DBS, OCBC and UOB push AI from pilots into daily banking
Source: Fintech News SG
DBS, OCBC and UOB have pushed hundreds of AI deployments into production, from agentic credit memos to wealth avatars, just as MAS starts enforcing new AI risk rules.
Singapore's three biggest homegrown lenders have quit treating artificial intelligence as a side project. DBS, OCBC and UOB each claim hundreds of live deployments — DBS past 430, its two rivals beyond 300 apiece. What counts, though, is what reached bankers and customers this year.
The tallies aren't directly comparable: UOB folds analytics into its total, so headline figures flatter. DBS put money behind the claim — its 2025 annual report ties roughly S$1 billion of economic value to analytics, AI and machine-learning work spanning more than 2,000 models. All this lands as the Monetary Authority of Singapore tightens the rulebook.
The sharpest shift is toward agentic systems — software that chains tasks together instead of answering a single prompt. DBS handed roughly 1,500 relationship and credit risk managers a tool in August that steps through 70-plus tasks to draft corporate credit memos, scaled from a 150-person pilot; humans still make the lending calls. Joy and digibot assistants gained generative agentic features, now reaching over 10 million users across Singapore, Hong Kong and Taiwan. OCBC's wealth push includes WoW, live since 1 July, where avatars Wendy and Wayne field portfolio questions from in-house research for staff and invited Premier Private Client customers. HELIOS, its agentic platform, targets a 15-business-day median for opening private banking accounts against an industry norm near six weeks. UOB's AURA covers anti-money-laundering, know-your-customer, credit and collections work, while Project Magnet lifted genuine-alert detection by 148% and exception rates by 130%.
The regulator is moving in step. MAS issued AI risk management guidelines on 7 October, effective 7 October 2027, with parts phased in through October 2028, and will consult in 2027 on whether agentic AI needs extra guidance. Its MindForge project closed a second phase in March with an AI Risk Management Toolkit produced with 24 financial institutions; a SAFR framework for agentic finance followed in July, and an AI cyber and technology risk task force came from the Association of Banks in Singapore. MAS also pledged a S$220 million, three-year commitment under FSTI 4.0.
Why it matters for Singapore: Singapore wants to be both a fast adopter and a strict referee, and these bank rollouts are where that balance gets tested. With MAS setting a 2027 compliance clock and FSTI 4.0 funding wider adoption, the financial sector is a live laboratory for governed AI at scale. Proven returns under real scrutiny would give other industries and regional markets a template to copy.
