Sumitomo Mitsui Banking Corp. (SMBC) has launched a Singapore-based agentic AI startup to accelerate the development of enterprise automation tools and draw talent that traditional banking structures often struggle to attract.
Announced in June 2025, the new venture is set up to function more autonomously, allowing for faster experimentation, controlled testing of automation solutions, and freedom from the slow internal approval processes typical of large financial institutions.
“We want the best of both worlds: an independent entity backed by SMBC that can attract top talent, experiment quickly, and operate with agility,” said Mayoran Rajendra, managing director and head of SMBC’s AI transformation office, in an interview with Asian Banking & Finance.
Agentic AI differs from generative AI in that it can make decisions, execute tasks, and manage multi-step objectives with limited supervision. It can autonomously tackle complex goals while reducing the risks associated with open-ended language models.
According to Rajendra, this approach helps reduce uncertainty in operations by breaking work into smaller, well-defined tasks. “You’re not relying heavily on a large language model, but on many small judgments needed in repetitive workflows,” he said.
The bank is initially applying its agentic AI to corporate onboarding and know-your-customer (KYC) processes—areas with heavy documentation requirements. Multinational clients often require compliance teams to review 400 to 500 pages of documents to extract jurisdiction-specific data points.
Agentic AI could take on much of this effort by pulling information, verifying accuracy, and cross-checking data across authorised sources. Rajendra said this could cut corporate account opening times from five days to two, while loan processing—sometimes as long as seven months—could be reduced to as little as five days, depending on regulations.
Despite the automation, human oversight remains central. SMBC designs its workflows so staff can always trace exactly where each extracted data point came from. “We automate repetitive work, but it’s always auditable, traceable, observable,” Rajendra noted, adding that 100% accuracy can never be assumed.
To safeguard sensitive information, all AI-processed data stays strictly within SMBC’s internal systems and is never used to train external models. “We ensure no information is fed into an LLM for training. It stays within SMBC,” Rajendra said.
Beyond its AI push, the bank is offering corporate clients free CFO dashboards with real-time operational analytics and continues to invest in startups across India and Southeast Asia through the SMBC Asia Rising Fund.
Rajendra summed up the mission simply: “We’re bringing agentic AI tools and capabilities to help automate our customers’ operations.”
Click here for more on Banking
