The Singapore bank started its path to use AI in 2018 by focusing on four main areas: the creation of analytics capabilities, data culture and curriculum, data upskilling, and data enablement.
The company’s goal was to create a data culture that pushed all employees to always consider how data and AI could assist them in their work as well as the relevant use cases and talent, such as machine learning engineers. It entailed offering a training course that instructed personnel on when and how to use data and when not to.
The bank is working on establishing its infrastructure to encompass AI adoption to its data platform, data management structure and data governance. It established a framework that all of its data use cases must be evaluated. PURE is for Purposeful, Unsurprising, Respectful, and Explainable. According to DBS, these four principles are fundamental in directing the bank's responsible use of data.
With the help of its data platform, ADA, the bank is better able to ensure data governance, quality, discoverability, and security.
It has been discovered that 95% of the data turned out to be useful and crucial for DBS’ AI-based operations. For a fact, the platform consists of more than 5.3 petabytes of data, with 32,000 datasets including videos and structured data.
However, Gupta revealed that getting to this point “proved a mammoth task.” He explained that organizing this data and making it discoverable in particular needed a lot of effort, primarily manual and human expertise. It took countless hours to identify the metadata, and there are very few tools available to automate this process.
Taking into account the data that spreads across different systems, "a lot of heavy lifting was needed to bring datasets onto a single platform and make these discoverable. Employees must be able to extract the data they need and the bank had to ensure this was done securely,” he said.
Today, DBS govern more than 300 AI and machine learning program, yielding a revenue uplift of SG$150 million ($112.53 million). Additionally, the company saved SG$30 million ($22.51 million) in 2022 for their efforts in risk mitigation, for example from bettering their credit monitoring. Gupta notes that these AI use cases involve a range of functions, like human resources, legal, and fraud detection.
Moreover, he highlighted the need to ensure that the use of AI applications maintains the PURE principles and Singapore's FEAT principles – serving as a framework for the sector’s AI use. It will also be necessary to evaluate other recognized hazards, such as hallucinations and copyright violations, he said.
He added that the company needs to work on its technical elements, including building mechanisms to monitor AI use and gather feedback in order to analyze the ongoing operation and identify areas of improvement. This will consequently ensure that the organization will learn from its AI application, and will be able to make necessary changes whenever needed.
In regards to DBS’ AI usage for predicting outages, such as the disruptions it witnessed last year, he confirmed that the bank is in fact working in identifying how it can do better, including tapping data analysis. He said there is potential to apply AI, for instance, in operations to spot anomalies and choose the best course of action. He also noted that a variety of circumstances might lead to surges in demand.
DBS is the biggest bank in Singapore and currently employs 1,000 data scientists, engineers, and engineers. It runs 600 AI and machine learning algorithms, in order to facilitate its five million customers spread across the regions of China, Indonesia, and India.
With an increase in revenue to SG$350 million ($262.56 million), the bank's AI projects are targeting to produce additional economic value and cost avoidance benefits this year. In the following three years, it aims to reach SG$ 1 billion ($750.17 million).
