KUALA LUMPUR: YTL-Sea Digital Bank is expected to be launched in the coming weeks, with a public rollout slated for January 2025.
YTL-Sea Digital Bank chief product officer Foong Chee Mun said the technology infrastructure was completed as early as January this year, but regulatory approval from Bank Negara Malaysia is required for the official launch.
“We are probably going to get approved in the next couple of weeks. Then we will have an internal launch. The tech stuff was done in January this year. My tech guy was like, ‘What are we doing here?’ But we can’t launch until Bank Negara says yes,” he told SunBiz on the sidelines of the SCxSC FinTech Summit recently.
Foong hinted at advanced artificial intelligence (AI) features, including Large Language Models (LLM), to be incorporated into YTL-Sea Digital Bank to streamline risk monitoring and improve customer support.
“Because of this big change in the nature of interactions with computers in finance, there are two types of interactions where we can expect to see innovation from now on.
“The first type is that humans talk into an LLM and in turn, the LLM will instruct a bunch of agents to do whatever the human wants to do. Second, because of the LLM, the large language model, is able to understand human language,” he explained.
This means it will also be able to ingest artifacts, like articles, news articles, and books that were generated for human consumption, and it will be able to understand and create agents out of them as well,” Foong said. “So these are examples of how it is being used, potentially in financial institutions. And this is closely inspired by a use case that our chief risk officer told me.”
Essentially, Foong said, this is a risk-based system where it ingests news articles from around the world, 24-7, and the output is a risk rating and signals.
“And if potentially ... to a bank, it will signal the right personnel. So you can see a lot of people saying that AI will replace human tasks, but in this particular case it doesn’t replace human tasks. In this particular case, it does a superhuman task where a human couldn’t possibly ingest all these news articles 24-7 and remain vigilant all the time,” he added.
Furthermore, Foong said, the second use case we have in a bank is that instead of having customers click through a few screens, banking agents will be able to take in chat, comments, natural language comments, pictures and even voice to instruct the computer system to do something, such as payments, deposits or onboarding.
“These are examples of potential commands that we can use to command the computer system to do something, like pay 26 bucks to someone for lunch. It will also be able to understand short forms, and it will be able to understand multiple languages. You can instruct the LLM to do math as well,” he shared.
Foong said customers can also combine two actions together, such as sending the same amount of money to two people. “You can combine different actions together, potentially paying someone a certain amount and letting them know when the transfer is complete, or sending them messages or WhatsApp notifications.”
The other thing financial institutions can do with a large language model – or in this particular case, a visual language model – is to be able to take in screenshots, according to Foong.
“Most of us sometimes get a WhatsApp instruction from a friend. So, you would be able to upload it to the app, and it will generate the payment instruction from the images. Or invoices, whatever the source – since most payment instructions come from invoices – the large language model will be able to generate them,” he added.
Foong said these represent the new generation of interactions with computer systems that is not generally seen in the industry.
“And I hear certain rumours that these features might be launching at a certain bank. I cannot tell you which bank it is, but you can guess,“ he quipped
