Banks are accelerating AI adoption, but Nitin Rakesh, CEO and Managing Director, Mphasis.ai, says decades-old legacy infrastructure remains a major barrier to innovation, security, scalability and long-term competitiveness.
Scratch beneath the surface of even the most sophisticated global bank and you will often find technology that predates the Internet itself.
Over 90% of the world’s largest banks are still operating on legacy IT systems built in the 1970s and 80s, many of them powered by COBOL and similarly outdated programming languages. These systems were engineered for stability in a different age – one of limited connectivity, predictable demand and slower cycles of change. Today, they are being asked to support a radically different reality: one defined by real-time transactions, hyper-personalised services and the rapid rise of artificial intelligence (AI).
This is the uncomfortable truth facing the banking sector. The industry is attempting to build its future on foundations laid half a century ago.
The collision course between AI ambition and ageing infrastructure
Customer expectations have changed beyond recognition. Banking is no longer confined to physical branches or even desktops – it is embedded in everyday life, accessible instantly via mobile devices and expected to work seamlessly across borders and time zones. Increasingly, customers assume that AI will anticipate their needs, detect fraud before it happens and deliver frictionless, intuitive experiences.
Banks are responding in kind. From AI-powered fraud detection and risk modelling to algorithmic trading and intelligent customer service, institutions are racing to embed machine learning into every layer of their operations.
Yet there is a fundamental disconnect.
More than 43% of banking systems still rely on COBOL. The average core banking platform is between 20 and 30 years old. And in the past five years, only 18% of banks have undertaken meaningful core modernisation. These figures point to a structural challenge that cannot be ignored: the infrastructure underpinning modern banking was never designed to support the data intensity, speed or integration demands of AI.
The result is a growing tension between ambition and reality. Banks are under pressure to innovate, but they are doing so on systems that constrain their ability to execute safely and effectively.
What lies beneath: risk, fragility and hidden complexity
Legacy systems are often described as a barrier to innovation. In truth, they are something more complex – and more dangerous. These platforms have endured because they work. They process vast volumes of transactions reliably and have been refined over decades. But their resilience can mask underlying fragility. Many are highly customised, poorly documented and dependent on shrinking pools of specialised talent. Small changes can have unpredictable consequences, making transformation both risky and resource-intensive.
Layering AI into this environment without addressing its limitations introduces a new class of risk.
Data is the lifeblood of AI, yet legacy systems often operate in silos, generating fragmented and inconsistent datasets. This undermines the accuracy and reliability of machine learning models, leading to flawed insights and potentially costly decisions. At the same time, older architectures often lack the robust security frameworks required to withstand modern cyber threats, increasing exposure as systems become more interconnected.
Regulatory pressure adds another dimension. Data governance, transparency and accountability are no longer optional; they are central to maintaining trust. Banks that deploy AI without ensuring the integrity and traceability of their data risk not only operational failure but also reputational damage and regulatory penalties.
Perhaps most critically, legacy infrastructure limits speed. In a competitive space shaped by fintech disruptors and digital-native entrants, the ability to iterate quickly is a strategic advantage. Yet many incumbent banks remain constrained by systems that make even minor changes slow, complex and costly.
The illusion of quick fixes
Faced with these challenges, some institutions are tempted to pursue rapid AI adoption as a shortcut to transformation. This is a mistake.
AI is not a layer that can be added on top of existing systems to deliver instant results. Without a modern, flexible and well-governed data foundation, its impact will be limited at best and damaging at worst. In some cases, rushed implementations can exacerbate existing weaknesses, creating new points of failure in already complex environments.
Equally, the idea of wholesale system replacement – ripping out legacy infrastructure and starting anew – is often impractical. The scale, cost and risk involved make such approaches difficult to justify, particularly for institutions that must continue operating without interruption.
The reality is that there are no shortcuts. Modernisation is not an event; it’s a journey.
Rewiring the engine: a pragmatic path to modernisation
The banks that will succeed in this new era are those that approach transformation with both urgency and discipline.
Rather than attempting to replace legacy systems outright, leading institutions are adopting phased, incremental strategies. This includes decoupling core components, introducing API-led architectures and leveraging cloud technologies to create more flexible and scalable environments. These approaches allow banks to modernise progressively while maintaining operational continuity.
Within this framework, AI becomes a powerful enabler. It can be used not only to enhance customer-facing services but also to optimise internal processes, identify inefficiencies and accelerate the modernisation journey itself. For example, AI-driven tools can help analyse legacy codebases, automate testing and support the migration of workloads to more modern platforms.
Crucially, successful transformation requires a renewed focus on data. Establishing robust data pipelines, improving data quality and ensuring strong governance are foundational steps. Without them, even the most advanced AI capabilities will fail to deliver a meaningful difference.
Survival of the most adaptable
The stakes for the banking sector are high. This is not simply a question of improving efficiency or reducing costs; it is about long-term competitiveness and, ultimately, survival.
Banks are custodians of trust. Their customers expect reliability, security and transparency. Meeting these expectations in a digital-first world requires more than incremental change; it demands a fundamental rethinking of how technology supports business.
Those institutions that navigate this transition successfully will be defined by their ability to balance innovation with resilience. They will recognise that AI is not a silver bullet, but a double-edged sword – capable of both amplifying strengths where foundations are solid and intensifying risks where they are not, depending on how it is deployed.
The uncomfortable reality is that many banks are still at the beginning of this journey. They are managing the complexity of maintaining decades-old systems while simultaneously trying to build the capabilities of the future.
But the direction of travel is clear. In a world where technology defines competitive advantage, the ability to modernise legacy infrastructure – thoughtfully, strategically and at pace – will separate the leaders from the laggards. AI will play a central role in this transformation, but only for those institutions willing to address the foundations on which it depends.
The question for banks is no longer whether they can afford to modernise. It is whether they can afford to delay.
