AI Cash Leakage Expense Auditor

Stop losing money to unmanaged spend and billing errors. This skill acts as a forensic accountant to analyze expense reports, card statements, and ledgers to find cash leakage. It identifies duplicate payments, subscription sprawl, price increases, and missed discounts.

What it does

• Detects duplicates by identifying reused invoice references and identical transactions.
• Audits subscriptions to find redundant SaaS tools, unused seats, and silent price drift.
• Quantifies leakage by calculating recoverable cash versus recurring annual waste.
• Ranks findings using a priority score based on impact, confidence, and recovery effort.
• Generates recovery drafts using templates for vendor disputes and credit requests.

How it works

1. Intake profiles the data to map dates, vendors, and amounts while flagging quality issues.
2. Scan runs deterministic tests for outliers, round amounts, and weekend spending.
3. Analyze applies a leakage taxonomy to find tax errors, missed early-pay discounts, and contract mismatches.
4. Quantify computes the at-risk amount and assigns confidence levels to every finding.

Frameworks & tools

Uses a specialized leakage_scan.py tool for first-pass detection. Leverages a proprietary leakage taxonomy and detection playbook for deep forensic analysis.

Why this beats prompting it yourself

Generic prompts miss subtle billing patterns like price drift or split purchases designed to bypass approval limits. This skill enforces a forensic accounting framework that separates recoverable cash from control weaknesses to ensure you don't double-count savings.

Use cases

• Auditing monthly corporate card statements for subscription sprawl.
• Reviewing accounts payable ledgers before a fiscal year-end close.
• Identifying overcharges and duplicate payments in high-volume vendor invoice lists.
• Flagging expense policy violations and missing documentation for tax compliance.

Known limitations

Requires at least date, vendor, and amount columns to be useful. Analysis quality drops significantly if invoice numbers or reference codes are missing from the input data.