Today’s business banking models are changing. Internal rating models enjoyed the spotlight for a long time, but new topics are already catching on. Machine learning and artificial intelligence are key new elements in addition to traditional techniques for model development.

EU and international regulators have also taken an active interest in AI and machine learning, and while they recognize the benefits that both can bring to financial markets, consumers and their own work, they are also increasingly mindful of the potential risks and unintended consequences that the use of these new technologies may have. AI and machine learning topics have been addressed in the IFC-Bank Indonesia International Workshop and Seminar on Big Data in Bali held on 23-26 July 2018, and a 970-page document with contributions was published recently (source: https://www.bis.org/ifc/publ/ifcb50.pdf).

At a worldwide level, central banks consider that the opportunities these new techniques provide can be significant. But we need to address related challenges, including accuracy, transparency, accountability and compliance.

Banks already experimenting with AI

Everyone knows that it is a journey. But banks around the world are now experimenting with machine learning and AI approaches, thanks to the availability of new technologies. These allow banks to address almost all related challenges.

For example, the transparency of machine learning techniques was once the main constraint. But now interpreters have been built to provide an understanding of the business logic behind the results. This is similar to the business logic that traditional model techniques – such as logistic regression and tree – easily provide.

How SAS can help

GARP and SAS conducted an online survey in December 2018 to learn more. The survey, Artificial Intelligence in Banking and Risk Management, included more than 2,000 total responses from across the financial services industry, including banking, investment banking/securities and wealth/asset management. It also covered different departments, including risk (48%), finance (14%) and IT (9%), as well as several stakeholders, including executives (28%), team leader/senior manager/manager (36%) and analyst (31%).

Executives, risk analysts and IT all have different needs. Executives are looking for automatic, user-friendly and ready-to-use results and reports, governance and transparency of the entire process, and better business results.

Whereas the main needs of risk analysts cover topics like:

• Automation of manual processes through an integrated interface for data preparation, data exploration, model development and evaluation, and model deployment.
• Machine learning model interpretability.
• Open source use and integration.
• Easy model deployment.
• Better statistical results.

IT users need to govern the whole platform and integrate different programming languages. They appreciate easy model deployment and use in a production environment, as well as better performance execution results.

In Italy, we are running a workshop called Next-Generation Credit Scoring, AI & Machine Learning on Feb. 13 (in Italian). We will explore the topics of machine learning and AI, as well as the needs of professionals like executives, risk analysts and IT.

We recommend reading our latest white paper, The Future of Risk Modeling.