Microsoft Build 2025 So Far: Innovations Unveiled

This year’s announcements at Microsoft Build 2025 signal a major leap forward for developers, makers, and enterprises embracing AI-powered solutions. With over 50 updates across the Microsoft ecosystem, Build 2025 showcases the future of agentic workflows, AI development, enterprise automation, and secure app innovation.

 

Key Takeaways

  • Copilot Tuning allows organisations to fine-tune AI models using proprietary data.
  • Agent2Agent protocol brings secure, decentralised agent collaboration.
  • Microsoft Copilot Studio gains powerful pro-code and low-code tools.
  • Azure AI Foundry introduces cutting-edge models, tools, and orchestration features.
  • Microsoft 365 Copilot Wave 2 includes powerful AI enhancements in Outlook and Pages.
  • GitHub Copilot evolves into an autonomous, team-integrated coding agent.
  • Edge browser APIs and on-device AI make web AI development easier and more private.
  • New security controls in Microsoft Purview and Entra protect enterprise AI environments.

 

AI at Work

Copilot Tuning and Domain-Specific Intelligence

Microsoft is introducing Copilot Tuning, enabling enterprises to train AI agents using internal knowledge to solve domain-specific challenges. This allows, for example, a legal team to generate documents aligned with firm-specific language, or consultants to answer industry-specific questions based on regulatory documents.

  • Copilot Tuning becomes available in June for customers with 5,000+ Microsoft 365 Copilot licenses.
  • Agent Builder in Microsoft 365 Copilot allows enterprises to deploy tuned agents with contextual awareness.

Collaborative and Smarter Agents in Microsoft Teams

The future of work is collaborative between humans and intelligent agents. Microsoft Teams is evolving with new development tools to create smarter, secure agents.

  • Agent2Agent (A2A) protocol supports secure, decentralised agent communication.
  • Updated Teams AI Library in JavaScript and C# for building collaborative agents.
  • Agentic memory supports personalised experiences based on past interactions.
  • Developer portal analytics enable real-time performance insights.
  • Meeting agent capabilities allow building agents that act in real-time via Insights API.

Find out about our AI Solutions here.

 

Microsoft Copilot Studio

Pro-Code Tools for Enterprise-Scale Agents

Developers now have access to an expanded pro-code toolkit to build multichannel, secure agents that integrate across Microsoft 365 and beyond.

  • Microsoft 365 Agents Toolkit integrates Copilot Studio and Azure AI tools.
  • Copilot APIs enable secure, retrieval-augmented generative AI.
  • Agent Store centralises deployment and discovery of custom agents.
  • BYOM (Bring Your Own Models) allows use of 1,900+ Azure AI Foundry models.
  • Enhanced Power Platform connector SDK simplifies building structured data integrations.

Empowering Makers with Natural Language Tools

Low-code developers get a significant upgrade with Copilot Studio tools that enhance collaboration and simplify agent creation.

  • Multiple agent system support allows agents to work in coordinated workflows.
  • Model Context Protocol (MCP) ensures agents access governed data.
  • Computer Use Tool uses AI vision to automate desktop and web tasks.
  • Code Interpreter lets agents execute Python code for advanced reasoning.
  • Dataverse search and operational database create real-time, knowledge-rich agents.

 

Microsoft 365 Copilot

A New Wave of AI Productivity Tools

The Wave 2 update transforms everyday workflows with deeper AI integration across Microsoft 365.

  • New Copilot App supports human-agent collaboration.
  • Copilot Notebooks, Search, and Memory roll out with advanced capabilities.
  • Researcher and Analyst agents leverage reasoning for enterprise users.

AI in Outlook and Copilot Pages

Outlook and Pages now feature smarter summarisation, preparation, and content generation.

  • Summarise search results and files in Outlook.
  • Turn Copilot chats into editable Pages with charts and code blocks.
  • One-click convert Pages into Word documents.

 

Azure AI Foundry

A Factory for Enterprise AI Agents

Azure AI Foundry has become a central hub for building and scaling intelligent agents.

  • New models like Grok 3 and Flux Pro 1.1 expand choice.
  • Foundry Agent Service enables multi-agent orchestration.
  • Foundry Observability tracks performance, cost, and safety with precision.
  • Model Router automates optimal model selection and A/B testing.

From On-Device to Secure Development

Foundry also supports powerful local development and advanced security tooling.

  • Foundry Local brings AI inferencing to Windows and macOS devices.
  • Prompt Shields, Task Adherence, and PII detection enhance AI safety.
  • Developer tier for model fine-tuning now available with no hosting fees.

 

GitHub and Agentic DevOps

GitHub Copilot Evolves into a True AI Partner

GitHub Copilot becomes a full-fledged agent, participating in DevOps alongside developers.

  • Coding Agent handles refactoring, test coverage, and feature implementation.
  • SRE Agent automates production issue mitigation and root cause analysis.
  • Agent Mode expands Copilot’s role within IDEs.
  • GitHub Models provides MaaS (Models as a Service) embedded in GitHub workflows.

 

Windows and Edge

Local AI Development and MCP Integration

Windows is fully embracing AI with new tools for local model development and agent interaction.

  • Windows AI Foundry supports full AI lifecycles with cross-silicon compatibility.
  • Semantic Search APIs, LoRA, and AI Toolkit simplify building performant AI apps.
  • Model Context Protocol on Windows enables app discovery by AI agents.
  • App Actions increases app engagement via AI-triggered functionality.

AI Features in Edge Browser

Edge delivers smarter browsing with embedded AI capabilities.

  • On-device AI APIs support summarisation, writing assistance, and translation.
  • Full PDF translations in one click.
  • Sales Assistant Agents and meeting summarisation in Edge for Business.

 

Security and Governance

Enterprise-Grade Controls for AI Apps and Agents

Security is central to Microsoft’s AI rollout, with new tools for visibility, risk management, and compliance.

  • Microsoft Purview SDK integrates compliance into AI apps.
  • DSPM for AI provides risk posture, auditing, and insider risk management.
  • Defender for Cloud delivers runtime protection and threat detection.
  • Data Loss Prevention (DLP) for Copilot agents prevents oversharing.
  • Entra Agent ID tracks and manages agent identities and access.

 

Frequently Asked Questions (FAQ)

What is Copilot Tuning?

Copilot Tuning is a feature allowing organisations to train AI models with proprietary knowledge, producing customised agents for domain-specific tasks.

What’s the significance of Agent2Agent protocol?

A2A enables decentralised, secure agent collaboration without intermediaries—crucial for handling enterprise workflows at scale.

Can I build agents without deep coding skills?

Yes. Microsoft Copilot Studio supports low-code creation while offering pro-code extensions for developers.

What’s new in GitHub Copilot?

GitHub Copilot now includes agent mode, turning it into an autonomous coding partner. It also integrates deeply with DevOps workflows and model selection tools.

How does Azure AI Foundry support local development?

Foundry Local brings AI models and inferencing directly to Windows and macOS, reducing latency and increasing privacy.

How is Microsoft ensuring AI safety?

Through features like Prompt Shields, PII detection, task adherence, and Microsoft Purview integrations, Microsoft ensures safety, security, and compliance.

What is the Model Context Protocol?

MCP is a standard for agent-to-agent communication, enabling agents to access data and functionality securely across ecosystems like Windows, Azure, GitHub, and Copilot Studio.

Find out more from Microsoft here.