Digital Banking: Opportunity Must Move with Discipline

Bangladesh has entered a new phase in its banking journey. On 24 September, Bangladesh Bank issued Letters of Intent (LoIs) to five proposed digital banks: bKash Digital Bank, Boost Digital Bank, DK Digital Bank, Nova Digital Bank and Kori Digital Bank. However, these are not final banking licences. The institutions must fulfil regulatory, technological and other conditions before they can begin full banking operations. The move creates significant opportunities for Bangladesh. Digital banks can expand access to formal finance, improve customer experience and reduce dependence on physical branches. They can potentially serve small businesses, young customers, informal workers, migrant families and people who remain underserved by traditional banking.

But digital banking should not be viewed simply as banking through a mobile application. It is a technology-driven banking model that depends on reliable data, strong risk management, secure infrastructure and continuous regulatory oversight. More importantly, it gives us an opportunity to rethink traditional banking processes instead of simply putting those processes on a screen.

One of the most important foundations is customer identity. Bangladesh Bank's revised e-KYC guidelines are based on the National Identity Card (NID) and associated biometric information. Financial institutions can use the NID database to verify customer identity and authenticity.

For digital banks, however, identity verification should not end when an account is opened. The correctness, completeness and current status of customer information should remain important throughout the customer relationship. Changes in address, mobile number, occupation, business activity, nominee information and other relevant details should be updated through appropriate processes.

This is particularly important because a digital bank may have millions of customers but very few physical touchpoints. If identity information is wrong or outdated, technology can make the problem faster rather than safer. 

India provides a useful warning. In 2024, the Indian government reported around 29,000 fraud incidents involving the Aadhaar Enabled Payment System (AePS) through its National Cyber Crime Reporting Portal. There have also been reported cases involving the misuse of Aadhaar information and cloned fingerprints to conduct unauthorised transactions through AePS.

The lesson for Bangladesh is not that biometric identification is unsafe. Rather, identity data itself must be treated as a critical financial asset. Strong verification, secure storage, continuous monitoring and mechanisms to detect unusual changes in customer information are essential.

Digital banks will also face fraud risks such as account takeover, SIM replacement, phishing, social engineering, mule accounts and fraudulent transactions. Traditional rule-based monitoring can identify known patterns, but digital banking creates enormous volumes of transactions and customer interactions.

This is where machine learning can make an important difference.

Instead of looking only for predefined rules, machine-learning models can learn from transaction behaviour and identify unusual patterns. Transaction frequency, amount, device changes, location, beneficiary behaviour, login patterns, account age and transaction velocity can be combined to create a dynamic fraud-risk score. For example, a customer who normally makes small local transactions suddenly changes the device, logs in from a different location and transfers a large amount to a newly added beneficiary. No single event may prove fraud. But together, these signals may justify additional authentication or human investigation.

India is already moving in this direction. The Reserve Bank of India has deployed MuleHunter.AI, an AI/ML-based solution designed to help detect mule accounts. Bangladesh's digital banks can learn from such developments while developing models suited to local transaction behaviour and fraud patterns.

Credit should also be handled carefully. Digital banks may use transaction behaviour, business cash flows and other permitted information to support lending to customers with limited formal credit histories. But alternative data should not automatically be treated as proof of creditworthiness. Data quality, customer consent, model bias and affordability remain important.

The bigger opportunity, however, is to think beyond traditional banking concepts.

Digital banking should not mean taking the existing branch model, loan application, account-opening process or service request and converting it into an app. It should ask a different question: If we were designing banking today from the beginning, what would we do differently?

Instead of waiting for customers to report problems, an intelligent banking platform could identify unusual behaviour and intervene early. Instead of repeatedly asking customers for information already available through verified sources, systems could securely reuse verified data. Instead of offering the same service to everyone, analytics could identify different customer needs and provide more relevant services.

The same thinking should apply to compliance. AI and analytics can help identify unusual behaviour, prioritise alerts and support investigators. The objective should not be to generate thousands of alerts. It should be to identify the signals that matter.

Cybersecurity is another critical area. For a digital bank, the technology platform is effectively its main branch. If that platform becomes unavailable or compromised, customer service, payments and trust can be affected. Strong identity and access management, encryption, secure software development, vulnerability testing, continuous monitoring and effective incident response are therefore essential. Third-party risks also require attention because digital banks may depend on cloud providers, telecom operators, fintech companies, payment processors and technology vendors.

Technology alone, however, cannot create a safe digital banking environment. Customers need to understand how to protect their accounts, recognise phishing attempts, avoid sharing OTPs and PINs, and identify fake customer-care numbers or suspicious applications. Employees also need continuous learning in data analytics, cybersecurity, AML, fraud investigation and AI model governance. Ultimately, the success of digital banking will depend on trust. Customers must believe that their money is safe, their identity is protected, their data is used responsibly and their problems will receive fair attention.

Digital banks should therefore not measure success only by account numbers, app downloads or transaction volumes. Sustainable success should also be measured through customer identity quality, fraud prevention, AML effectiveness, technology resilience, service quality and customer trust.

Bangladesh now has an opportunity to build a modern and inclusive banking ecosystem. But the opportunity is larger than simply creating branchless banks. It is an opportunity to rethink how banking is designed, how customer identity is managed, how fraud is detected and how decisions are made. Innovation should not merely automate traditional banking. It should make banking fundamentally smarter, safer and more customer-centric.

_____________________________________

The writer is former banker and AI and analytics professional