Back to industries
Public Sector
Social benefits fraud detection
Client · National social-protection agency
A control-targeting system combining ML and graph analysis, multiplying the fraud detection rate by 3.5× at constant headcount.
01
Challenge
- Control targeting historically based on known business rules, easily circumvented by organized fraudsters.
- Very limited control capacity vs. >10M beneficiary base.
- Low recovery rate on classic controls (<15%), generating a sense of inefficiency.
- Strong requirements on explainability, algorithmic fairness and regulatory compliance.
02
Solution
- Risk-scoring model per file, combining behavioral, declarative and cross-administrative features.
- Fraudulent-network detection module via graph analysis (shared bank accounts, addresses, IPs) using GNN.
- Algorithmic audit framework: fairness tests, drift monitoring, full documentation.
- Investigator tool with graph visualization and automated reporting.
03
Business impact
- Targeted control detection rate: 15% → 52% (3.5×).
- +€85M additional recoveries identified in year one.
- −35% investigation time per file.
- Audit framework cited by the regulator as a sectoral best practice.
Other industries
Banking
Real-time transaction fraud detection
Insurance
Health underwriting automation with GenAI
Retail
Local assortment personalization & dynamic pricing
Energy
Wind production forecasting & market arbitrage
Manufacturing
Predictive maintenance on F&B production lines
Logistics
Route optimization & demand forecasting
Let's identify where AI can create value in your business.
Book a 30-minute introductory call with a founder.