when AI helps say yes or no to a borrower

    Home / Business Personal finance / when AI helps say yes or no to a borrower

when AI helps say yes or no to a borrower


Is AI the lethal weapon against payment defaults? BNP Paribas Personal Finance (Cetelem, Cofinoga, AlphaCredit, etc.) which supports “European leader in consumer credit”, has chosen the MondoBrain startup to be able to protect itself from it. Applied to the bank, this “augmented intelligence” software (which combines artificial, human and collective intelligence) would allow the establishment of personal loans to be taken “The best possible decision” in granting credit on the basis of available data, explained Bruno Saint-Cast, MondoBrain marketing director, during a conference in Paris on Wednesday 3 October.

“It helps us accelerate the optimization of our processes and protect our decision-making process,” said Khalid Saad-Zaghloul, head of digital and fintech risk at BNP Paribas Personal Finance.

Based on the same principle of credit “scoring”, this software implemented in April 2017 at the subsidiary BNP Paribas, uses data compilations to generate more reliable indicators than traditional tools. Analyzes and crosses a series of criteria (expenses per card, professional situation of the borrower, age, income, etc.) in order to assess a level of risk, but with “computing power multiplied by exclusive algorithms”presented by MondoBrain.

During a MondoBrain online seminar last July, Bruno Saint-Cast estimated that his software had allowed, only once, to 25% improvement in recovery in Italy and a 13-point increase in the share of assisted decisions at BNP Paribas Personal Finance. During the conference, he did not want to communicate a figure.

From predictive to prescriptive

Before AI, the consumer credit group used the score (standard analysis) in the 1970s to identify customers who are likely to repay their loans. But with 27 million customers to manage, checking borrower profiles needs to be even more sophisticated and efficient. The branch of BNP Paribas thinks MondoBrain allows it to go a step further.

During the conference, Augustin Huret, founder of MondoBrain, said that augmented intelligence makes this possible “Going from predictive to prescriptive . Predictive analysis helps predict a customer’s creditworthiness, while prescriptive analysis goes further, evaluating what the risks are and how they materialize based on criteria decided by the transaction manager, all in rconstituting in natural language the reasons for the decision-making process. In other words, it allows you to optimize the choice whether or not to accept a customer who wishes to borrow, having an overview and avoiding the “black box” effect. (when there is no one able to analyze with what reasoning one arrives at this or that result).

“Using MondoBrain is as easy as a video game. We have simply translated the sum of dozens and dozens of statistics into a color code “Augustin Huret explained.

“Green” represents a low risk, “yellow” a limited risk and “red” a high risk of failure. But Khalid Saad-Zaghloul pointed out that it is good “human intelligence” who makes the final decision and who will never be there ” auto reply “specifying that this process is in any case mandatory in Europe unlike the United States.

Reduced response time

MondoBrain also ensures that its technology enables BNP Paribas Personal Finance “to improve competitiveness and profitability, reduce costs and build customer loyalty”, according to Bruno Saint-Cast. Analyzing billions of combinations “in seconds”MondoBrain would significantly reduce the response times to credit applications.

“Augmented intelligence allows for a more optimal decision, better shared and faster implemented. The decision making is done 5 to 20 times faster than a conventional process. We spend a few hours on a file to minutes, “says Khalid Saad-Zaghloul.

Currently, the number of MondoBrain users of BNP Personal Finance is limited to about thirty people, “Really able to define problems, then find the type of data to use and find a solution. [Elles] are certified “according to Khalid Saad-Zaghloul, adding however that this software could be extended to a hundred people within the group’s credit branch, due to its ease of use.