Jeremy Neely Text Jeremy

AI & Agents · Production

Hatz User Watch

Warns admins before any person runs out of AI credits.

Built for: The AME Group, the managed IT provider I work for, watching AI usage across about 20 client companies

The tenant overview, then the users table filtered to one person and their credit detail (sample data).
The admin overview, shown with mock tenant data.
The admin overview, shown with mock tenant data.

The problem

Hatz AI credit limits are set per role and per person. Someone can burn through their allowance unnoticed until they are blocked in the middle of a task, and admins had no quick way to see who was over, about to go over, or had gone quiet.

What I built

I built a Windows tray app that watches every user in a client's AI tenant against their credit budget, with a forecast for each person and plain-language reasons. It can email a user a branded usage summary with recommendations and a button to request more credits, and it sends threshold alerts automatically.

How it works

  1. Watch every person in one tenant, not just the tenant total.
  2. Work out each person's effective limit and forecast the month.
  3. Color the tray icon and list who needs attention.
  4. Email users a branded summary when they cross a threshold.
  5. Keep a usage history for trends and AI-written insights.
How Hatz User Watch is wiredOne instance per tenant. Alert emails go once per billing period; MSP-only advice never reaches the end user.
Comes inInside Hatz User WatchGoes out Hatz admin API per-user usage, each poll Alert rules percent of each limit Effective limit role and personal caps Forecast pace, history, usual shape Verdict on track, likely over, over AI recommendations checked against real models Tray icon count needing attention Usage email branded, via Outlook History & PDFs MariaDB, per-user report

Built with

  • C# / .NET 8
  • WPF
  • WebView2
  • Chart.js
  • MariaDB
  • Hatz AI
Text Jeremy about this Next: AI Hours Tracker All 38 tools Texting opens your Messages app with a note already written. He replies himself.

More screens

The users table with credit use, limits and status per user (sample data).
The users table with credit use, limits and status per user (sample data).
One user's credit detail and outlook (sample data).
One user's credit detail and outlook (sample data).
Users by credit band and by status, with notable changes (sample data).
Users by credit band and by status, with notable changes (sample data).
The tenant overview, then the users table filtered to one person and their credit detail (sample data), a frame from the demo clip (sample data).
The tenant overview, then the users table filtered to one person and their credit detail (sample data), a frame from the demo clip (sample data).

More of Jeremy's work

All 38 tools, by category