Implement A Fraud Strategy Within Minutes
| Task | Manual Process | aiAutoPilot ML ® |
|---|---|---|
| Determine the performance of the current rules and models in the system and know where they are performing well or not | 10 minutes per rule | N/A |
| Determine the best data filters to apply to detect as much fraud as possible whilst reducing the non-fraud data | 0.5 – 1 day | N/A |
| Create machine learning rules and models using aiSmartScore and aiSmartRule from the filters | 1 day per run | N/A |
| Put the rules and models into production | 1 day per run | N/A |
Key Functionality
-
Automated machine learning model configuration and training triggered by arrival of new transaction and fraud data
-
Scoring against historical models for continual model monitoring.
-
Ingestion of active fraud strategy performance, both ML-based and manually authored rules
-
Identification of the best models and/or rules which maximise fraud detection
-
Identification of ‘delta’ changes required in active rule strategy to move to the optimal strategy
-
Secure API connectivity to aiRiskNet® ensuring changes can be implemented at the click of a button
-
Extensive audit trail of all models built, changes proposed, review procedures and changes implemented
Key Features
Automation
Can quickly and easily detect 80% of the fraud already being detected with a combined reduction in alerts
Optimisation
It can contribute to upwards of 20% reduction in fraud losses, compared with managing fraud manually
Cost Savings
80% of the fraud analyst team can be applied to other business areas due to the vast reduction in alerts.
Effective Detection
Produces highly effective rules, many suitable for immediate card blocking
Our Innovation Articles
