Demonstrating ai’s fraud prevention expertise for FI’s
The ai Corporation (ai) has worked with a large regional Financial Institution (FI) to demonstrate how ai’s fraud prevention expertise and RiskNet products can reduce the FI’s fraudulent activity and number of false alerts.
ai’s RiskNet is a Pre-Authorisation and Post-Authorisation self-service rules engine for the detection and prevention of fraud and other suspicious transactions, supporting all card scheme recommended rule sets and can be deployed with or without neural models.
ai worked with the FI via a consultative approach throughout the life cycle of the RiskNet Pre-Authorisation implementation project.
The results of the implementation exercise showed that ai was capable of increasing the FI’s fraud detection while reducing the number of false alerts – over $1m of fraud detected in 4 months and reducing false alerts by 8,000.
Other ai recommended best practices around rules analysis and rule management have also been adopted by the FI to ensure rules are always targeting the latest fraud trends.
With ai’s RiskNet Post-Authorisation module already successfully deployed, the FI had a requirement to implement Real Time monitoring and intervention capability via RiskNet’s Real Time module.
The FI’s experience with fraud risk management was limited, so it required ai to take a significant role in the project from all aspects.
ai worked with the FI via a consultative approach throughout the life cycle of the RiskNet Pre-Authorisation implementation project.
- An ai Business Analyst and a Project Manager were assigned from the start to agree and document the full requirements of the system and to agree the project plan with the FI.
- A Technical SME and a Risk SME were assigned from ai to the project to discuss and document data mapping and Technical & business operational workflows
- Full training of the RiskNet system was provided on-site by an ai Product SME
- Technical Consultants carried out an offsite deployment of the RiskNet Pre-Authorisation system
- The FI performed their User Acceptance Testing, and Go Live was agreed.
- ai supported the FI with the implementation of a go-live rules analysis
- ai provided Hypercare support for an agreed period whilst the FI built familiarity with the new system and processes
- Hypercare ended when ai agreed with the FI that the system was running operationally in a “business-as-usual” state
ai undertook analysis, using the FI’s transactional data and confirmed fraud data, to evaluate performance of it’s existing fraud tool implementation. Of the 16 rules implemented by the FI, only 3 had a detection rate lower than 75:1, showing an urgent need for more efficient rules.
Following ai’s analysis, 12 high performing rules were implemented in the Near Real-Time capacity with the following results:
- Fraud detection of 76% by value of losses
- A false positive ratio of 15:1
In addition, 9 further rules were implemented in the Real-Time environment with the following results:
- Fraud detection of 17% by value of losses, where fraud is stopped at first attempt
- A false positive ratio of 5:1
The results of the implementation exercise showed that ai was capable of increasing the FI’s fraud detection while reducing the number of false alerts – over $1m of fraud detected in 4 months and reducing false alerts by 8,000.
The rule sets have been implemented for over 6 months and there has been great feedback from the FI regarding their effectiveness.
The performance of RiskNet has been fully endorsed to the rest of the FI, and opportunities to deploy RiskNet to other connected divisions are in discussion.
