Microsoft will ship three new Copilot capabilities to Copilot+ PC users within months, granting the AI assistant direct control over local files and the option to delegate workloads to on-device models. Microsoft outlined the changes on October 7, 2026 at its Hybrid Intelligence event.

The rollout extends Copilot from a cloud assistant into an on-device agent that can move files, change Windows settings and draw context from recent user activity. Microsoft Executive Vice President Jacob Andreou framed the access as permission-bounded.

"It has the same level of the ability to control and change things that I do as a user, but always with permission," Andreou said.

What changes for Windows users?

Three capabilities will land "in the coming months," Microsoft said:

• Local file and activity access, giving Copilot additional context for its outputs.
• On-device action execution, such as moving files and modifying system settings.
• Hybrid inference that routes simpler tasks to local AI models and reserves cloud calls for the hardest workloads.

"Copilot will still use the cloud for the hardest tasks," Andreou said, "but for times when cost or privacy matter more, it can delegate down to local models that run directly on your computer."

How does the hybrid model address enterprise cost concerns?

Forrester principal analyst Biswajeet Mahapatra said the design reflects buyer demand for routing flexibility rather than a full on-premises stance.

"The demand we see is not for AI that runs exclusively on-device, but for AI that can intelligently decide which workloads should run locally and which belong in the cloud," Mahapatra said. "Enterprises increasingly want both options available."

Microsoft also showed Autopilot, a long-running Copilot agent formerly known as Scout, running against locally hosted models. Autopilot remains in private preview.

Which hardware targets these workloads?

Microsoft unveiled two reference devices aimed at on-device AI: the Surface Laptop Ultra and the Surface RTX Spark Dev Box. The company did not disclose pricing or ship dates.

Tom Mainelli, group vice president of Device & Consumer Research at IDC, tied enterprise uptake to hardware refresh cycles.

"Most companies are still getting their heads around AI, token usage, which models to use, and where to run them," Mainelli said. "As these companies get more savvy and look to scale AI to more employees, the need to run AI locally — including Microsoft Copilot — will increase."

"As more systems ship with the right combination of CPU, GPU, NPU, and memory, and as local models continue to improve, I expect more AI jobs to run on the device," he added.

What is Microsoft Execution Containers and why does it matter?

Microsoft Execution Containers (MXC) reached general availability on Windows 11 alongside the Copilot announcements. MXC lets IT teams define which files and networks an agent may touch, an answer to the governance gap that comes with autonomous tools.

Supported agents at launch include OpenAI Codex and OpenClaw. Anthropic Claude Code, Box and Manus are slated to follow. Mainelli called the controls a differentiator.

"That's what Microsoft is bringing to the table: a level of security and traceability in these processes that should give companies confidence in using them," Mainelli said.

How should enterprises govern the rollout?

Mahapatra advised IT leaders to apply automation-platform discipline to Copilot agents rather than trusting the AI itself.

"Microsoft's announcement is significant because it acknowledges that agents require a different security model than traditional applications," he said. "Technologies such as Microsoft Execution Containers are designed to limit what agents can access, identify which agent took an action, and enforce policy-based controls."

Mahapatra cautioned against granting broad autonomous file access immediately. He named role-based access, read-only assistance, document summarization and search as the safest early use cases. He flagged autonomous file modification, regulated processes and systems of record as areas that demand stronger approval, monitoring and human oversight.

Forrester's research pointed to a slow enterprise PC refresh as a brake on adoption. "Most enterprise fleets will take several years to refresh," Mahapatra said. "The immediate opportunity is therefore concentrated among organizations already investing in AI PCs and high-performance devices."