New capabilities in Azure AI Foundry to build advanced agentic applications 

Every era has its inflection points—moments when niche technologies suddenly become foundational. Today, AI stands on the brink of a similar leap, evolving from assistants awaiting instruction to agents capable of serving as workplace teammates. At Microsoft’s 50th anniversary, we’re not simply celebrating our history—we’re laying the groundwork for tomorrow’s digital economy with Azure AI Foundry

The AI landscape has become much more complex in the past two years. As models continue to rapidly advance, organizations are balancing how to effectively integrate proprietary data, domain-specific actions, and secure environments with frontier models at the edge in order to get the most out of their AI investments. Fine-tuning and distillation are now crucial for customizing models and unlocking enterprise value. 

A foundation for AI advancement 

That’s why we introduced Azure AI Foundry last fall—a unified platform for models, tools, and workflows to help developers and enterprises design, customize, and manage AI apps and agents. Think of it as an agent factory—a modular production system designed for modern intelligence. Just as factories standardized manufacturing, Azure AI Foundry standardizes how we build intelligent systems through pre-built components, standardized processes, and continuous improvement via operational feedback. With access to more than 1,800 models and sophisticated tooling, more than 60,000 customers are already building their futures with Azure AI Foundry. 

More recently, AI applications have evolved to include agentic capabilities with tens of thousands of organizations using our Azure AI Agent Service since we announced its public preview in late 2024. One of my favorite examples is Atom, an AI agent built by Atomicwork, a leading agentic service management platform. Atom helps IT and HR teams enhance productivity and their early adopters have reported significant productivity gains, reduced operational costs, and increased employee satisfaction. 

Today, I’m excited to share new capabilities that will further help customers build and scale advanced AI agents in a trustworthy way

A new era of multi-agent systems 

First, we’re announcing the general availability of our agent framework—an extension of Azure AI Foundry’s open-source kit Semantic Kernel, specifically designed to simplify the orchestration of multi-agent systems. 

Available today, the Semantic Kernel agent framework makes it easier for agents to coordinate and dramatically reduces the code developers need to write. Organizations like KPMG are using Semantic Kernel to orchestrate workflows among specialized agents, dramatically reducing development complexity. 

At the core of Azure AI Foundry is a feedback system, designed to enable a continuous improvement cycle through real-time telemetry and user feedback. This approach is complemented by our advanced observability suite—evaluations, tracing, and A/B testing—giving developers unprecedented visibility into agent behaviors and outcomes. These tools aren’t just about monitoring; they’re about creating compounding improvements with every generation of your AI system. 

Consider Fujitsu: after using Azure AI Agent Service and Semantic Kernel to develop an intelligent, scalable AI agent for sales automation, sales proposal creation became 67% more productive, freeing employees up to focus on customer engagement and strategic planning. 

We are using Microsoft’s Semantic Kernel and Azure AI Agent Service to orchestrate multiple specialized AI agents and an orchestrator AI to coordinate them to answer questions as a team.

Hirotaka Ito, Lead Engineer of AI, Digital System Platform Unit, Fujitsu 

Scaling trust with AI Red Teaming 

As agent capabilities become smarter, managing risk is more essential than ever before. Yet, MIT Technology Review highlights that more than half of businesses still manually evaluate AI—an unsustainable approach at scale. 

To address this, we’re introducing the AI Red Teaming Agent—now in public preview. This agent systematically probes AI models to uncover safety risks, integrating Azure AI Foundry’s robust evaluation systems with Microsoft Security’s PyRIT (Python Risk Identification Tool) framework. The agent generates comprehensive reports, tracking improvements over time, creating an AI safety testing ecosystem that evolves alongside your system—a critical asset as agent capabilities expand and potential risks multiply.

At Accenture, we’re creating more agentic applications for our clients than ever before. Azure AI Foundry gives us a one-stop shop for all the right tools and services. To meet the growing demand while ensuring our applications are safe and secure, we’re looking into the AI Red Teaming Agent to automatically scan and ensure responsible development. 

Nayan Paul, Managing Director, Accenture 

Additionally, our new agentic evaluations—also available today—provide comprehensive risk and quality assessments, empowering teams to build trustworthy, compliant, and advanced AI agents from day one. These evaluation metrics help teams systematically identify and mitigate potential risks in agent behavior, providing another layer of security and compliance. 

Meeting developers where they work 

The vision of creating technology that feels so natural and fitting that it becomes second nature guides how we build for developers. Tools should feel native, not novel. Over the past year, we’ve strengthened our collaboration with GitHub, starting with the launch of GitHub Models to its 150 million-strong developer community, and more recently through a series of open integrations with Azure AI Foundry and widely used open-source frameworks.

Today, we’re extending this philosophy further with our new Azure AI Foundry extension for Visual Studio Code, now available in public preview. Developers can now build, test, and deploy agent-based applications entirely within their Integrated Development Environment (IDE)—no context switching required.

Additionally, we’re making Visual Studio Code even more powerful by making agent mode in GitHub Copilot available to all users. Agent mode, MCP support, and code review takes GitHub Copilot beyond code suggestions to become your peer programmer, performing complex, multi-step changes, running tests, and guiding you through its reasoning.

Building the future together 

Whether streamlining business processes, transforming customer interactions, or pioneering multi-agent applications, Azure AI Foundry provides the powerful infrastructure developers need to create intelligent agents at scale. 

As Microsoft marks its 50th anniversary, we’re looking forward to the exciting possibilities ahead. Join us at Microsoft Build from May 19-22, 2025, in Seattle and online to learn more about what’s next for Azure AI Foundry. Build the next 50 years together. 
 

The post New capabilities in Azure AI Foundry to build advanced agentic applications  appeared first on Microsoft Azure Blog.

This site uses cookies to offer you a better browsing experience. By browsing this website, you agree to our use of cookies. Click More Info to view Privacy Policy.