Audit Analytics in Python: Duplicate Payments, Benford's Law and Journal-Entry Testing

A supplier rang Kaveri Crunch Foods, a Pune snack maker, to say it had been paid twice. The audit committee's response: test every payment and every manual journal entry of the year, not a sample of 25. Ira Deshmukh, the internal audit lead, has the ERP exports and a Python script to write.Across five lessons you build that script on your own copy of the data. You prove the data first: voided cheques, rupee amounts in Indian digit grouping, day-first dates, hand-typed invoice numbers and a vendor master that multiplies rows. Then you run the three tests auditors rely on. Duplicate payments: exact and near matches on the supplier's GSTIN, which see through duplicate vendor records and ignore monthly rent. Benford's law: first-digit shares, z-scores, chi-square and Nigrini's MAD, followed down to payments just under the approval limit. Journal-entry testing: manual entries posted on weekends and holidays, late at night, in round sums, self-approved or unbalanced.In every lesson you build the test yourself first, then give the same job to your AI (ChatGPT, Claude, Copilot or Gemini) and learn exactly where it goes wrong: errors="coerce" hiding amounts, keep=False doubling the value at risk, a weekend test that misses Saturdays, a case-sensitive self-approval check. In the capstone your AI writes the whole audit script from one brief, and you prove every exception before it reaches the audit committee.Bartman runs your script on its own servers against your data and checks the files it writes, so every check is about your real output, not a screenshot.

Tools: Python, pandas, SciPy, Audit Analytics, Duplicate Payment Testing, Benford's Law, Journal-Entry Testing, matplotlib, AI Assistants for Python