AI-ASSISTED BUILDAUDIT LOG

2025 · government process automation

a government paper trail, compressed
to minutes.

A desktop app that tracks exams at a government university and generates the payment vouchers staff used to spend weeks preparing by hand. Each voucher is validated and prints with one click.

built in 5 days with ai coding tools

the problem

Paper vouchers are slow, and when one goes wrong nobody can trace where. In government institutions, payment workflows run on pre-printed forms, hand calculations and whatever staff remember about the process. There's no audit trail and no rollback, so nobody can check what went wrong or when. Each handover adds delay, and the delays stack up over months.

paper vouchers, passed desk to desk
the product decision

The system tracks each exam, and the vouchers are generated from that record. It follows each exam through scheduling, the sitting, marker handovers and mark entry. Staff enter data as each step happens, and the system applies the voucher rules so they don't have to.

exam lifecycle tracking
the result

One button produces every voucher, filled in, checked and ready to sign. Vouchers match the government-approved templates exactly. The system calculates and fills in each one, backs it up, and logs every step that produced it. What took weeks by hand now prints in about three minutes.

one-click voucher generation
5days to ship
~3minfrom click to print
80+papers tracked last semester
400+vouchers generated
5+departments

decision log

  1. 01Model the exam lifecycleStaff already work exam by exam: scheduling, the sitting, marker handovers, mark entry. The vouchers are generated from that record.
  2. 02Reproduce the approved templates exactlyPrinted vouchers have to match the government-approved forms.
  3. 03Log every step and back up every voucherOn paper there was no audit trail, so nobody could check what went wrong or when.
next case studyThe client wasn't happy, and no metric showed itread it →
working on something similar?I’m happy to talk it through.