2026 · learning project
AIS Audit Sandbox
Fraud detection practice, internal controls, JavaScript
A practice tool for spotting purchasing problems. Each round makes up a company with its own procurement policy, staff and approval limits, approved vendor list, and a year of 18 to 22 payments. Four to eight of them break the rules: duplicate payments, purchases split to stay under an approval limit, unregistered vendors, invoices that don’t match the PO or the receiving report, and approvals from someone without the authority. You check each payment’s documents against the company files, flag what you find, and get scored on both what you miss and what you flag wrongly.
I built it with a lot of help from AI, as a way to practice spotting these schemes in a ledger. The first version double-counted duplicate payments in the at-risk total, and later the answer key turned out to mark things the evidence didn’t prove. The case study covers both fixes.