Detect Vendor/Employee Collusion

AI-Powered Procurement Fraud Detection Network Entity Resolution & Network Analysis AI builds a vendor network graph: connects entities sharing bank accounts, addresses, phone numbers, or beneficial owners. Identifies 'shell vendor' patterns: same bank account across 3 vendor IDs, registered addresses at UPS stores or virtual offices, and newly created vendors with no web presence. Behavioral Anomaly Detection AI establishes baseline spending patterns per vendor and buyer, then flags deviations: sudden volume increase with a specific vendor, invoices

consistently at 95-99% of approval threshold, round-number invoices ($5,000.00, $10,000.00), and weekend/holiday invoice submissions. Segregation of Duties Analysis AI maps the actual 'Who can do what' across the procure-to-pay cycle: vendor creation, PO approval, goods receipt, invoice approval, and payment release. Flags when the same person controls 2+ steps (SoD violation) or when a manager approves their own expenses. Kickback Detection AI looks for kickback indicators: employee-vendor relationship patterns (same address, shared phone), vendor

pricing above market rate by >15%, sole-source justifications that bypass competitive bidding, and employee lifestyle anomalies (expense patterns inconsistent with salary level). Investigation Case Management When AI flags a potential fraud pattern, it generates an investigation file: all related transactions, vendor profile, employee profile, network diagram, and recommended interview questions. Maintains chain of custody for evidence in case of prosecution. Run 'Ghost Vendor' scans monthly — AI compares the vendor master to state business registries.

Vendors with revoked business licenses, expired registrations, or matching employee home addresses are immediate red flags. Implement 'Cumulative Limit' tracking — a buyer who splits a $25K purchase into five $4,900 invoices (each below the $5K threshold) should trigger a cumulative threshold alert. Analyze the 'Time Between PO and Invoice' — legitimate vendors take days to weeks. Fraudulent invoices often appear within hours of PO creation because the same person creates both. Don't rely on deterrence alone — fraud controls must be detective (find it

after it happens) and preventive (stop it before it happens). AI provides both. Don't investigate based on AI flags alone without corroborating evidence — false positives erode trust. AI should provide a confidence score and recommend verification steps. Don't ignore 'small' fraud — the median procurement fraud scheme runs for 18 months before detection. Small amounts that compound over time are often larger than one-time big hits. The 'Fraud Triangle' Assessment Every fraud requires three elements: Pressure (financial need), Opportunity (weak controls),

and Rationalization (self-justification). AI can assess organizational 'Fraud Risk' by monitoring: Pressure indicators (salary freezes, layoff rumors, excessive overtime), Opportunity indicators (SoD violations, override frequency, manual payment percentage), and Rationalization indicators (declining employee satisfaction scores, ethics hotline call volume). High scores on all three dimensions indicate elevated fraud risk requiring enhanced monitoring.