Risk Management and Ai: Towards more performance

In concrete terms, what are the main needs of the companies that turn to you?

We are primarily engaged by companies in the management of their customer and supplier relationship. We help them in the analysis of the financial health of their trading partners whose risk of default is assessed over a time horizon of one year, and beyond according to needs. Accurate and rapid analysis of financial information is a challenge even for experts in the field.

That’s why we have developed a simple-to-use scoring system on a scale of 1 to 10. Our customers thus easily integrate this indicator into their decision-making process, which evaluates the sustainability of their business partners.

Depending on their aversion to risk, they set the threshold from which they decide whether or not to finalize their relationships. For some of our customers, we have the possibility to propose a custom-made scoring system, which adapts to their risk policy, for better decision-making.

In addition, our customers are also interested in our credit advisory which allows, in complementarity with the score, to assess the monthly purchasing capacity that their trading partner is able to cope with. The credit advisory is therefore linked to the score, in the sense that it decreases with a risk of default that increases.

In this context, how do you capitalize on AI?

Ellisphere has been working on the integration of AI since 2015. We have moved from 2nd generation systems based on a mix of statistics and expert rules, to a 3rd generation system based on purely statistical models. Today, our algorithms rely 100% on AI to establish the scoring formulas independently.

To build the 3rd generation scores, we relied on the information of millions of French companies through a data history of more than 30 years, as well as on our business expertise in order to have the best predictive performance possible. Our Gini index has increased from 70% to 80% thanks to the use of AI.

What are your stakes?

AI in scoring systems is only in its infancy, and will grow with the enthusiasm of private and public actors around its use. Moreover, AI is still seen as a black box, which is why many companies have a certain apprehension of using it in finance-related activities. To meet this challenge, we have transformed our algorithms into “white boxes” to allow our customers to understand the reasoning behind our scores.

Our second challenge is to provide constant quality data in greater quantity to our score models, and therefore to think about new sources such as unstructured data.

For example, we are in the process of setting up a scoring system based on the press information published about companies, which, after analysis, is able to determine whether the company will experience economic and/or financial difficulties in the near future.

Interview published in the January 2020 issue of the magazine La Jaune et la rouge de Polytechnique