Microsoft Build 2025: Unpacking the Agentic Future

In partnership with

Lots of news coming from Microsoft this week as their Build conference kicks off. We participated in many of the events and are highlighting the main points for Day 1 - incorporating key announcements, themes, and explanations of what is coming. Can you say “agentic agents”? 😀 

Head over to the Book of News for all the details. The Microsoft Build 2025 Book of News is your guide to key news items that we are announcing at Microsoft Build.

If someone forwarded this email to you, please consider clicking the button below to join 100K+ other cloud professionals on their cloud certification journey!

Also check out our Learn Azure and Learn AWS apps to help pass those exams on the first try.

Microsoft Build 2025 Recap: Charting the Course to the Agentic Future

Microsoft's Build 2025 conference keynote was a definitive declaration: the company is "doubling down on Agentic AI". Far from just a series of product updates, the event unveiled a full-blown roadmap to the 'agentic' future, signaling a fundamental platform shift towards an "open agentic web at scale". According to Satya Nadella, this moment is akin to pivotal past shifts like the rise of Windows or the cloud, marking the "middle innings" where innovation scales dramatically. The core takeaway is clear: "AI is not a helper anymore; it’s a doer".

Microsoft showcased a comprehensive strategy focused on building AI agents that can autonomously perform tasks, improve workflows, and create new opportunities across products, platforms, and the web.

AI Agents Move to the Forefront

A key theme was the evolution of AI from assistants to capable peers.

GitHub Copilot as a Peer Programmer: GitHub Copilot is stepping up from an assistant to a "peer programmer". It can now autonomously handle tasks such as bug fixes, feature additions, and code maintenance when assigned issues, streamlining the software engineering process. This capability evolves beyond code completion and chat, enabling developers to delegate tasks to a "real virtual teammate".

Multi-Agent Orchestration: A significant advancement is the support for multi-agent orchestration in Copilot Studio. This allows multiple AI agents to collaborate on complex, cross-functional tasks, such as automating employee onboarding across different departments like IT, HR, and marketing. This framework is designed to orchestrate workflows across "every role, every business process".

Unlock AI-powered productivity

HoneyBook is how independent businesses attract leads, manage clients, book meetings, sign contracts, and get paid.

Plus, HoneyBook’s AI tools summarize project details, generate email drafts, take meeting notes, predict high-value leads, and more.

Think of HoneyBook as your behind-the-scenes business partner—here to handle the admin work you need to do, so you can focus on the creative work you want to do.

Try HoneyBook for free

Building the Agentic Ecosystem

Microsoft introduced platforms and protocols designed to empower developers to build and deploy agentic applications.

Microsoft 365 Copilot Tuning: A new low-code tuning tool is now available for Microsoft 365 Copilot. This tool allows organizations to customize Copilot using their specific internal data and workflows, enabling it to learn the company's unique tone, language, and expertise. This allows Copilot to provide answers and generate documents tailored to specific business needs or industry knowledge.

Azure AI Foundry: Positioned as a "production line for intelligence" and the "complete app platform for the AI age", Azure AI Foundry received major upgrades. It now provides access to over 10,000 AI models, including offerings from partners like Grok, Hugging Face, Meta, and Mistral. Developers can select and fine-tune models and build agents, integrating components like sophisticated retrieval systems (RAG) and orchestration layers. Notably, Grok from xAI is coming to Azure AI Foundry. The platform offers enterprise guarantees, including security, compliance, and observability features.

Windows AI Foundry: The local counterpart to Azure AI Foundry (formerly Windows 11 Copilot Runtime) is now Windows AI Foundry. This platform enables model selection and deployment directly on devices (Windows and Mac) and in the cloud. It includes Foundry local, providing a high-performance runtime for developing agents at the edge. Developers can customize built-in models using techniques like Laura adapters and utilize semantic APIs for integrating local data securely into RAG applications.

Connecting the Digital World with Agents

New protocols aim to make interactions with digital services more natural and agent-friendly.

Model Connectivity Protocol (MCP): MCP is highlighted as a crucial standard for the emerging agentic web. Integrated into Azure and Windows 11, MCP allows AI models and agents to securely access and manipulate data across disparate systems like business tools, content repositories, and development environments. MCP servers expose system functions, enabling more advanced agentic workflows. In Windows, it provides native support with built-in servers for core functions like file systems and settings, managed via an MCP registry for control and security. MCP is an "open standard protocol" described as being "akin to HTTP in the internet," designed for reliable, interoperable communication between agents and services.

NLWeb: Natural Language Web Protocol: Microsoft introduced NLWeb, an open standard designed to allow websites to communicate conversationally with users in natural language. This transforms traditional web navigation into more intuitive, conversational experiences. NLWeb is designed to help website owners easily make their sites "an agent application". Thought of as "HTML for the agentic web", every NLWeb endpoint is also an MCP server, making these agent applications accessible to any agent that uses MCP.

Start learning AI in 2025

Keeping up with AI is hard – we get it!

That’s why over 1M professionals read Superhuman AI to stay ahead.

• Get daily AI news, tools, and tutorials

• Learn new AI skills you can use at work in 3 mins a day

• Become 10X more productive

Sign up and start mastering AI

Foundation: Data, Infrastructure, and Security

Supporting the agentic shift requires robust underlying layers.

Data Integration: The data tier is paramount for AI applications. SQL Server 2025, now in public preview, features direct AI integration and built-in vector search capabilities. Cosmos DB is integrated directly into Foundry for managing agent conversational history and RAG applications. Microsoft Fabric, the central data and analytics platform, now incorporates Cosmos DB alongside SQL, enabling organizations to unify structured and semi-structured data for AI readiness. AI-driven ETL capabilities are introduced in OneLink, and Copilot in PowerBI allows users to chat with their data for analysis.

World-Class Infrastructure: Microsoft is taking a systems approach to optimize infrastructure for AI workloads, aiming for "the lowest cost, highest scale infrastructure". This includes deploying cutting-edge hardware like Nvidia GB200s on Azure at scale, expanding their global data center footprint, and building a high-capacity AI WAN backbone. Innovations like ARM-based Cobalt VMs and Azure Local for low-latency scenarios are also part of this effort.

Security and Compliance: Enterprise-grade security is built-in. Security tools like Entra, Defender for Cloud, and Purview are embedded in Azure AI Foundry and Copilot Studio. A major announcement was Entra Agent ID, which provides agents with their own identity, permissions, policies, and access controls, extending existing identity management rails to AI agents. Purview integration ensures end-to-end data protection for agents, and Defender protects them from threats.

Commitment to Openness and Real-World Impact

Microsoft emphasized its commitment to open source and showcased AI's real-world applications.

Open Source: Core components of the Windows Subsystem for Linux (WSL) are now open source under the MIT license, a move requested by the community nearly a decade ago. The GitHub Copilot Chat Extension for VSCode is also being open-sourced. The release of NLWeb as an open standard further reinforces this approach, aiming to foster an open ecosystem.

Demonstrated Impact: Real-world use cases included the NFL Combine app for player analytics, the Met Office using AI for advanced weather forecasting, and Stanford Medicine using multi-agent orchestration in Foundry to improve cancer care tumor boards. A significant new platform, Microsoft Discovery, was unveiled for accelerating scientific R&D, using specialized agents and graph-RAG to speed discoveries from months to days or hours. The potential for AI's broader societal impact was also highlighted, such as a World Bank study showing positive educational outcomes using Copilot with students and teachers in Nigeria and Peru.

Learn AI in 5 minutes a day

What’s the secret to staying ahead of the curve in the world of AI? Information. Luckily, you can join 1,000,000+ early adopters reading The Rundown AI — the free newsletter that makes you smarter on AI with just a 5-minute read per day.

Sign up to start learning.

Build 2025 painted a compelling picture of an "agentic" future where AI agents are powerful, capable, and integral to workflows, powered by a comprehensive platform spanning infrastructure, data, and application layers, all built with a focus on openness and security. The message to developers is clear: the tools and opportunities to build this new wave of AI-powered applications are here.

Until next week, keep learning and building!

Sponsor Us:

Want to reach 100,000+ cloud and data enthusiasts? Sponsor our newsletter and gain valuable exposure for your brand! Send us an email to learn more.