Technology leader Ashish Chatterjee explains why modern applications, trusted data and clear governance must come before large-scale AI adoption.
Banks are being asked to modernise decades-old systems while adopting artificial intelligence at speed. The most visible discussion is about generative AI, intelligent assistants and autonomous agents. The harder question is whether the technology underneath is ready to support them.
Ashish Chatterjee, a technology leader with more than eighteen years of experience across banking, financial services, energy, healthcare and telecommunications, believes many organisations are beginning the conversation in the wrong place.
“AI transformation does not start with AI,” he says. “It starts with reliable applications, trusted data, secure APIs, scalable cloud platforms and clear governance. If those foundations are weak, adding AI only makes the weaknesses harder to control.”
His view comes from leading application modernisation, data migration, Open Banking, cloud transformation and technology delivery programmes in regulated environments. Across these programmes, the common challenge has not been access to new technology. It has been connecting new capabilities to complex systems without disrupting services that customers and institutions depend on every day.
Modernisation is only the beginning
Many financial institutions still operate a mixture of core banking platforms, mainframes, monolithic applications, fragmented data systems and batch processes. These systems may continue to perform their original functions, but they can become difficult to adapt when customer expectations, regulations and digital products change quickly.
Modernisation programmes have traditionally focused on moving applications to the cloud, creating APIs, introducing microservices and improving deployment processes. That work remains essential, but Chatterjee sees it as the foundation of a longer journey.
The earlier transformation agenda was largely about digitising, modernising, migrating and automating. The next stage is about using connected and trusted information to understand, predict, recommend and optimise. AI becomes useful at that stage, but only when the underlying technology estate is accessible, secure and properly governed.
Chatterjee began his career in Java engineering, enterprise integration and application architecture before moving into programme and transformation leadership. That technical background continues to shape the way he approaches AI. He treats it as one component of an operating system rather than a separate product that can be attached to any environment.
Open Banking created the connectivity layer
Open Banking is a useful example of how the banking technology stack has changed. Secure APIs made it possible for customers and authorised providers to access banking information and initiate payments through connected digital services.
In a recent UK banking engagement, Chatterjee led the delivery of Payment Initiation Services and Account Information Services through secure API-based integration. The programme operated within requirements including PSD2, Financial Conduct Authority guidance, UK Open Banking standards and data protection obligations. Project records report 99.9 percent service availability and an improvement of approximately 20 percent in time to market.
The longer-term importance of Open Banking goes beyond payments and account aggregation. It created a controlled connectivity layer through which financial data and services could be made available in real time.
That connected environment can support stronger fraud controls, more relevant customer experiences and better operational decisions. AI and machine learning may help deliver those outcomes, but only where data access, consent, security and quality are managed properly.
Data migration should create trusted data
Chatterjee is currently leading a large mortgage data migration connected with the integration of Nationwide Building Society and Virgin Money. The work covers data profiling, transformation, validation, reconciliation, governance and cutover preparation in a closely regulated environment.
Large migrations are often described as exercises in moving information from one system to another. Chatterjee argues that a successful migration should also improve the quality, ownership, traceability and usefulness of that information.
“The opportunity starts after the data has moved,” he says. “Trusted data can support better customer insight, fraud analysis, document processing, operational forecasting and regulatory reporting. If the data is incomplete or poorly governed, the intelligence built on it will inherit the same problems.”
This changes how migration programmes are designed. Data quality, lineage, access rights, reconciliation and intended use must be considered from the beginning. These controls are not additional documentation. They determine whether the information can later be used safely for analytics, automation and AI-assisted decisions.
Cloud provides scale, not intelligence
Cloud platforms have changed how banks build and operate technology. They provide flexible infrastructure, support automated deployment and make it easier to scale applications. Moving to the cloud, however, does not automatically make an organisation intelligent.
Chatterjee describes the relationship in practical terms. Cloud provides scale. Data provides the raw material. APIs provide connectivity. Applications execute services. AI can add intelligence. Governance creates trust.
All of these elements must work together. A language model connected to a chatbot may create a visible AI feature, but it does not solve weak data quality, fragmented applications or unclear accountability.
This is particularly important in financial services. An AI-generated answer can be technically plausible and still create risk if it uses information without the correct permission, cannot explain a recommendation or operates without effective human review.
Governance must be designed into the system
During his work on an AI and Agile transformation for a United States healthcare technology platform, Chatterjee helped establish an approach covering risk, compliance, ethics and explainability. The platform serves more than two million consumers across over twenty health plans.
Healthcare and banking share several concerns when deploying AI. Both manage sensitive personal information, regulated processes and decisions where errors can have serious consequences.
Chatterjee’s approach is to begin with the business problem, assess data readiness and define what a successful outcome would look like. Model selection comes later.
Responsible deployment requires privacy controls, cybersecurity, model risk management, explainability, bias monitoring, human oversight, auditability and clear ownership. These controls cannot be added shortly before launch. They have to be part of the design.
Intelligent engineering may deliver the earliest value
Some of the most immediate benefits of AI may come from the way technology itself is maintained and modernised. AI tools can assist with legacy code analysis, documentation, dependency discovery, test generation and defect investigation. They can also help teams understand large systems before beginning a migration or redesign.
Chatterjee describes this as intelligent engineering. The purpose is not to remove engineers or architects from the process. It is to reduce repetitive cognitive work and allow experienced people to spend more time on decisions that require judgement.
“The engineer remains accountable, the architect remains accountable and the delivery leader remains accountable,” he says. “AI can assist the work, but responsibility cannot be handed to the tool.”
For banks with extensive legacy estates, this may create value sooner than highly visible customer-facing AI. Better documentation, faster testing and clearer system understanding can reduce modernisation risk while improving the foundations needed for more advanced use cases.
The shift from digital banking to intelligent banking
The longer-term vision is a banking environment where applications, data, APIs, cloud infrastructure and AI operate as one governed system.
A customer could ask a question through a digital channel. The platform could understand the context, retrieve trusted information, apply the appropriate policy, invoke a secure banking service and preserve a complete record for compliance and review. That is more than a chatbot. It is an intelligent operating environment.
The institutions most likely to reach that point may not be those that announce the largest number of AI pilots. They may be the ones doing the less visible work of modernising applications, improving data quality, strengthening governance and assigning clear accountability.
For Chatterjee, the strategic question is no longer whether banks should adopt AI. It is how they can connect AI to applications, data, cloud platforms and business processes in a way that produces measurable and trusted value.
The future of banking may be intelligent, but reaching it will depend on the quality of the digital foundations underneath.
