging ONNX Runtime, Foundry Local prioritizes privacy, reduces costs, and conserves bandwidth. Steven Bathiche, during a presentation, remarked on the rapid pace of client-side AI advancements: "We're constantly adapting to new trends. Predictions we make now often materialize within months, which is exhilarating yet demanding. Still, it keeps us energized!"

Security and Identity Management Address Enterprise Challenges

As AI agent adoption grows, so do concerns around security and compliance. Microsoft addressed these pressing issues with several features aimed at preventing "agent sprawl," a term referring to uncontrolled proliferation of agents within organizations. "Microsoft Entra Agent ID, currently in preview, assigns unique identities to agents created in Microsoft Copilot Studio or Azure AI Foundry," the announcement noted. This ensures enterprises can securely manage agents from inception, avoiding blind spots caused by unchecked expansion. Additionally, Microsoft integrated Purview data security and compliance controls with its AI platforms, equipping developers with enterprise-grade safeguards. New capabilities for detecting sensitive data in AI interactions further bolster these efforts.

Accelerating Scientific Research Through AI Agents

One of the most ambitious applications showcased at Build involves Microsoft Discovery, a platform designed to expedite research and development across industries like pharmaceuticals and materials science. Jason Zander, CVP of Strategic Missions & Technologies at Microsoft, shared in an interview with VentureBeat how this platform identified a non-PFAS immersion coolant for data centers in just 200 hours—a process that typically spans years. "The traditional timeline for such discoveries can stretch into decades," Zander explained. "With this framework, we screened 367,000 candidates in under 200 hours. Partnering with others allowed us to synthesize those findings efficiently."

Standards for Interoperable Agents Across Platforms

A cornerstone of Microsoft's vision revolves around establishing open standards for agent interoperability across platforms and services. The Model Context Protocol (MCP) plays a pivotal role here. Microsoft joined the MCP Steering Committee and contributed two notable additions: an updated authorization specification and a design for an MCP server registry service. Jay Parikh, leader of Microsoft's Core AI team, underscored the importance of openness: "Inside Microsoft, fostering rapid learning is paramount. Embracing speed aligns with the dynamic nature of technology, competition, and application landscapes." Introducing NLWeb, a novel open project, Microsoft envisions it functioning akin to HTML for the agentic web. By enabling websites to provide conversational interfaces tailored to user preferences and data, NLWeb opens exciting possibilities for interaction models.

Positioning Microsoft at the Forefront of the Next Computing Paradigm

Microsoft's expansive announcements at Build 2025 highlight its unwavering dedication to AI agents as the next frontier in computing paradigms. Kevin Scott, reflecting on his enthusiasm, recalled the excitement of the '90s when building blocks felt comprehensible and creative. Analysts observe Microsoft's holistic approach—combining cloud and edge AI, open standards alongside proprietary solutions, and developer tools with business applications—as positioning the company centrally within the burgeoning agentic ecosystem. For enterprises, the initial impact may manifest through heightened automation of intricate workflows, smarter responses to business events, and the ability to construct bespoke agents incorporating domain expertise. As we transition from an information-centric web to an agent-driven one, Microsoft's strategy mirrors its earlier trajectory in cloud computing—providing robust tools, platforms, and infrastructure while advancing open standards. The pertinent question now isn't whether AI agents will alter business operations but rather how swiftly organizations can adapt to a world where machines anticipate needs, make informed decisions, and redefine work dynamics.

Related article

Comments (5)