• Ramón Trías and Irving Juárez, executives of the AIS Group, a pioneer in the development of Artificial Intelligence applications, show their potential in the financial sector, improving the granting of loans and reducing defaults and fraud.

• The development of AI applications, particularly Machine Learning models, focused on financial activity, allow us to better manage risk, optimize portfolio monitoring, detect signs of possible default in advance, automatically appraise real estate, prevent fraud and optimize the credit recovery, among others.

"Beyond fashions, Artificial Intelligence must generate real value in the company," says Ramón Trías, president of AIS Group, a company specialized in the application of Artificial Intelligence (AI) for the development of decision-making systems based on in data, also known as data-driven.
A pioneer in the development -in 1987- of artificial intelligence applications in Spain and various Latin American countries, including Mexico, the president of AIS affirms that AI is adding value in different sectors, such as healthcare, transportation or the selection of personnel . But if a sector can be a catalyst for the use and benefit that these techniques can bring, it is the financial sector. "Banking currently immersed in digital transformation processes has enormous potential in the application of AI," he says.
Trías points to different experiences carried out by his team in different entities where Machine Learning (ML) techniques, from the AI family, allow working with a greater number of variables and finding more interactions, multiplying the predictive power of the models who use them against traditional statistical techniques. "They make it possible to improve the granting of loans, by attracting customers with a lower probability of default, offering a hit rate that doubles that of traditional techniques, which translates into a highly positive impact on commercial and business objectives."
In addition, for him, working with these methodologies brings enormous advantages in detecting fraud or recovering defaults, because “by offering these models a more granular score, they allow managing groups of unpaid customers that are homogeneous and very different from the others , and define differentiated strategies for each group, even going so far as to determine which is the best strategy for each one of them ”.
The President of AIS Group adds that, in the field of AI, the figure of the expert is very relevant. “Many believe that artificial intelligence learns on its own and is totally autonomous, but it is false. Artificial intelligence has limitations. She is like a child who must learn and, like children, she needs parents to accompany her, to show her new cases, to avoid overfitting, to avoid bias. The role of the expert is fundamental, since he is the one who will make it possible for artificial intelligence to bring real value. ”
In which areas can Machine Learning Models be applied?
Irving Juárez, AIS director for the region of Mexico, Central America and the Caribbean, comments that “for some time now, it has been normal to think of AI applications to strengthen risk management, boost commercial and business strategy, solve needs in processes, among others. In this context, Machine Learning methodologies stand out for different reasons: the new Big Data platforms are more compatible with these models; Thanks to technological evolution, very acceptable response times are achieved and, in some cases, they can be managed with free software (open source); and finally because conventional methods use few variables, which limits the prediction (anticipation) ”.
According to the president of AIS Group, the results obtained in recent projects reveal that the level of success in the models improves between 25% and 50% when using Machine Learning algorithms compared to traditional techniques.

Fighting the black box effect

According to what the AIS manager commented, the most recent challenge has been the application of Machine Learning methodologies within financial institutions in areas related to risk management, says Juárez. “Despite the excellent predictive levels, these algorithms used to be seen as black boxes, which were not feasible to be approved by the risk, audit or comptroller areas, and much less by regulatory entities. At AIS Group we have successfully implemented Machine Learning models in various fields, thanks to a significant investment in R&D, which allows, through processes and methodologies, to document and make its operation transparent. This has opened the doors for us to implement these models in recent times within risk areas and very important entities in Spain, Mexico and Chile, among other countries. ”

How is development in north american regions?

Juárez highlights that “many local financial institutions are working on digitization and cybersecurity processes. In this framework, the implementation of Machine Learning models will become, in the short term, a strategic factor for competitiveness. AIS Group is a pioneer in the implementation of these models in first-class financial institutions, applying them in processes for the early detection of alerts that allow anticipating default before it occurs, which offers many more alternatives to find an adequate solution that guarantees the viability of the client (company) and the entity's business ”.
The director of AIS in Mexico also highlights that Machine Learning models allow the incorporation of data that until now was not being used, or at least not automatically. This is the information that is published on social networks or in the media about clients (companies), which can be very powerful early indicators of future problems that may later be reflected in the client's balance sheet and, therefore, in their payment behavior.
In fact, says Juárez, we are currently implementing, in one of the main banks in Mexico, a process based on a Machine Learning methodology that allows for timely monitoring of the deterioration of the loan portfolio, incorporating unconventional sources of information ( social networks or the media), in order to strengthen traditional credit analysis, and provide timely, dynamic and efficient indicators and tools that help strengthen customer monitoring and promote policies and actions in this regard. ”