Digital Economy

Can artificial intelligence help the refactoring of obsolete systems

Says Paolo La Torre, Managing Director of Capco Italy

23 September 2024

2' min read

The management of many banks, financial institutions and public administrations is still entrusted to information platforms built with COBOL (COmmon Business-Oriented Language), a programming language that shows all its years (it was invented in the United States in the 1950s): rigidity and high maintenance costs.

Dismantling this management system is not easy, but it is necessary: if you do not change, it is impossible to innovate the offer, to bring the quality of services up to the level that users now expect and, ultimately, to remain competitive. The problem is that dismantling is a costly and complex operation, and therefore a lengthy one: a system generally governs many of an organisation's services at the same time, and you cannot do everything at the same time. In addition, change, at least until now, required professionals with knowledge of both languages, the old and the new. It is like looking for a bi-lingual translator to translate a text: they are there, but they are few.

Banks, in recent years, have made significant strides in change and therefore do not have to innovate in all service areas, but in some yes, even strategic ones, for example payments, one of those on which user attention is highest and customer loyalty is at stake, especially of the youngest.

Generative AI can help evolve towards more efficient information systems and migrate to the cloud. It can, to put it in computer jargon, facilitate the refactoring of legacy (i.e. obsolete) systems. The transition from the old programming language has so far been done almost manually, implying the aforementioned need to know the two languages. With generative AI, on the other hand, it can be done with source-to-source compilers, which allow the examination of millions of lines of code in reduced times, thanks to the automatic translation of text from one language to another (Machine Translation) and the neural networks that enable it.

Technological innovation, however, is not neutral: that is, it cannot be introduced leaving everything as it was before. One must, first, understand how the organisation uses data and how it is shared by the different functions. Business management and development policies must therefore be reviewed, ensuring that systems are always scalable, secure and flexible and that migration to the cloud does not entail excessive costs.

The audit trail, schematically, first involves an automated analysis of the code of the applications in use (whatever their language) to assess the system configuration. This is followed by the classification of data and the definition of a taxonomy of legacy systems and their monolithic applications based on industry standards. In the case of banks, the Banking Industry Architecture Network (BIAN) or ISO 20022, an international standard for the exchange of electronic messages between financial institutions.

It is the combination of these two phases that ensures the balance between the quality of the systems and the resources deployed. From this point on, the planning of interventions begins, which must assign a priority level to the breakdown of each monolith. This planning is the best guarantee for successful refactoring and facilitating the maintenance of the new system.

* by Paolo La Torre, Managing Director of Capco Italy