CASE STUDY

Crédit Agricole

Banking Giant Slashes Cost of Payment Fraud with Prevision.io.

With Prevision.io, our anti-check fraud processes are more reliable and accurate. We are saving significant time and money. We achieved ROI in under 3 months.

Francis Glomon,

Head of Customer Digital Data Innovation / Crédit Agricole, Provence Côte d’Azur

Improving the security of checking processes

Credit Agricole Group is the world’s largest cooperative financial institution. Crédit Agricole Provence Côte d’Azur is one of the 39 regional banks making up its retail banking network.

Check volumes have been declining steadily at a rate of around 11% year over year. Nevertheless, they remain an essential payment method for both professional private customers alike in France. 

Although they only account for 7% of payment volumes, checks suffer the highest level of fraud of all payment methods: 43% of payment fraud in total. This is pushing many in the industry to redouble their security efforts.

Fighting fraud more effectively with AI

Methods to counter check fraud have been refined over the years. However, in 2018, check fraud made an unexpected jump of 52%. This translates to an industry loss of 450 million euros nationally. 

Credit Agricole PCA has not been immune to the phenomenon. With check fraud costing it a constant 5 million euros year on year despite declining volumes, the bank sought to further optimize its due diligence and investigatory processes. 

Yet, there were significant barriers moving to the production phase. In particular, there were no automated means of monitoring the ongoing reliability of a machine learning model once in actual use. This made it impractical to put them into production.

A clear boost to operational performance

Prevision worked with the bank’s data science teams to understand how the users in the anti-fraud team were using the existing data feeds and scores. Using the Prevision enterprise AI platform, within the space of just a few hours they were able to automatically build, benchmark and validate the relative performance of 40 different models. Each of these outperformed the existing models the team had at its disposal. 

The result is a unified data app that draws on multiple data feeds and analyzes 10 000 to 15 000 checks from private customer accounts that pass through the system every day. Each check is sorted by probability of fraud. A specialist team then takes over and examines approximately 100 to 150 that are flagged for attention.

The team can instantly access the characteristics of the check, the details of the person who issued it, and the reasons for which the transaction was flagged as suspicious. They can then decide whether further investigation is required. Importantly, they can give feedback on the recommended actions generated by the app and thus ensure the underlying model remains reliable and free from bias.

A reliable AI platform for future projects

Prevision’s AI platform has already improved  check fraud avoidance above and beyond our existing processes. All while saving 3 to 4 hours’ work every day within the investigatory team,” said Francis Glomon, Head of Customer Digital Data Innovation at Crédit Agricole, Provence Côte d’Azur. 

Prevision was quickly adopted internally. Our data science teams are confident in the platform’s ability to help them save time and effort with no compromises in terms of performance, reliability and explainability in a whole variety of projects we have planned, such as risk analysis and marketing lifetime scores. Likewise, its ability to deploy to our private cloud is essential for security reasons and well appreciated by our IT teams. 

All told, our anti-fraud AI app paid for itself in less than 3 months.

ROI
Achieved in under 3 months

Hundreds of Thousands of Euros
of avoided loss each year.

5 hours saved
very day by in-house anti-fraud team

Accelerate Your Enterprise AI Projects Today

Get in touch to discuss your business challenges, to request a demo, or to open an account.

Get started