At the 8th China Fintech Forum, held on September 9 in Beijing under the theme "Technology Empowering - Digital Innovation and Application in the Financial Industry," Gong Xiaojun, Chief Operating Officer and Chief Information Officer of DBS Bank China, delivered a keynote speech on the bank's AI transformation journey.

Gong began by expressing gratitude for the invitation, sharing insights from DBS's AI initiatives. As Singapore's largest bank, formerly the Development Bank of Singapore, DBS embarked on its AI journey in 2014. He noted that the bank initially employed traditional AI concepts, transitioned to machine learning around 2018, and began substantial investments in generative AI (GenAI) in 2023. Since last year, they have been accumulating experience and exploring Agentic AI.

Gong emphasized that AI, particularly generative AI, represents a decisive competitive advantage. He referenced earlier remarks about market-wide AI adoption, pointing out that banks operate within a highly regulated environment. As corporate and retail clients increasingly integrate AI, a bank's readiness to embrace AI becomes critical. Those moving faster could capture significant market share and unlock substantial growth potential.

The bank focuses on three key priorities: scaling AI adoption across the organization, reducing the full-lifecycle costs and workload of AI applications, and achieving exponential growth in economic outcomes.

Gong outlined DBS's internal deployment framework, which has been systematically advanced this year. The enterprise-level application relies on four drivers: use cases, data and technical capabilities, responsible AI security with strict quality and safety measures, and employee care. The latter is crucial, as AI adoption often triggers fears about job displacement, making successful implementation dependent on addressing workforce concerns.

Application scenarios are categorized into horizontal and vertical types. Horizontal applications benefit all employees through DBS-GPT, an internal AI platform with integrated security and risk management. Employees can query the platform, create simple agents, and develop more sophisticated ones for deep reasoning tasks. Vertical applications require specialized AI capabilities focused on specific areas, such as agents for relationship managers and generative AI for customer service staff.

In customer service centers across multiple markets, every employee is supported by an AI agent that performs two main tasks. First, it listens to customer calls, analyzes needs, and identifies potential requirements—for instance, recognizing that a customer calling multiple times about account opening may have investment interests, and preparing this information for the human agent. Second, it conducts quality assurance by assessing call tones and evaluating service quality.

The operations team is also exploring numerous use cases. Gong described a future where customers, especially corporate clients, can input needs in natural language, and the system generates tailored solutions automatically. Currently, DBS-GPT contains operational process manuals and procedures in its knowledge base. In tests, queries about account opening with missing documentation prompt quick, accurate solution proposals with alternative options. Accuracy depends largely on knowledge base quality, given the bank's extensive repository of documents. High-quality knowledge bases enable rapid, effective responses.

At the next level, agents are classified into personal, team, and enterprise types. Personal agents allow employees to perform simple tasks like reading meeting minutes, generating follow-up actions, and drafting summaries, with easy sharing capabilities. Team agents emerge when specific team-level customer data is required, while enterprise agents need more governance. At the top sits operational model transformation, where every department explores organizational and workflow changes. The goal: equip every employee with an AI assistant, handle every task with an agent, and provide AI-powered customer service for every client.

Enterprise agent development differs from traditional system development, which typically involves collaboration between business and technology teams. Now, data teams play a more integral role, as data accuracy and knowledge base quality are critical determinants of agent capability, particularly with unstructured data.

Gong noted examples including enterprise customer agents and automated approval and document memorandum generation. He concluded by highlighting the importance of employees in operational transformation. This transition presents opportunities: redirecting staff toward client-facing roles that generate more value, and creating new positions focused on managing AI models. He has observed many employees thriving in new roles, achieving remarkable results through successful transitions.

He concluded, "That concludes my sharing. Thank you!"