AI in payments modernization: Leveraging the full potential  

Modernizing payments architecture is a top priority for financial institutions, with a recent survey by KPMG finding 79% of US banks are planning to modernize multiple payment systems over the next few years. This determination is based on multiple drivers, from the emergence of new standards like ISO 20022, to a tightening competitive environment as fintechs and technology companies muscle in on the payments space. Put simply, updating legacy infrastructure has become imperative. 

However for banks and other established payments businesses, modernizing can be a painful prospect. Aging systems have often been bootstrapped together by previous generations of software professionals who have since retired, taking their knowledge of the intricacies and dependencies of that architecture with them. Especially in the payments space, the consequences of any amount of downtime or disruption can be severe. Rather like an old house that has been extended over the centuries, organizsations are nervous about making renovations, in case what seems to be a plasterboard partition turns out to be a supporting wall. Small wonder that in one study by Forbes, banking leaders cited legacy infrastructure as their number one challenge.

Putting AI to work

In working closely with clients on modernization projects across the financial services sector, we’ve been trying to share some good news: Modernization is not the same process it was a few years ago, and businesses can draw on new forms of support. A lot of hype has accompanied the emergence of GenAI, but modernization is one area where we’re already seeing it deliver on a regular basis. In fact, AI can help institutions address challenges throughout every stage of the legacy modernization process – provided it’s understood, and applied in the right way.  

While it’s relatively new, the use of AI in modernization is already surrounded by misconceptions. One is that it’s most useful as a tool to write code. AI can indeed make software engineers more productive – typically by 10% to 30%, according to our reckoning. But coding is only a small element of the modernization process, and isn’t where we see AI delivering the most advantage. 

Another misconception is that AI is a ready-made, all-in-one tool that can simply be set loose to perform the tasks or deliver the answers the organization is seeking. An ‘all or nothing’ mindset around AI means when it’s deployed without much planning or supervision and (inevitably) fails to meet expectations, businesses quickly default back to more traditional, labor-intensive ways of working. In this way they miss an opportunity. 

We believe the most transformative impact from AI will be in mainframe modernization, and enhancing understanding of legacy codebases, particularly those that have little in the way of reliable documentation - two of the most problematic aspects of the entire modernization process. 

We’ve developed an accelerator called CodeConcise that helps human software engineers rapidly identify the dependencies and interconnections within a codebase, translate the code into current computing language, and then reverse engineer it into a modernization project.

At the same time, we’re encouraging clients to leverage AI for more than migrating legacy systems into modern stacks. The reality is that today’s modern code is tomorrow’s legacy, and simply re-engineering systems will create new legacy-related issues that need to be resolved in the future.

To be future-proof, modernization should be grasped as an opportunity to incrementally re-engineer business processes, as well as systems. Part of this is identifying which parts of a legacy system map to specific domain capabilities, so focus can be applied to reverse engineering the areas that are of highest strategic value to the organization or its customers.  

This level of re-engineering also requires business flows to be mapped out in a way that business stakeholders can easily comprehend and reference. Providing that visibility enables these stakeholders to give input on areas of strategic priority, and where there’s potential for processes to be retooled or improved. 

GenAI can provide a solid foundation to develop this kind of understanding and visualization. We’ve worked with clients to develop capability trees that assign subdomain capabilities at the code level, which are then aggregated into larger, more abstract domain capabilities,  until they converge at the highest level of abstraction at the level of business domains, such as compliance or customer service. 

The result is an accurate blueprint of the enterprise that can be enhanced by creating additional documentation, such as glossaries or flowcharts, associated with each domain. This paves the way for a more informed assessment of which areas of the old system should be focused on and re-engineered first, encouraging the organization to pursue modernization incrementally, rather than attempting to overhaul the entire system at once.

Steps on the path to value 

The incremental approach is core to what we call value-driven modernization – pursuing tangible, relatively quick wins in areas that matter to the organization, to underpin the business case for modernization and build momentum. In a vital function like payments, incremental steps also help minimize the likelihood of disruption or governance oversights.

Modernization can be made a lot easier with the transparency and understanding GenAI brings to the process. However we still urge customers to seek advisors, rather than simple AI vendors, to guide them on this journey. The fact is GenAI is only one possible item in the modernization toolkit; a plethora of other techniques, rooted in both machine learning and other technologies, can be employed to analyse and reverse engineer legacy systems. The most effective approaches will leverage a variety of solutions to provide the best starting point for modernization, and ongoing support. 

AI and other tools also need to be tailored to each customer’s context. Different technology stacks may require different types of analysis, solution or techniques to be effectively understood or updated; depending on GenAI alone risks it becoming a generic regurgitator of information that does not add any real understanding of the functioning of the legacy system. Having an experienced partner that can work with the organization to identify the optimal combination of tools and approaches to apply, and to highlight additional relevant (and sometimes implicit) domain knowledge when GenAI solutions are used to answer questions or generate documentation, can significantly enhance the efficiency and outcomes of a modernization program. 

Fundamentally, we counsel financial institutions to see GenAI for what it is: not a silver bullet, but a tool that can help reverse and forward engineering teams find a better starting point, and inform productive conversations with business stakeholders. It’s hard to overstate the positive impact this can have in terms of ensuring teams have a clear sense of priorities, and maximum bandwidth. 

GenAI should add to the impetus for banks and other payment providers to embark on their own modernization journey as soon as possible. As standards shift and payments increasingly become a tech industry, hesitating or failing to take action are starting to look like greater risks. 

Thoughtworks has both the technological capabilities and industry expertise to identify exactly where AI-driven automation is valuable, and where more traditional software approaches should be retained in a payments modernization program. We partner with banks and other enterprises to drive the transition to composable architecture that keeps them at the cutting edge of payment innovation and smart compliance, ensuring they’re prepared for a future in which they must be innovators in their own right.