PlanningHigh riskComparison recommended

AI Expense Reimbursement Fraud Pattern Analysis Playbook

A regional bank's ethics hotline received a complaint that a senior VP in commercial lending is submitting duplicate and inflated expense reports. The VP manages a $420M loan portfolio and travels frequently. You have 28 months of expense submissions totaling $194,000.

When to use this playbook

  • Use this playbook when the decision looks like the situation above: A regional bank's ethics hotline received a complaint that a senior VP in commercial lending is submitting duplicate and inflated expense reports.
  • It is a fit when you have source files in hand and need a structured, reviewable analysis — not a generic chat answer about "Expense Reimbursement Fraud Pattern Analysis".
  • Do not use it as a substitute for licensed, legal, clinical, or authorized official judgment in the domain.

What you'll need

  • 28 months of expense reports (all submissions, all categories)
  • Receipts archive (scanned PDF)
  • Corporate card statement for the same period
  • Approved vendor/restaurant list for the bank's entertainment policy

Attachments: Multiple attachments (PDFs, Documents)

The Prompt

You are a forensic accountant investigating expense reimbursement fraud by a senior VP at a regional bank. I am attaching:

Work only from the attached source files. If a conclusion is not supported, say so.

Produce:
1. Identify all duplicate submissions: same amount, same date, same vendor across different expense reports or report periods.
2. Flag receipt amounts that differ from submitted amounts by more than $10, and identify the pattern (systematic rounding up vs. isolated incidents).
3. Identify entertainment expenses submitted on weekends or holidays and cross-reference against the VP's calendar for those dates.
4. Calculate the total estimated overpayment if the pattern holds across all 28 months.
5. Tell me whether the evidence supports termination for cause, criminal referral, or both—and what additional documentation I need to close each path.

Call out where independent models are likely to disagree, and list follow-up documents a reviewer should request.

What to expect

  • Duplicate transaction register with dollar totals
  • Receipt variance analysis
  • Timeline of suspicious submissions
  • Employer action decision framework with evidentiary thresholds

Review before you act

  • Validate this output against source files before relying on it: Identify all duplicate submissions: same amount, same date, same vendor across different expense reports or report periods.
  • Validate this output against source files before relying on it: Flag receipt amounts that differ from submitted amounts by more than $10, and identify the pattern (systematic rounding up vs. isolated incidents).
  • Validate this output against source files before relying on it: Identify entertainment expenses submitted on weekends or holidays and cross-reference against the VP's calendar for those dates.
  • Validate this output against source files before relying on it: Calculate the total estimated overpayment if the pattern holds across all 28 months.
  • Confirm every cited figure, date, counterparty, or requirement against the attached originals — models compress and can drop a qualifier.
  • Treat disagreement between models as a review item, especially on classification, materiality, and recommended next action.
  • Do not authorize an operational, clinical, legal, credit, or enforcement action solely because the models agree.

Why compare models on this

For Expense Reimbursement Fraud Pattern Analysis, running the same attachments across independent models is useful because the hard part is classification and completeness, not fluency. The workflow is already designed to surface duplicate transaction register with dollar totals; receipt variance analysis; timeline of suspicious submissions; employer action decision framework with evidentiary thresholds. Those are comparison artifacts — they only exist if more than one model runs. Models often split on qualitative materiality, intent versus error, and whether a newly formed counterparty is a red flag or a legitimate intermediary. Those splits are the review queue — not noise.

Forensic AccountingOccupational FraudPlanningHighMultiple attachments

See governed multi-model AI on your own prompt

Compare GPT-5, Claude, and Gemini side by side, with human review and a decision record built in.