PDQ Assistant Use Cases

Use PDQ Assistant to investigate device issues, build packages, and troubleshoot deployments with plain-language requests. Adapt the examples below to your environment.

Important: The Assistant requires your approval before running a command, changing a package, or starting a deployment. Review each proposed action before approving it.

Open PDQ Assistant

  1. Open a device, deployment, or package in PDQ Connect.
  2. Click Assistant in the upper-right corner.
  3. Enter a question or describe the task you want help with.

The Assistant uses the current page as context. You can also use @ mentions to reference specific devices, packages, or groups. Replace the example names and bracketed details in these prompts with your own.

Investigate a slow computer

When someone reports a slow computer, start by asking the Assistant to review available information and suggest what to check next.

Try this:

Review this device for possible causes of slow performance. Start with available inventory, then suggest diagnostic commands for anything you need to check live.

Review the findings and any proposed commands. Ask the Assistant to focus on a particular symptom, such as slow startup or an unresponsive application.

Follow up with:

The user reports that startup is slow. Help me investigate applications configured to start automatically.

Investigate startup applications with Autoruns

For a more detailed investigation on a Windows device, deploy the SysinternalsSuite package from the Package Library. After the deployment completes, ask:

Look in C:\PDQ\SysinternalsSuite for Autoruns. Use it to identify applications configured to start automatically on this device, then summarize entries that may warrant further investigation.

Review and approve the proposed command, then review the returned output and summary. An unfamiliar startup entry is a starting point for investigation, not proof of a problem.

Investigate an application crash

When someone reports an application crashing on a Windows device, use the Assistant to look for related errors around the time of the issue.

Try this:

[Application] crashed on this device around [time]. Propose a command to retrieve up to 20 relevant Windows Application event log entries from the surrounding hour. Summarize the errors and any recurring patterns.

Review and approve the proposed command, then compare the findings with the reported symptoms.

Follow up with:

Explain what these errors suggest and recommend the next diagnostic step. Identify any additional information needed to narrow down the cause.

Review and fix a failed deployment

Open a failed deployment and ask the Assistant to identify the relevant errors.

Try this:

Review this deployment. Identify the failed devices and steps, summarize the relevant errors, and recommend the next action. Group devices with the same failure.

Review the findings and suggested fix. If a package change is needed, ask the Assistant to propose it and review the change before approving it.

Follow up with:

Deploy the updated package to @Test-Device so I can check the result before retrying the other failed devices.

After the deployment completes, ask the Assistant to review the result. Continue with the remaining devices when you're ready.

Tip: Include the expected result and any differences you've noticed between successful and failed devices.

Compare a working device with an affected device

When an application works on one device but fails on another, comparing installed software can help identify differences worth investigating.

Try this:

Compare the installed software on @Working-Device and @Affected-Device. Show applications that appear on only one device and version differences for software installed on both.

The comparison uses available inventory data. Check the scan timestamps if recent changes aren't reflected.

Follow up with:

Focus on [application] and its related components. Explain which differences may be relevant to the problem and what to check next.

Software differences can help guide troubleshooting, but don't necessarily explain the failure.

Build a silent installation package

Use the Assistant to research installation options and help create a Windows package.

Try this:

Research unattended installation options for [application and version]. Identify the installer information you need from me, then help create a Windows package.

Provide the requested installer details. Review the proposed package steps, installation arguments, and restart behavior before approving package creation.

Follow up with:

Explain each step in this package and identify anything that could interrupt someone using the device.

When the package is ready, ask the Assistant to deploy it to a test device. After that deployment completes, ask it to review the result before deploying more broadly.

Note: Package creation and editing currently support Windows packages. Device and deployment assistance can include macOS devices.

Summarize recent deployment failures

Ask the Assistant to summarize recent failures when you need to identify recurring problems or prepare an update for your team.

Try this:

Review deployment failures from the past 24 hours. Group similar errors, list affected devices, and create an HTML report with recommended next steps.

Review the report and narrow the request if you need more detail about a particular package or error.

Follow up with:

For the most common failure, summarize the relevant device logs and suggest what I should investigate first.

More things to try

Task Example prompt
Troubleshoot printing on a Windows device Users report that printing stopped working on this device. Propose commands to check the Print Spooler service and related errors from the past 24 hours. Summarize the findings before suggesting changes.
Understand an existing package Explain what this package does, including its steps, installation arguments, and restart behavior.
Deploy software to a group Deploy @Package to @Group.
Prepare a troubleshooting handoff Summarize this conversation for a ticket: the reported problem, findings, commands run, changes made, and remaining questions.

Get better results

  • Include the device or package, the symptom or desired outcome, and any relevant time window.
  • Request a table, summary, or report when you have a preferred format.
  • Keep follow-up questions in the same conversation so the Assistant can use earlier findings. Separate conversations don't share memory.
  • For workflows with multiple actions, prompt the Assistant to continue after an action completes. For example, approve package creation, then ask it to deploy the package.

For information about how your data is handled, see the PDQ Connect AI Data and Safety FAQ.

Was this article helpful?