Microsoft’s Next Windows 11 AI Push Could Put Autonomous Agents on Your PC

Microsoft wants Windows 11 to become a platform for local AI, with more models and AI agents running directly on PCs instead of relying entirely on cloud services.

At IFA 2026, Windows and Devices CVP Mark Linton described Microsoft’s strategy using the phrase “unmetered intelligence,” as Windows Latest reports.

Rather than adding more visible Copilot buttons, Microsoft wants Windows to provide the underlying infrastructure needed to run AI models and autonomous agents locally.

Microsoft calls local AI “unmetered intelligence”

Microsoft uses “unmetered intelligence” to describe AI workloads that can run locally without sending every request to paid cloud infrastructure.

Running models directly on a PC could reduce cloud inference costs while also allowing users and developers to access AI capabilities without relying on an internet connection for every task.

Microsoft does not expect local models to replace cloud AI completely. Instead, the company sees both approaches working together depending on the workload.

Windows PCs could run much larger AI models

Microsoft also discussed Windows PCs capable of running models containing tens or even hundreds of billions of parameters locally.

More powerful hardware could make local AI practical for developers, creators, and users who regularly work with AI-powered applications.

This approach would move Windows beyond individual AI features and turn the operating system into a platform that applications can use to access local AI capabilities.

Microsoft wants Windows to become an operating system for AI agents

Microsoft says it wants Windows to become “the best platform for AI,” with greater emphasis on developer tools, local model execution, and infrastructure for autonomous agents.

The company is also working on support for long-lived agents that can perform tasks across applications and continue operating over extended periods.

Future Windows releases should give users more control over what those agents can access and which actions they can perform.

Users and admins will get more control over AI agents

Microsoft is developing auditing tools that could let users and administrators review agent activity and inspect logs showing what an AI agent has done.

Enterprise administrators should also gain policies and restrictions for controlling how agents operate across corporate systems.

Microsoft is additionally working on isolated containers and subsystems for AI agents. These environments could limit access to personal files, corporate data, and other sensitive resources.

Microsoft will reveal more later this year

Microsoft plans to share more details about its Windows AI and agent strategy later in 2026.

The broader direction suggests that AI will remain deeply integrated into Windows 11, but Microsoft increasingly wants models, agents, and AI infrastructure to operate at the platform level rather than appearing mainly through dedicated Copilot buttons.

In other AI news, OpenAI admits that its agents used a programming wiki to organize attacks, while GPT-6 Astra has been launched.