Agentic AI integrates automated and semi-automated processes to reduce translation time by at least 50% with 85% execution-ready accuracy.
Financial systems worldwide remain heavily reliant on COBOL, a programming language first introduced in the 1960s. Despite its age, this language supports critical infrastructure such as ATMs, stock exchanges, and banking platforms, with an estimated 800 billion lines of code still in operation.
Yet, this dependency on legacy technology is increasingly fraught with risks. Transitioning to more adaptable and contemporary technologies necessitates the modernisation of COBOL — but this is no simple task. Here, I explore how Agentic AI is addressing these challenges to make modernisation both efficient and forward-looking.
The Risks of Continued COBOL Dependency
COBOL underpins approximately 43 per cent of all banking systems worldwide, maintaining a significant presence in the financial sector. European banks, in particular, remain reliant on COBOL, following deeply ingrained and conservative IT practices.
However, this longstanding reliance comes at a mounting cost. The dwindling pool of qualified COBOL developers, the vulnerability of outdated systems to modern security threats, and the difficulty of integrating new technologies collectively pose serious challenges. To date, efforts to modernise these ageing COBOL-based systems have been limited and piecemeal, leaving much to be done.
The Challenges of COBOL Modernisation
Modernising COBOL systems is no minor undertaking. The obstacles are multifaceted and deeply entrenched, including the following key issues:
Rigid data structures: COBOL relies on fixed data types and structures, which stand in stark contrast to the flexibility of modern programming languages like Python. Migrating such legacy data formats is both time-consuming and complex.
Erosion of expertise: With the average age of COBOL developers steadily increasing and insufficient training opportunities available, the risk of a critical talent shortfall looms large.
Legacy systems and limited reproducibility: Many COBOL environments in active use today lack comprehensive documentation or well-maintained code. Additionally, some traditional functions, such as IBM’s CICS transaction processing, are challenging to replicate within modern frameworks.
Revolutionising COBOL Modernisation with LLMs and Agentic AI
A breakthrough solution has emerged in the form of Agentic AI, a class of AI agents that do more than simply analyse code and data. By autonomously making decisions and adapting to evolving requirements, these agents bring a much-needed dynamism to the modernisation process.
Agentic AI leverages advanced cognitive architectures to develop robust translation strategies. Unlike traditional approaches, this method requires minimal human intervention, thanks to intelligent planning and self-correction mechanisms that reduce errors and improve efficiency.
A New Approach to Agentic AI Solutions
SoftServe is at the forefront of using Agentic AI to provide tailored solutions for COBOL modernisation. These solutions facilitate seamless translation of COBOL into contemporary languages such as Python and Java, addressing the unique requirements of each client’s IT environment.
A proof of concept developed by SoftServe delivered some remarkable results, including:
Accelerated tanslation: Agentic AI integrates automated and semi-automated processes to reduce translation time by at least 50%.
Cost savings: Companies can cut project costs by 15-20% by reducing reliance on specialised development teams and extensive manual oversight.
High reliability: Without requiring extensive technical support, 85% of translated code lines were executed with complete accuracy.
Building a Resilient Code Base for the Financial Industry
The modernisation of COBOL systems has become imperative for financial institutions aiming to remain competitive in an increasingly dynamic marketplace. For the first time, Agentic AI offers a scalable and innovative approach to making this transition more effective and less risky.
If you are keen to learn more, join me at the „AI in Banking 2025“ event on 3-4 September, where I’ll be hosting a masterclass on COBOL modernisation and its critical role in shaping the future of financial technology.
Author
Antonina Skrypnyk is a Head of FSI EMEA at SoftServe, with more than 11 years extensive industry experience delivering high-impact results.
Antonina oversees the business of SoftServe financial services Clients portfolio in Europe, Canada, Singapore, and the Middle East.
Antonina’s experience extends from leading projects in Enterprise crediting, credit portfolio management, and financial investment projects. She also leads high-profile initiatives in development cycles of innovative products acceleration via SoftServe’s Financial Services Lab.
Antonina supports complex negotiations on delivering IT solutions during pre- and post-sales, and the on-site activities with financial Services clients