From traditional to trailblazing: Generative AI’s impact on fintech software development
In the pursuit of innovation, a notable 91% of enterprises have prioritized modernizing their legacy applications, as reported by VMware. This push for modernization has yielded significant results, with a May 2022 Forrester report revealing that 77% of enterprises have witnessed substantial revenue growth.
Even traditionally slow-to-adapt financial services companies now recognize the urgency of modernization. Their main hurdle in achieving innovation lies in their legacy systems. While industry leaders envision banks thriving for another century, it’s evident that this vision can’t be realized if outdated core systems continue to weigh them down. This presents a prime opportunity for advanced fintech software development to yield modern solutions, empowering them with competitive capabilities in a rapidly evolving market.
Whether they provide commerce-enabling solutions to global merchants, operational solutions for financial institutions, or products and services for digital payment transactions, they require the necessary tools, resources, and tech-savvy talent to innovate and meet rising demands. Despite the considerable benefits of legacy modernization, embarking on such a digital transformation can appear daunting for many fintech companies. This is primarily because on average, code modernization incurs a cost of $1.5 million and takes about 16 months to complete.
Enter generative AI: Augmenting developer productivity
Fortunately, generative AI presents many applications in financial services, including an effective solution for modernizing legacy code without the need for full-blown, disruptive, and costly digital transformations. The beauty of this approach lies in its ability to enable existing development teams to collaborate with AI, eliminating the need to hire multiple experts in modern technologies. The concept is straightforward: let generative AI handle a portion of the coding while developers focus on crucial tasks. Developers can guide generative AI in replicating and advancing the functionalities of the legacy core, resulting in time and resource savings.
In fact, at Grid Dynamics, we conducted a study to quantitatively assess the performance improvements achieved by generative AI tools across various software development tasks and project types. Simultaneously, we evaluated the organizational and process implications of these tools and provided best practice recommendations for engineering teams looking to adopt generative AI.
Generative AI had a significant impact on various development tasks, and we measured metrics for eight standard tasks in our crowdsourcing study. We found that it influences developers’ productivity at every task level.
While the results from our study are fascinating, let’s explore its implications for the fintech industry. We’ll first delve into the challenges faced by fintech companies offering comprehensive solutions like transaction processing, electronic bill payment, business process outsourcing, and other software systems. Then, we’ll explore how such fintech pioneers can harness the potential of generative AI to streamline software modernization based on our research findings.
Managing complex operational dynamics
Financial technology companies often deal with intricate payment journeys, involving multiple backend processes from authorization to settlement. Streamlining operations for merchants, banks, credit unions, and other financial institutions while catering to diverse market needs is a multifaceted challenge.
💡Example scenario: Consider a fintech company that offers online payment processing services to merchants. Each payment involves a complex journey, starting from the customer’s authorization and ending with the settlement of funds. This journey includes multiple backend processes, such as fraud detection, currency conversion, and reconciliation. Streamlining these operations while catering to the unique needs of various merchants, each with their preferred payment methods and currencies, presents a substantial operational challenge.
Harmonizing diverse interface standards
Catering to a plethora of clients necessitates harmonizing disparate interface standards across merchant acceptance, fintech operations, and payment networks as well as reducing confusion for developers who must work with varying data structures and operations. Ensuring a consistent customer experience across varied solutions becomes a high-stakes endeavor.
💡Example scenario: Imagine a fintech firm that provides payment gateway solutions to a range of clients, including e-commerce platforms and traditional brick-and-mortar retailers. These clients may use different data formats, communication protocols, and security standards for processing payments. To harmonize these disparate interface standards, the fintech company must invest in robust middleware and API translation layers to ensure a consistent payment experience for all clients and their customers.
Ensuring agility in compliance
Mastering regulatory compliance, especially in transaction processing and electronic commerce, demands agility. Safeguarding against financial risks, crimes, and data breaches is paramount. Yet, legacy systems and scattered data can hamper real-time vigilance. Adapting to evolving compliance without operational hiccups is a complex task.
💡Example scenario: A fintech startup operates as a cross-border payment aggregator, serving customers in multiple countries. Each jurisdiction has its unique regulatory requirements for financial transactions, such as anti-money laundering (AML) and know-your-customer (KYC) regulations. The fintech company must continuously adapt its compliance processes to align with evolving regulations in each region while maintaining uninterrupted transaction processing.
Balancing risk and speed
Efficiently managing risks associated with digital payments, especially in the global ecosystem, demands real-time monitoring and streamlined risk mitigation strategies. Balancing risk exposure while ensuring swift payment processing is a tightrope walk.
💡Example scenario: Consider a global fintech platform that facilitates peer-to-peer lending. The platform must efficiently manage the risk associated with lending to borrowers from diverse backgrounds and credit histories. Real-time monitoring of borrowers’ creditworthiness and developing streamlined risk mitigation strategies is essential to maintain a healthy loan portfolio while ensuring quick loan approval and disbursement.
Maximizing data value
Extracting value from the vast data generated in electronic transactions and financial operations is a complex task. Transforming fragmented data into actionable insights and enhancing its relevance across diverse operations remains an uphill battle.
💡Example scenario: A financial institution processes millions of credit card transactions daily. These transactions generate vast amounts of data, including transaction timestamps, locations, and transaction amounts. Extracting value from this data involves not only aggregating and analyzing it but also leveraging machine learning algorithms to detect unusual spending patterns that may indicate fraud. Transforming fragmented data into actionable insights can help the institution prevent financial losses due to fraudulent activities.
Meeting rapidly shifting market demands and swift integration of innovative payment capabilities requires an agile infrastructure. Overcoming siloed systems to respond promptly and roll out new products and services is a formidable challenge.
💡Example scenario: An established fintech company is facing increased competition and changing customer expectations in the payment industry. To stay competitive, the company must rapidly innovate by introducing new payment features, such as contactless payments or cryptocurrency integration. However, the challenge lies in seamlessly integrating these innovations into existing systems, especially when dealing with siloed legacy systems that hinder quick deployment.
Code modernization and API enablement: Tackling the challenges of legacy systems
Code modernization and API enablement are pivotal in addressing the multifaceted challenges confronting the fintech industry. Modernizing legacy code streamlines backend operations, resulting in heightened efficiency and responsiveness. This translates into swifter transaction authorization and settlement, simplifying intricate processes.
API enablement further enhances these efforts by harmonizing interface standards across solutions, ensuring a consistent user experience. It boosts compliance agility, facilitates rapid adaptation to changing regulations, and bolsters risk management with real-time monitoring and streamlined mitigation. Additionally, API enablement maximizes data value through simplified aggregation and analysis, empowering fintech companies for swift innovation and capability integration.
Generative AI roadmap for fintech modernization
As you embark on your code modernization journey with generative AI, a robust roadmap can help you with organizational readiness by transforming task execution, shaping team composition, expediting development processes, and reevaluating processes, especially across diverse project types.
Step 1: Identify suitable project profiles
Start by pinpointing projects within your fintech organization that are ripe for generative AI integration. Look for areas where automation and efficiency gains can make a substantial impact. For instance, consider modernizing some of your operational solutions used by financial institutions to run their business. Generative AI can streamline complex backend processes for deposits, loans, investments, and currency exchange, improving transactional efficiency.
Step 2: Establish working groups
Create cross-functional teams that include developers, data scientists, and domain experts. These groups should work collaboratively to understand the specific needs of your fintech projects. For instance, if you’re developing a solution to serve merchants globally, assemble a working group to focus on optimizing payment authorization processes.
Step 3: Define tooling and security policies
Generative AI requires robust tooling and security measures. Define clear policies for data handling, access controls, and model management. For example, when dealing with financial institutions, ensure that sensitive data is adequately protected through encryption and access controls.
Step 4: Initiate pilot projects
Start small to test the waters. Choose a pilot project, such as enhancing fraud detection in digital payment transactions. Implement generative AI models in a controlled environment, closely monitoring their performance and impact.
Step 5: Plan strategic changes
Based on the insights gained from your pilot projects, devise a strategic plan for broader generative AI adoption. Identify areas across your fintech operations where AI can optimize processes. For instance, if you provide operational solutions for financial institutions, explore how generative AI can streamline transaction reconciliation.
Step 6: Create a generative AI center of excellence (CoE)
Establish a dedicated center of excellence focused on generative AI. This CoE should consist of experts in AI research and development, serving as a knowledge hub for the organization. For example, if you’re offering commerce-enabling solutions to merchants, your CoE can continuously innovate payment processing algorithms.
Step 7: Scale up organization-wide adoption
With your CoE in place, scale up generative AI adoption across your fintech organization. This includes integrating generative AI models into your existing software development processes. For example, if you provide financial institutions with products for digital payment transactions, ensure seamless integration of AI-enhanced fraud detection capabilities.
Now, let’s explore a few use cases with a focus on fintech.
Code translation
Code translation involves converting code from one programming language to another, a critical task in fintech legacy migration projects, particularly when transitioning from obsolete languages like COBOL (still prominently used in ATM swipes and banking systems) to modern ones like Java. In our study, we used tasks that required translating substantial classes with business or technical components typically seen in enterprise applications. For instance, we translated the LRUCache class from Python to Java.
In our evaluation, we compared the code quality of the translated solutions. Initially, the raw output from GenAI tools had an average error rate of around 15 errors per 100 lines of code, including functional bugs and code quality issues. However, after addressing these errors, using GenAI tools resulted in performance enhancements of up to 45%.
To improve the effectiveness of GenAI in code translation, we found that employing multi-stage translation chains was highly beneficial. This approach involves converting legacy code into a natural language description of the business logic, then into a block diagram, and finally generating code based on this diagram. By using such techniques, productivity gains of up to 90% can be achieved, albeit with the need to develop custom tools for efficient implementation.
API schema development
An API schema is like a database schema definition but for APIs, to make integration between platforms easier for developers. A schema is metadata—data about how data is structured. This is important information for a developer wanting to interact with an API and its endpoints. For fintech, this may represent a concise blueprint that defines how financial data and functions are structured, accessed, and shared through an application programming interface (API). It encompasses endpoint definitions, data object specifications, API behavior contracts, security measures, versioning guidelines, and compliance details. This schema ensures secure and standardized interactions between financial systems, adhering to industry regulations and promoting efficient data exchange.
For example, the code snippet below shows a service the schema needs to be generated for:
Leveraging GenAI tools, we achieved remarkable results—completing the task in a mere 45 minutes, a substantial time-saving compared to the estimated 120 minutes without AI assistance. This translates to an impressive 62.5% reduction in time.
Given the task’s straightforward nature, fintech organizations can harness GenAI tools effectively. Our recommendation is to use precise prompts that specify details such as the OAS version and specific model requirements (like Lombok, Java records, etc.) to optimize API schema generation and boost efficiency in the fintech sector.
Conclusion
By embracing generative AI for legacy code modernization, fintech companies can shed the “laggard” label and revamp their innovative portfolio. While GenAI undoubtedly bolsters operational efficiency in tasks like code writing and API schema generation, its potential spans a wide array of software development tasks. The highest gains are observed for low- low-complexity, low-risk, routine tasks such as code translation, test writing, and documentation writing, where productivity improvements can easily reach 70-90%.
For a comprehensive understanding of GenAI’s influence across diverse software development use cases and guidance on selecting the right generative AI tools, delve into our complete study. We’ve also tailored recommended best practices for each use case. The effectiveness and suitability of GenAI tools hinge on the specific task at hand, underscoring the importance of tool selection and best practices.
Our Grid Dynamics experts advocate building a modular infrastructure that seamlessly integrates third-party and private language models (LLMs) with IDE plugins. Partner with us and launch proof-of-concept projects, tackle legacy migrations, and address technical debt and internal infrastructure enhancements. Reach out to us today to embark on your journey harnessing the power of GenAI.
