Microsoft is turning Windows 11 into an AI operating system

At a glance:

• Microsoft is integrating local AI models and agents directly into Windows 11 to enable natural language system control.
• The strategy focuses on "unmetered intelligence," allowing users to run AI workloads locally without token costs or data leaving the device.
• Developers can leverage the Windows ML framework and the Foundry portfolio to build offline, hardware-accelerated AI applications.

The shift toward local intelligence

For years, Microsoft has positioned Windows 11 as an AI-ready operating system, but recent developments at the Build event suggest the company is finally moving from hype to architectural reality. Rather than relying solely on cloud-based chatbots like ChatGPT or Gemini, Microsoft is embedding AI models directly into the OS and its core applications. This approach aims to transform the user experience from manual menu navigation to intent-based interaction through natural language.

A key pillar of this transformation is the concept of "unmetered intelligence." According to Anastasiya Tarnouskaya, product manager for Windows ML, running AI locally provides several critical advantages: there are no token costs for the user, sensitive data remains on the device for enhanced privacy, and latency is significantly reduced. This shift is supported by a massive installed base; Tarnouskaya noted that more than 500 million PCs are already running local AI workloads, proving that the hardware foundation is already largely in place.

Expanding the developer ecosystem

The proliferation of AI-capable applications on Windows 11 is being driven by the release of Windows ML last fall. This framework allows developers to create offline AI applications by mapping software to localized models and hardware components like GPUs and neural processors. To support this, Microsoft has introduced the "Foundry" portfolio, which includes:

• Foundry Local: Designed for running open-source models directly on Windows devices.
• Windows AI APIs: Tools that automate complex tasks such as conversation summarization, speech recognition, and video upscaling.

Major software players are already capitalizing on these tools. Microsoft’s own suite—including Office, Photos, and Teams—is utilizing on-device capabilities. For example, Outlook uses the Phi Silica model alongside a PC's GPU to summarize emails. Beyond Microsoft, companies such as Adobe, WhatsApp, Canva, Affinity, and Speechify are reportedly building local AI-powered experiences to take advantage of this new architecture.

The rise of agentic AI and personal assistants

Beyond simple task automation, Microsoft is moving toward "agentic AI," where the OS acts as a proactive assistant. Samantha Song, a product manager for Windows, demonstrated how users can use natural language to execute complex system changes. Instead of digging through settings menus, a user can simply describe how they want to personalize their wallpaper, colors, or menus, and an AI agent will execute the change as a single, coherent action.

To make this work, developers must create "skills files" that define how an agent should behave. These files allow specific skills to be reused across different contexts, enabling the agent to map natural language commands to real system outcomes. This level of automation is expected to change the fundamental way users interact with their computers, moving away from a command-and-control model toward a conversational one.

Hardware implications and enterprise strategy

The move toward an AI-centric OS is forcing a significant rethink of hardware procurement, particularly for large organizations. Because AI workloads are highly dependent on specific silicon, such as Neural Processing Units (NPUs), the hardware requirements for a modern workstation are shifting. Darren Oberst, co-founder of LLMWare.ai, noted that running models locally on an NPU allows for optimized performance and scheduled automated agent runs.

Industry analysts suggest that this transition will create a divide in the PC market. Jack Gold, principal analyst at J. Gold Associates, recommends that enterprises prioritize the purchase of "AI PCs" during their next upgrade cycle to ensure compatibility with these new features. Meanwhile, hardware manufacturers like Samsung and Lenovo are already beginning to roll out their own versions of "personal AI" features, though analysts like Leonard Lee warn that the industry must move carefully to ensure safe and secure deployment of these agentic capabilities.