Copilot PC Insights: AI-powered hardware status queries land in Windows with caveats for performance

Microsoft is testing a new Copilot feature called PC Insights that lets users ask natural language questions about their PC's hardware and storage. For AI builders, this signals a shift toward on-device system introspection via AI assistants, but early reports raise concerns about resource consumption.

What happened

According to a Microsoft support document spotted by Windows Latest and reported by Digital Trends, PC Insights can query CPU, RAM, and GPU usage, calculate free storage space, and check folder sizes in locations like Downloads or Documents. It can also see connected USB devices, external drives, printers, webcams, and the state of Bluetooth and Wi-Fi connections.

The feature is opt-in and asks for permission each time a user asks a hardware-related question, unless the user switches to "Always allow." Copilot cannot open individual files, and for now it is limited to flagging problems rather than fixing them.

Why AI builders should care

PC Insights represents a pattern where AI assistants move from answering general knowledge questions to performing system-level introspection. For product teams building AI-assisted diagnostics, this opens up possibilities for natural language interfaces to hardware monitoring, but it also introduces design challenges.

The biggest tension is between utility and resource usage. Early testing found that the Copilot app uses up to 1GB of RAM at idle and ships with its own private copy of Microsoft Edge. If you are building an on-device AI feature that queries system state, you need to account for the runtime overhead of the AI assistant itself. A tool that promises to diagnose slowdowns should not be a major contributor to them.

The opt-in permission model is another consideration. Each query requires explicit consent unless the user opts for always-allow. This is good for privacy but may reduce spontaneous usage. Builders should think about how to make permission flows frictionless while maintaining trust.

Practical implications

If you are designing an AI-powered diagnostics tool, consider these points:

• Make permissions explicit and reversible. The PC Insights model of per-query consent with an always-allow option is a reasonable baseline.
• Clearly separate diagnosis from remediation. Users will expect the assistant to not only flag issues but also suggest or apply fixes. Managing expectations is critical.
• Measure the resource footprint of your AI runtime. Bundling a full browser engine or large language model can negate the performance benefits of the diagnostics feature.
• Consider whether the feature should run locally or in the cloud. Local execution may reduce latency but increase RAM and storage usage.

Caveats

The PC Insights feature is still in testing. No official rollout date or Windows version availability has been confirmed. The current Copilot app's resource usage (up to 1GB RAM at idle, bundled Edge copy) raises questions about whether the diagnostic tool will be practical on lower-end hardware. Additionally, the feature is diagnostic only: it can flag problems but cannot fix them, which may limit its usefulness for less technical users.

Builders should watch for official Microsoft announcements and benchmark the feature's actual performance impact before relying on it in their workflows.