DBS Group Holdings Ltd. is using more artificial intelligence than ever while reducing the cost per interaction, a trend its CEO describes as the “token paradox.”
Tan Su Shan, speaking in an interview, said the Singapore-based bank’s spending per AI token has fallen even as usage increases. “The more you use, the cheaper it gets,” she stated. “Our people are getting a lot smarter on how to optimise.”
The savings stem from several changes. DBS implemented a memory cache to avoid reprocessing identical queries. It also directs simpler questions to smaller language models instead of relying on larger, costlier ones.
“You don’t need to burn an ocean to get a straightforward answer,” Tan added. The bank’s total technology spending remains at around 10% of revenue.
Related: UKM and BlackBerry launch embedded software course
Other financial institutions share similar concerns. JPMorgan Chase & Co. recently advised employees to avoid using high-cost models for routine tasks like summarizing reports. While first-half token spending was minimal, executives warned costs could rise significantly in the coming months.
DBS manages expenses through an open architecture. The bank evaluates multiple providers in proof-of-concept trials to “see who’s good at what” and is figuring out what codes and models should be done in-house. Since she took the top job more than a year ago, Tan has pushed ahead on AI, establishing a team and ensuring the resiliency and security of the technology.
This strategy impacts workflows across the bank. Tan noted that AI is transforming how people work, from traders to support staff—even the tea lady at the firm. “We’re changing the way people look at tech,” she said.
Efficiency efforts coincide with record profits. DBS raised its full-year guidance after a second-quarter profit surpassed analyst expectations, powered by a surge in wealth-led fee income. Shares reached record highs, matching gains at local rivals Oversea-Chinese Banking Corp. and United Overseas Bank Ltd.
Related: Toyota Boosts Profit Outlook with Big Buyback
Taiwan emerged as the most promising market for wealth management over the next two to three years. Tan highlighted the island’s expanding AI sector and efforts to diversify investments offshore. India also stands out, with foreign currency non-resident deposits attracting stronger interest after the central bank offered full hedging-cost support on the inflows for the first time in more than a decade.
Geopolitical risks remain a concern. Tan said the on-off war in the Middle East, together with managing the fast pace around AI and thwarting cyber-attacks, will be her biggest challenges over the next year.
“Investors and customers are looking at us to be that safe harbour,” she said.
