Bridging Traditional Infrastructure And Intelligent Automation: The Role Of AI/ML In Banking IT System

Abstract

AI/ML technologies are now maturing, available in performance and price, and secured as well as enabled by evolving and well-accepted cloud environments based on shared or hybrid infrastructures. AI/ML and advanced analytics are rapidly introducing new capabilities in banking as part of the digital transformation of traditional IT systems. On the one hand, IT systems traditionally have been a siloed set of specific applications, but during the last 50 years the R&D efforts were often focused on coping with the increasing complexity of legacy systems, accomplishing virtualization, and ensuring security and reliability. On the other hand, in recent years the emergence of new methodologies and new technologies has generated the possibility of advancing the functionality of already available systems. The first products of this opportunity have been realized as AI/ML systems that provide online deployment of pre-trained models. While the cost could be high, especially in the case of generating the training data set, for many use cases it could be well worth the investment. However, generative AI and other new capabilities are now challenging the banks in providing sufficient robustness and reliability in the production environment of on-premise systems. The speed and applicability of these newer technologies are foreseen by the R&D department of banks that made the necessary investment into the knowledge and resource foundation.

Today, the available options are: 1) low-risk options for the digitalization of specific services--mostly in the front-end applications and 2) (often high-risk) options into designing new solutions--in-house, partner-based or third party sourced. It is aimed to review the pros and cons of the existing options, and map the bank readiness in adopting them. [1]A project to assess the data readiness for CDP through a review of the existing data domains of clients will be outlined as well. All opportunities would depend on addressing the needed data and skill investment areas, but even with less than the project-worthy readiness there are still adequate low-risk solutions. On the one hand, data readiness needed to draw upon a well-defined data architecture framework through high-level architectural views to narrow down gap areas. The recently developed microarchitecture and opened libraries would be reviewed, along with the higher-level architectural elements developed for banking IT systems to serve as guidelines. On the other hand, to acquire skill readiness, an educational framework is needed to improve domain knowledge with aligned roadmap options for newly arrived technologies like AI/ML.