How banks are navigating the AI landscape

For decades, financial services firms have been using machine learning techniques to detect fraud and predict if customers are likely to default on loans. But the recent hype over artificial intelligence (AI) has once again cast the technology into the spotlight.

During a panel discussion at the recent Singapore Fintech Festival, industry leaders explored the future of banking in the AI era and addressed key challenges such as talent acquisition, cost management, data security and governance.

Dwaipayan Sadhu, CEO of Trust Bank, kicked off the discussion by highlighting the dual nature of the challenges around AI. While technical hurdles like talent shortages and data security are significant, he noted that “business and cultural challenges are much bigger”.

“Everyone is an AI expert, and hence you have called me here, but I know nothing about AI,” Sadhu quipped, noting the pervasive yet superficial understanding of AI. He advocated for a “value prism” framework, prioritising use cases based on value and feasibility.

“We are very diligent and disciplined in first ascertaining the value, and then we talk about feasibility,” Sadhu said, adding that this paves the way for better allocation of resources and a clearer path to production.

Geraldine Wong, chief data officer at GXS Bank, pointed to the difficulties faced by traditional banks in accessing and utilising their vast data stores. They also struggle to determine if the data is good enough to be used to fine-tune and train AI models.

“The challenges that we face today with generative AI [GenAI] are no different from 10 years ago when we were looking at traditional AI and data science,” she said. “The only difference is that now we have easy access to AI, and everyone’s been given that shiny toy, but is anyone governing its use by people in the organisation?"

Against this backdrop, Wong said it was important for organisations to establish clear guidelines for data access and privacy, as well as evaluate the quality of AI-generated outputs.

“Previously, it was all very structured and quantitative in nature. Today, it’s qualitative with text, images, and voice, so how do we know if the outputs are accurate? Do we have a consistent set of standards to ensure the quality of outputs?”

Kevin Lam, group managing director and CEO of Hong Leong Bank, focused on the commercialisation aspect of AI implementations. He noted that while many AI applications are technically feasible, their scalability and commercial viability remain a challenge, particularly in markets like Southeast Asia.

“The biggest challenge is really about finding that right prioritisation and knowing where to put your resources, which are very thin, and make sure that you get good payback from all these investments in AI,” Lam added.

Lam cited the successful implementation of voice bots for debt collection at Hong Leong Bank, illustrating the potential for AI to drive cost efficiencies. He also pointed to other benefits such as revenue generation and improving the productivity of relationship managers, suggesting that AI can augment human capabilities rather than replace them.

Addressing concerns about the impact of AI on customers, Sadhu noted that AI is still a tool and that banks should be responsible for its use, adhering to principles of fairness, ethics, accountability and transparency.

Wong shared GXS Bank’s experience with testing its chatbot before public release, stressing the need for humans in the loop and iterative improvements based on customer feedback. Lam advocated for transparency with customers, acknowledging that AI systems are not yet perfect and that banks must ultimately be accountable for decisions made by AI systems.

For organisations embarking on their AI journey, Sadhu advised them to start with a clear business outcome in mind, thinking big, starting small, scaling fast and establishing robust governance frameworks from the outset.

Wong encouraged newcomers to explore existing tools before building everything from scratch while Lam advised organisations against locking into specific platforms or vendors too early, given the fast pace of AI developments.

The panellists also discussed future use cases. Sadhu pointed to hyper-personalisation, leveraging data from the broader ecosystem to gain a holistic view of the customer and deliver tailored experiences.

Sadhu also highlighted the potential of AI in combating fraud and scams, shifting the focus from protecting banks to providing customers with tools and nudges so they are less likely to fall prey to scams. Wong added that GenAI can also be used to analyse unstructured data like documents and images in the fight against fraud.

On where financial services firms are in the AI hype cycle, Lam positioned his bank in the middle ground, acknowledging the benefits of AI while also emphasising the need for realistic expectations and a focus on demonstrable commercial value. “We think AI is definitely real,” he said. “If you can identify where the big revenues are, it will be a big uplift.”