DBS Bank, Singapore’s largest bank, has faced numerous challenges in its journey to adopt artificial intelligence (AI), realizing that success goes beyond simply developing training models. One major hurdle has been data, according to Sameer Gupta, DBS’ chief analytics officer. In 2018, the bank set out to leverage AI across four key areas: analytics capabilities, data culture and curriculum, data upskilling, and data enablement.
The bank aimed to make AI accessible to all employees and deliver economic value from it. It also focused on developing the right use cases and talent, such as machine learning engineers, and fostering a data culture that encouraged employees to think about how data and AI could enhance their work. DBS worked on establishing the necessary infrastructure for AI adoption, including the data platform, data management structure, and data governance.
The bank implemented a framework called PURE, which stands for purposeful, unsurprising, respectful, and explainable. This framework guides DBS in using data responsibly. The data platform, ADA, serves as a central source for data governance, quality, discoverability, and security. More than 95% of data deemed useful for DBS’ AI-powered operations is now discoverable on the platform, which holds over 5.3 petabytes of data and 32,000 datasets.
However, organizing and making the data discoverable proved to be a challenging task that required significant manual work. The bank used multiple applications, each holding data necessary for its AI initiatives. Bringing datasets onto a single platform and ensuring secure data extraction was a complex process that involved a lot of effort. Despite these challenges, DBS currently runs over 300 AI and machine learning projects, resulting in revenue uplift and cost savings.
DBS’ AI initiatives are expected to generate further economic value and cost avoidance benefits, with the bank aiming to reach SG$1 billion ($750.17 million) in the next three years. The bank currently has around 1,000 data engineers, data scientists, and data engineers. DBS is also exploring the use of generative AI through several pilot projects, but it is still in the early stages.
The bank aims to ensure that the use of AI applications complies with its PURE principles and Singapore’s FEAT principles, which guide the sector’s use of AI. DBS currently operates 600 AI and machine learning algorithms that power interactions with its five million customers across the region. Gupta emphasized that the number of AI models used is not as important as achieving optimal efficiency and accuracy.
Gupta highlighted the misconception that the AI model is everything, stressing the importance of working through all technical elements and continuously gathering feedback to identify areas of improvement. He emphasized the need for perseverance and acknowledged that there is no “magic bullet” for AI adoption. When asked about using AI to anticipate outages, Gupta mentioned that the bank is working to improve its capabilities, including leveraging data analytics to detect anomalies and determine the next course of action.
DBS is currently conducting a full review of its technology resiliency, led by a special committee comprising four board members. External experts have been engaged to assist with the review, and more details will be provided once it is completed.