Jeremy Neely Text Jeremy

Security & Compliance · Beta

Shadow AI Scan and Reporter

Finds the AI tools in use across a client's PCs and reports the risk.

Built for: The AME Group's sales and admin teams, and client leadership

Dropping in the fictional sample scans and reading the report's overview and key observations (sample data).
The overview of a Shadow AI discovery report with an overall risk rating (sample data).
The overview of a Shadow AI discovery report with an overall risk rating (sample data).

The problem

Employees adopt AI tools on their own, and leadership cannot see where company data is going. The vendor's stock scanner missed whole categories and kept results only in the vendor's dashboard.

What I built

I built AME's edition of the Hatz endpoint collector, a PowerShell scan pushed through N-able N-central that also keeps AME's own copy of every scan. A companion tray app files scans by client, redacts them, and builds HTML and PDF discovery reports.

How it works

  1. An admin sets up one automation object in N-central.
  2. Each PC scans browsers, desktop apps, extensions and npm packages.
  3. Hatz gets its copy; AME's copy goes to OneDrive through Power Automate.
  4. The Reporter app files scans by client and applies redaction.
  5. A report with rated observations goes out on a schedule.
How Shadow AI Scan and Reporter is wiredThe MSP keeps its own copy of every scan instead of trusting the vendor dashboard. Reports are redacted first.
Comes inInside Shadow AI Scan and ReporterGoes out Managed PCs via N-central automation Browser & app traces extensions, npm packages PowerShell scan runs as LocalSystem Dual upload Hatz + Power Automate copy File by client Reporter tray app Redact URLs, ids, emails Discovery report HTML and PDF, risk-rated Vendor dashboard Hatz's own copy Own scan archive OneDrive, by client

Built with

  • PowerShell 5.1 / 7
  • Pester 5
  • N-able N-central
  • Power Automate
  • Python
  • SQLite
Text Jeremy about this Next: AME CMMC All 38 tools Texting opens your Messages app with a note already written. He replies himself.

More screens

Key observations rated by risk, each with a recommended action (sample data).
Key observations rated by risk, each with a recommended action (sample data).
Three locations compared side by side (sample data).
Three locations compared side by side (sample data).
Usage by AI category and by person (sample data).
Usage by AI category and by person (sample data).
The findings explorer with redacted conversation links (sample data).
The findings explorer with redacted conversation links (sample data).
The report builder, where scan files are dropped in.
The report builder, where scan files are dropped in.
Dropping in the fictional sample scans and reading the report's overview and key observations (sample data), a frame from the demo clip (sample data).
Dropping in the fictional sample scans and reading the report's overview and key observations (sample data), a frame from the demo clip (sample data).

More of Jeremy's work

All 38 tools, by category