Digital lending is a major driver of Southeast Asia’s US$30 billion ($40.1 billion) revenue from digital financial services this year. This is due to high lending rates and consumer demand as underbanked consumers and small businesses participate in the digital economy, according to the e-Conomy SEA 2023 report by Google, Temasek and Bain & Company.
With micro, small and medium enterprises (MSMEs) and small businesses being key driving forces in the region’s economies, it presents banks an opportunity to widen their customer base. Banks, however, will need to use non-traditional data to identify better creditworthy borrowers who may not meet traditional credit requirements.
“Most banks in Southeast Asia have similar lending criteria and credit scoring processes [that tend to disfavour customers without formal credit history, which is why] there is a large proportion of unbanked and underserved individuals and businesses here. To tap on that segment, banks will need to look at alternative data points that can show patterns of good credit behaviour,” Guy Sheppard, chief operations officer of financial services at Aboitiz Data Innovation (ADI), tells DigitalEdge. ADI is the data science and artificial intelligence arm of the Philippines conglomerate Aboitiz Group.
For example, the Union Bank of the Philippines (UnionBank) uses ADI’s alternative credit scoring and risk model solution to facilitate more efficient loan provision for unbanked individuals and MSMEs. Powered by machine learning, the solution considers non-traditional data from publicly available sources, government data, and partners to assess creditworthiness more inclusively and accurately.
“This may include looking at the applicant’s office location (such as its proximity to an embassy) and the strength of the internet bandwidth there to correlate with credit risk because in the Philippines, scoring based on addresses is more effective than simply running a credit check based on the applicant’s names,” says Sheppard.
He adds that lenders that use ADI’s alternative credit scoring and risk model solution have seen delinquency performance of bookings decrease by 11%, approval rates double, and booking amounts increase by seven times.
See also: Alibaba anoints new chief in revamp of stalling commerce arm
Since biased decisions are a concern when using AI, banks must have guidelines ensuring the responsible use of AI and explainable AI. Explainability is the ability to express why an AI system reached a particular decision, recommendation, or prediction. This can be complex when AI engines continually ingest data from multiple sources to become “smarter”.
ADI addresses this by having “solid data architecture with strict data governance policies” and a team focusing on model validation to ensure the accuracy of the AI models (also known as model drift).
“We have a handbook that details how AI models should be built. It touches on coding best practices and how to decide what data to use for the model, among other things. We also have a system built in-house that tests models for potential bias (be it for gender, age, ethnicity and more) to self-police,” adds Sheppard.
See also: Break up Google? What’s at stake in antitrust action
Combating cyber criminals
Financial crime is another major concern for Southeast Asian banks, especially as cybercriminals become increasingly sophisticated.
The Bangladesh Bank heist showed that the cyber attackers understood how the Swift messages work and where the gaps [in terms of defences and security controls] in the financial system were. So, relying only on rules-based monitoring systems to identify suspicious transactions is no longer enough.
Guy Sheppard, chief operations officer of financial services, Aboitiz Data Innovation (ADI)
Alternative data and AI can help by enabling better detection of mule accounts used to transfer money obtained illegally. “The key challenge around mule account detection is that signals indicating a change in account use, legitimacy, and ownership can be very slight and often blend into the noise of daily monitoring. Customer collusion is also rife,” says Sheppard in a recent interview with DigiconAsia.
Developed in collaboration with UnionBank, ADI’s Mules Account Detection solution takes a two-level approach. Firstly, at onboarding, the AI model identifies the probability of a falsely set up mule account and further assesses the suspected account’s behaviour and likelihood of it being a mule.
Secondly, the AI-driven anomaly detection model continuously learns and improves to generate scored alerts of mule activity. The solution encompasses workflow, alert and case management for ease of deployment and offers configurable detection.
To stay ahead of the latest tech trends, click here for DigitalEdge Section
Sheppard adds: “Our Mules Account Detection solution can look at customers’ transactional profiles to find lookalikes and identify suspicious accounts before investigating them. This enables banks to shift from a reactive to a proactive approach to mule account detection. We’re also considering complementing it with other tools such as graph analytics as we believe technology shouldn’t be used in isolation [to combat financial crime].”
Additionally, he emphasised the need for more partnerships between public and private sectors and within the region, as financial crime today knows no boundaries. Bad actors can conceal their criminal activities by exploiting the complexity of the global financial system, the differences between national laws, and the speed at which money can cross borders.
Centralised platform
One way of overcoming this is by having an information-sharing platform. For instance, the Cosmic (Collaborative Sharing of ML/TF Information and Cases) platform enables financial institutions in Singapore to securely share information on customers or transactions where they cross material risk thresholds, enabling them to identify and disrupt illicit networks more effectively.
The platform is co-created by the Monetary Authority of Singapore (MAS) and six banks, namely DBS Bank, Oversea-Chinese Banking Corporation, United Overseas Bank U11 , Standard Chartered Bank, Citibank and HSBC.
MAS describes Cosmic as a centralised platform facilitating seamless collaboration among financial institutions through structured data sharing. The regulatory framework outlines shared information types and circumstances, aiding risk surveillance to detect illicit networks and enabling timely intervention.
By applying AI to data from more cross-industry and cross-border collaboration, Sheppard believes banks can better prevent financial crime and play a pivotal role in addressing societal issues like preventing human trafficking.
