Microsoft Tests Copilot PC Insights for Device-State Questions
Microsoft is gradually testing PC Insights, an opt-in feature in the Copilot app for Windows that answers questions using current device and file information after the user grants permission. The official support page says it can explain hardware, storage, performance, battery, connected-device, and selected file details, but cannot change settings, run fixes, or monitor the computer in the background. Digital Trends and PCWorld independently reported the rollout and its limitations. LDS treats the feature as a read-only diagnostic interface, not an autonomous repair agent, and recommends testing permission persistence, data minimization, answer accuracy, stale-state handling, and whether the explanation clearly identifies the system evidence behind each conclusion.
What happened
Microsoft is gradually testing PC Insights, an opt-in capability in the Copilot app for Windows. After a user asks a device question and grants permission, the feature can retrieve relevant system or file information and explain it conversationally. Microsoft lists examples involving hardware, storage, processor usage, battery health, network adapters, connected devices, and selected folders.
The official support page also draws a clear boundary: PC Insights provides information but does not change settings, run troubleshooting steps automatically, or monitor the device in the background. Digital Trends and PCWorld independently reported the rollout and noted that the feature remains experimental and may return incomplete or inaccurate answers.
Technical context
A conversational interface can reduce the effort required to navigate Settings and Task Manager, but diagnostic quality depends on the evidence supplied to the model. A plausible explanation is not useful if it is based on a stale snapshot, omits a relevant device signal, or confuses correlation with the cause of a slowdown.
| Evaluation layer | Useful test | Failure to watch |
|---|---|---|
| Permission scope | Ask for one device fact | Access broader than the question requires |
| State freshness | Change load or storage before asking again | Answer based on stale measurements |
| Evidence trace | Compare the response with system tools | Explanation without a verifiable signal |
| Uncertainty | Ask about an ambiguous performance issue | Confident diagnosis from incomplete data |
| Revocation | Remove persistent permission | Continued access after the setting changes |
For practitioners
Enterprise teams should test the feature under managed profiles before treating it as a support shortcut. Important controls include whether permission is per request or persistent, how conversation activity is handled, what information crosses account boundaries, and whether administrators can enforce consistent settings.
A useful benchmark should pair common support questions with known device states. Reviewers can measure factual accuracy, evidence coverage, abstention behavior, correction rate, and time saved relative to standard tools. The benchmark should include deliberately ambiguous problems where the correct answer is to request more evidence rather than name a cause.
Editorial analysis
LDS sees PC Insights as a narrow but practical form of tool-grounded assistance. Its value comes from translating structured device signals, not from free-form model intuition. Keeping the feature read-only is an important safety boundary; the next risk increase would occur if future versions can execute repairs or change settings.
What to watch
Watch regional availability, managed-device controls, permission defaults, evidence-level citations, diagnostic accuracy, conversation-data settings, and whether Microsoft adds action-taking capabilities with explicit confirmation and rollback.
Key Points
- 1Microsoft is gradually testing an opt-in Copilot feature that answers Windows device questions after the user grants permission.
- 2PC Insights can read relevant device information but cannot change settings, perform repairs, or monitor the computer in the background.
- 3LDS recommends permission, freshness, evidence, uncertainty, and revocation tests before organizations treat it as a support shortcut.
Scoring Rationale
An impact score of 5.0 reflects a useful experimental product capability with clear boundaries, tempered by gradual rollout and unproven diagnostic accuracy.
Sources
Primary source and supporting public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
