In 2026, artificial intelligence has moved from experimental enthusiasm to board-level expectation across the banking sector. For chief information officers (CIOs), it is no longer a question of whether to adopt AI, but how quickly it can be embedded into operations in a way that delivers measurable value. Yet the path to success remains uncertain.
The Massachusetts Institute of Technology’s State of AI in Business 2025 report found that 95% of enterprise generative AI pilot projects fail to deliver measurable returns within the first six months. In an industry where investment decisions are increasingly scrutinised, that figure presents a sobering reality. For banking leaders under pressure to demonstrate return on investment, AI success is proving far harder to achieve than initial hype suggested.
The Pressure to Prove AI Value
Across financial services, expectations are intensifying. A large majority of CIOs – around 92% – expect to harness AI by 2028 as a core part of their operating model. However, expectation alone does not translate into execution. The challenge now is not experimentation but scaling pilots into production systems that deliver sustained business impact.
At the same time, nearly 69% of CIOs are pursuing highly customised AI strategies, favouring in-house development or tightly tailored partner solutions. The rationale is clear: differentiation. In a competitive banking landscape, technology is increasingly seen as a source of competitive advantage rather than a shared utility.
But customisation at scale brings complexity. As organisations embed bespoke systems into legacy infrastructure, they risk creating fragmented architectures that are difficult to maintain, regulate, or evolve. The result can be innovation in the short term, but rigidity in the long term.
Modernisation as a Competitive Imperative
Banking institutions are also grappling with changing customer expectations. Digital-first experiences are no longer a differentiator; they are a baseline requirement. According to Capgemini’s 2025 World Banking Report, many banking customers remain either dissatisfied or indifferent to their current card and digital banking experiences.
Customer dissatisfaction is increasingly reflected in behaviour. Globally, account switching is on the rise, with digital experience emerging as a key driver. A growing share of customers now cite better online and mobile banking as a primary reason for changing providers. The trend signals a broader shift: customers are prioritising convenience and usability over traditional brand loyalty when choosing where to bank.
Established banks are therefore under pressure from digitally native challengers that offer faster onboarding, simpler interfaces, and more responsive services. In this context, modernisation is no longer optional. It is becoming a structural requirement for competitiveness. Doing nothing is no longer an option.
The Limits of Hyper-Personalisation
Against this backdrop, hyper-personalisation has emerged as a dominant strategy. Many banking CIOs see it as the logical endpoint of AI-driven transformation: deeply tailored services, dynamic pricing, and highly individualised customer journeys.
However, there is a growing debate about whether this level of customisation is sustainable as a long-term architectural strategy. While tailored solutions may deliver short-term gains, they often come at the cost of interoperability and adaptability.
Financial services operate in a highly regulated and rapidly evolving environment. Privacy requirements such as GDPR, alongside shifting regulatory expectations and geopolitical uncertainty, demand systems that can adapt quickly and consistently across the enterprise. Highly bespoke architectures can make this difficult, particularly when changes need to be implemented across multiple disconnected systems.
There is also the question of adoptability. A strategy built on tightly coupled custom systems can struggle when organisations attempt to expand successful pilots across different business units or geographies. What works in a controlled environment does not always translate cleanly into enterprise-wide deployment.
Agentic AI and the Acceleration Challenge
The emergence of agentic AI is intensifying these tensions. Unlike earlier generations of AI, agentic systems are designed to perform multi-step tasks autonomously, interacting with different systems and data sources to achieve defined outcomes.
According to recent industry research conducted with IBM, 17% of CIOs are planning to launch agentic AI initiatives in 2026, while a further 42% are already running pilot programmes. This indicates a rapid shift from theoretical exploration to practical implementation.
However, agentic AI increases architectural demands significantly. These systems require consistent access to data, clear governance frameworks, and reliable integration across legacy and modern platforms. Without a coherent underlying architecture, organisations risk introducing new layers of complexity rather than reducing them.
The challenge is therefore not just deploying AI, but ensuring the underlying systems can support it at scale.
From Customisation to Interoperability
As banks reconsider their AI strategies, a broader shift in thinking is emerging. Rather than prioritising maximum customisation, some institutions are focusing on interoperability and modularity as the foundation for long-term resilience.
This approach emphasises the ability to integrate, replace, or upgrade components without disrupting core operations. It is particularly relevant in environments where technology lifecycles are shortening, and regulatory requirements are continuously evolving.
In practice, this means moving away from monolithic architectures and towards more composable, service-based systems. Microservices and open standards are increasingly being used to enable flexibility, allowing banks to adopt new technologies without requiring wholesale system replacement.
The objective is not to eliminate customisation, but to ensure it does not become a constraint on future innovation.
The Concept of Coreless Architecture
Within this broader shift, the idea of “coreless” banking architecture has gained traction. Rather than relying on a single central core system that governs all functionality, coreless approaches distribute capabilities across modular components connected through standardised interfaces.
This structure allows institutions to add, remove, or update services more efficiently, without disrupting the entire system. It also supports faster experimentation, enabling banks to test new technologies and scale successful use cases more rapidly.
Importantly, this approach is not about replacing core banking systems entirely, but about reducing dependency on rigid, centralised architectures. By decoupling services, organisations can respond more effectively to regulatory change, market shifts, and emerging technologies such as AI.
The Human Factor in AI Transformation
Technology alone is not sufficient to deliver successful AI transformation. Cultural and organisational readiness play an equally important role.
As AI becomes embedded across banking operations, employees are increasingly required to understand not only how to use these tools, but also how to manage the associated risks. This includes recognising potential errors, understanding governance requirements, and ensuring compliance in day-to-day workflows.
Industry leaders have highlighted the need for a cultural shift in which responsibility for AI risk is distributed across the organisation, rather than confined to specialist teams. In this model, employees at all levels act as participants in maintaining oversight and accountability.
Without this shift, even well-designed systems can fail to deliver value, as human processes struggle to keep pace with technological capability.
Navigating the Next Phase of Banking AI
Looking ahead, banking CIOs face a strategic choice. The pursuit of hyper-personalised, highly customised AI systems may deliver immediate differentiation, but it risks creating long-term constraints in flexibility and scalability.
Alternatively, a focus on interoperability, modularity, and adaptable architecture may provide a more resilient foundation for sustained innovation. In an environment defined by regulatory change, geopolitical uncertainty, and rapid advances in AI capability, adaptability is increasingly becoming the defining success factor.
The risk for institutions is not simply failing to adopt AI but adopting it in ways that limit future options.
Conclusion: Building for Change, Not Just for Today
The evidence suggests that AI success in banking will depend less on the sophistication of individual pilots and more on the strength of the underlying architecture in which they operate. With most enterprise AI initiatives failing to deliver near-term returns, the emphasis is shifting towards structural readiness rather than isolated innovation.
Banks that prioritise flexibility, interoperability, and scalable design are more likely to translate AI investment into sustained value. Those that over-index on bespoke optimisation risk creating systems that are difficult to evolve as conditions change.
In a sector where transformation is constant, the most valuable capability may not be personalisation itself, but the ability to adapt it.
