Core banking modernization: Unlocking legacy code with generative AI

 Early trials with generative AI present that the tangle of legacy code that hinders each side of the transformation of banks’ core services and products might be rapidly and successfully resolved. By Alvaro Ruiz, International Core Banking Lead, Accenture.

On the coronary heart of most conventional banks is a mainframe pc operating the software program that defines and drives the group’s core processes. That is the place most of the financial institution’s issues begin. To satisfy prospects’ and regulators’ rising calls for, to enhance price effectivity, and to maintain up with the speedy tempo of change – which incorporates making the most of rising applied sciences – banks want the agility to innovate rapidly. This isn’t one thing the business is known for.

The reason being that the core system behind each banking product contains many hundreds of strains of pc code. Most of it was written a long time in the past, in languages that may now not meet the wants of contemporary banking. Over time it has develop into multi-layered, convoluted and fragile. This is because of repeated modifications, add-ons and connections with newer functions, and the final tendency to retain slightly than decommission outdated functions. These points, along with the truth that the code tends to be poorly documented, makes it obscure and improve it.

Banks have lengthy wished to re-engineer or change this code – convert COBOL to a contemporary language like Java, for instance – however this could be a mammoth job requiring a whole bunch of engineers and taking a number of years. Most supply code programmers have lengthy since retired, making it troublesome and costly to search out and recruit them in ample numbers. This, in flip, provides to the danger of the train. And so it stays on the again burner, an necessary precedence however one that’s laborious to justify within the present circumstances.

Legacy tech: A expensive roadblock to banking innovation

Ultimately, nonetheless, we’re seeing gentle on the finish of the tunnel. We consider that generative AI, mixed with new composable, interoperable and coreless architectures, may provide the answer to this decades-old drawback. We’re working with a handful of banking purchasers to check this strategy, and the early outcomes are very promising. As my colleague Michael Abbott, Accenture’s international banking lead, advised American Banker: “It’s early days, however we’re seeing 80% to 85% accuracy.”

The method begins by utilizing specialist generative AI fashions and a course of known as retrieval augmented technology (RAG) to reverse-engineer the code. This permits us to know and doc the necessities which the code is designed to satisfy.

The subsequent step is forward-engineering, for which we see two major paths. The primary is the automated recoding of the software program into a contemporary and versatile language, utilizing an ISO-functional strategy and/or specialised generative AI fashions, to re-imagine the performance that’s required to satisfy the present goals of the expertise and/or the enterprise. We then use generative AI to routinely check each a part of the brand new code and its efficiency, and to facilitate the transition from a mainframe {hardware} and software program stack to a contemporary set of frameworks and compute expertise.

Put merely, we change the outdated code with new packages which can be easier and extra versatile, and that help the financial institution’s modernization technique.

“It’s early days, however we’re seeing 80% to 85% accuracy.”

Having composable, interoperable and coreless architectures is essential to most banks’ modernization, because it permits elements of their legacy functions to co-exist frictionlessly with modernized elements and even with third-party merchandise from totally different sources. It additionally permits banks to make use of totally different internet hosting fashions on the similar time, which is crucial to assembly new enterprise necessities and producing well timed enterprise outcomes whereas legacy modernization is beneath manner.

The advantages of this strategy are doubtlessly game-changing. A core modernization undertaking that in some reported circumstances price greater than US$700 million and prompted years-long enterprise bottlenecks throughout their execution may now be accomplished in a fraction of the time with no unfavourable impression on the enterprise. The fee and danger benefits are apparent. In latest work with a world financial institution we transformed 25,000 strains of legacy code, reducing the reverse-engineering effort by 50% and boosting testing effectivity by 30%. This saved greater than 50% of the unique finances.

Extra importantly, our evaluation signifies that banks that modernize their core may doubtlessly enhance their return on fairness by 8.3 share factors by bettering their manufacturing, distribution and servicing. They might additionally dramatically facilitate danger administration and regulatory compliance.

It’s after all very important that we guard towards the tendency of generative AI to plagiarize and, if it doesn’t know the reply, to manufacture. RAG helps on this regard by drawing on the financial institution’s information base, together with its current code repository, slightly than unverified exterior sources. Nevertheless, till generative AI matures and these failings are remedied, the brand new code does should be fastidiously checked and examined.

From legacy to main with a contemporary digital core

Each financial institution is aware of {that a} trendy digital core is important to its capability to compete and meet buyer wants. I consider we’re on the cusp of placing this elusive objective inside the attain of each monetary companies group – rapidly and affordably, whereas conserving a good management on danger.

To search out out extra about this subject now we have a bit titled The Key to the Core in our Top 10 Trends for Banking in 2024. We’ve additionally simply printed a brand new report, The Age of AI – Banking’s New Reality, which explores the potential position of generative AI all through the financial institution.

If you want to learn how our code conversion trials are progressing, please contact me immediately at LinkedIn.

Disclaimer: This content material is offered for basic data functions and isn’t supposed for use instead of session with our skilled advisors. Copyright© 2024 Accenture. All rights reserved. Accenture and its emblem are registered emblems of Accenture.