As of 13 August 2026, AI can only partly check the security of employee devices.

What to actually do

1. Hand it to a person

   The route this page recommends

   A person who owns the outcome does this end to end, worth it when the failure is dear.

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, power-user skill, and roughly 1 hour until you can act on the result.

   How to actually do it

   1. Open the approved device-management, endpoint-protection and patch-management consoles, and export current device status without including passwords, private keys, personal files or unnecessary employee identifiers.
   2. Gather the organisation's approved security baseline, including supported operating systems, required endpoint protection, firewall requirements, disk-encryption requirements and account-security controls.
   3. Paste the sanitised exports and the approved baseline into a chatbot with the prompt above, keeping each report labelled by source and date.
   4. Ask the chatbot to separate confirmed findings from likely findings and unknowns, then save the resulting triage report with the source exports.
   5. Open the relevant vendor console and documentation for every flagged device or control, and compare each finding against the live status rather than relying on the AI report alone.
   6. Have an IT or security colleague confirm the evidence, decide which findings are genuine, approve remediation and record what was changed in the authorised management system.

   Prompt

   Act as a cautious cybersecurity analyst. I will provide sanitised exports or screenshots from approved company systems that show employee-device security status. Do not claim to have accessed any device, endpoint agent or live network. Do not ask for passwords, private keys, full employee names, personal files or other unnecessary personal data. Analyse only the information supplied.
   
   Produce:
   1. A table of devices or device groups with the evidence supplied, detected gaps and confidence level.
   2. Separate findings into confirmed, likely and unknown.
   3. Check for missing operating-system updates, unsupported systems, inactive endpoint protection, disabled firewalls, weak disk encryption status, stale device records, missing multi-factor authentication evidence and accounts or devices that appear unmanaged. Only report a condition when the supplied evidence supports it.
   4. For each finding, explain the security risk in plain English, the evidence behind it and the safest next investigation step.
   5. Give remediation actions that an IT administrator can carry out, but do not invent commands, settings, product names or policy requirements. If a product-specific check is needed, say which vendor documentation or console screen must be consulted.
   6. List limitations, missing evidence and any conclusion that cannot be made from this data.
   7. End with a short handover checklist for an IT or security colleague to verify before anyone changes a device.
   
   Treat the result as a triage report, not proof that the devices are secure. Keep employee identifiers to the minimum needed. Here is the approved evidence:
   [PASTE SANITISED DEVICE AND SECURITY REPORTS HERE]
   
   Company requirements or baseline, if available:
   [PASTE APPROVED SECURITY POLICY OR BASELINE HERE]

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.