At the moment, AI stands getting ready to an analogous leap, evolving from assistants awaiting instruction to brokers able to serving as office teammates. At Microsoft’s fiftieth anniversary, we’re not merely celebrating our historical past—we’re laying the groundwork for tomorrow’s digital economic system with Azure AI Foundry.
Each period has its inflection factors—moments when area of interest applied sciences all of a sudden turn out to be foundational. At the moment, AI stands getting ready to an analogous leap, evolving from assistants awaiting instruction to brokers able to serving as office teammates. At Microsoft’s fiftieth anniversary, we’re not merely celebrating our historical past—we’re laying the groundwork for tomorrow’s digital economic system with Azure AI Foundry.
The AI panorama has turn out to be rather more complicated up to now two years. As fashions proceed to quickly advance, organizations are balancing the right way to successfully combine proprietary information, domain-specific actions, and safe environments with frontier fashions on the edge in an effort to get essentially the most out of their AI investments. Fantastic-tuning and distillation are actually essential for customizing fashions and unlocking enterprise worth.
A basis for AI development
That’s why we launched Azure AI Foundry final fall—a unified platform for fashions, instruments, and workflows to assist builders and enterprises design, customise, and handle AI apps and brokers. Consider it as an agent manufacturing facility—a modular manufacturing system designed for contemporary intelligence. Simply as factories standardized manufacturing, Azure AI Foundry standardizes how we construct clever programs by pre-built parts, standardized processes, and steady enchancment by way of operational suggestions. With entry to greater than 1,800 fashions and complex tooling, greater than 60,000 prospects are already constructing their futures with Azure AI Foundry.
Extra not too long ago, AI functions have developed to incorporate agentic capabilities with tens of 1000’s of organizations utilizing our Azure AI Agent Service since we introduced its public preview in late 2024. Certainly one of my favourite examples is Atom, an AI agent constructed by Atomicwork, a number one agentic service administration platform. Atom helps IT and HR groups improve productiveness and their early adopters have reported important productiveness features, diminished operational prices, and elevated worker satisfaction.
At the moment, I’m excited to share new capabilities that may additional assist prospects construct and scale superior AI brokers in a reliable manner.
A brand new period of multi-agent programs
First, we’re saying the final availability of our agent framework—an extension of Azure AI Foundry’s open-source equipment Semantic Kernel, particularly designed to simplify the orchestration of multi-agent programs.
Out there as we speak, the Semantic Kernel agent framework makes it simpler for brokers to coordinate and dramatically reduces the code builders want to write down. Organizations like KPMG are utilizing Semantic Kernel to orchestrate workflows amongst specialised brokers, dramatically decreasing growth complexity.
On the core of Azure AI Foundry is a suggestions system, designed to allow a steady enchancment cycle by real-time telemetry and consumer suggestions. This method is complemented by our superior observability suite—evaluations, tracing, and A/B testing—giving builders unprecedented visibility into agent behaviors and outcomes. These instruments aren’t nearly monitoring; they’re about creating compounding enhancements with each era of your AI system.
Think about Fujitsu: after utilizing Azure AI Agent Service and Semantic Kernel to develop an clever, scalable AI agent for gross sales automation, gross sales proposal creation turned 67% extra productive, releasing staff as much as give attention to buyer engagement and strategic planning.
We’re utilizing Microsoft’s Semantic Kernel and Azure AI Agent Service to orchestrate a number of specialised AI brokers and an orchestrator AI to coordinate them to reply questions as a workforce.
Hirotaka Ito, Lead Engineer of AI, Digital System Platform Unit, Fujitsu
Scaling belief with AI Pink Teaming
As agent capabilities turn out to be smarter, managing threat is extra important than ever earlier than. But, MIT Know-how Overview highlights that greater than half of companies nonetheless manually consider AI—an unsustainable method at scale.
To deal with this, we’re introducing the AI Pink Teaming Agent—now in public preview. This agent systematically probes AI fashions to uncover security dangers, integrating Azure AI Foundry’s strong analysis programs with Microsoft Safety’s PyRIT (Python Threat Identification Device) framework. The agent generates complete studies, monitoring enhancements over time, creating an AI security testing ecosystem that evolves alongside your system—a important asset as agent capabilities develop and potential dangers multiply.
At Accenture, we’re creating extra agentic functions for our shoppers than ever earlier than. Azure AI Foundry provides us a one-stop store for all the suitable instruments and providers. To fulfill the rising demand whereas guaranteeing our functions are protected and safe, we’re trying into the AI Pink Teaming Agent to robotically scan and guarantee accountable growth.
Nayan Paul, Managing Director, Accenture
Moreover, our new agentic evaluations—additionally obtainable as we speak—present complete threat and high quality assessments, empowering groups to construct reliable, compliant, and superior AI brokers from day one. These analysis metrics assist groups systematically establish and mitigate potential dangers in agent habits, offering one other layer of safety and compliance.
Assembly builders the place they work
The imaginative and prescient of making know-how that feels so pure and becoming that it turns into second nature guides how we construct for builders. Instruments ought to really feel native, not novel. Over the previous 12 months, we’ve strengthened our collaboration with GitHub, beginning with the launch of GitHub Fashions to its 150 million-strong developer group, and extra not too long ago by a sequence of open integrations with Azure AI Foundry and broadly used open-source frameworks.
At the moment, we’re extending this philosophy additional with our new Azure AI Foundry extension for Visible Studio Code, now obtainable in public preview. Builders can now construct, check, and deploy agent-based functions totally inside their Built-in Growth Surroundings (IDE)—no context switching required.
Moreover, we’re making Visible Studio Code much more highly effective by making agent mode in GitHub Copilot obtainable to all customers. Agent mode, MCP assist, and code overview takes GitHub Copilot past code options to turn out to be your peer programmer, performing complicated, multi-step adjustments, working checks, and guiding you thru its reasoning.
Constructing the longer term collectively
Whether or not streamlining enterprise processes, reworking buyer interactions, or pioneering multi-agent functions, Azure AI Foundry offers the highly effective infrastructure builders must create clever brokers at scale.
As Microsoft marks its fiftieth anniversary, we’re trying ahead to the thrilling potentialities forward. Be a part of us at Microsoft Construct from Could 19-22, 2025, in Seattle and on-line to study extra about what’s subsequent for Azure AI Foundry. Construct the following 50 years collectively.