DBS is no stranger to artificial intelligence (AI). The bank is leveraging the technology to enhance both internal processes and the customer experience, such as detecting fraudulent transactions, pre-screening candidates during its mass recruitment and offering features like DBS NAV planner, a digital advisory tool to help consumers with financial planning.
“AI and machine learning are becoming a core competence. If we look at Formula One, every decision [the driver and his team make] during a race is enabled by real-time data instead of historical data. I believe the same thing will happen in the banking industry, [which is why] we want to embed AI in every part of the bank,” says Sameer Gupta, DBS’s chief analytics officer, in an interview with DigitalEdge.
He also shares that there are currently more than 300 AI and machine learning use cases across the bank. This resulted in a revenue uplift of $150 million in 2022 (double that in 2021), on top of an additional $30 million from cost avoidance due to enhanced credit monitoring, fraud prevention and productivity gains.
Supporting DBS’s pervasive AI
Having the right technology is one of the key pillars for DBS’s success with AI. “When we started our data-first programme in 2018, we were very clear that data would be one of our defining competitive advantages. But achieving that means we need to be able to leverage AI and machine learning at scale. This is why we focused on building technology platforms that can support that,” says Gupta.
Central to DBS’s data-first programme is ADA (Advancing DBS with AI), an internal self-service data platform. It brings the bank’s data together to provide a single source of truth that ensures data discoverability, quality, governance and security.
See also: Testing QA New Section BDC Feature Winner 1
The bank has also developed an AI protocol platform called ALAN and integrated it with ADA to enable AI to be used across the entire organisation. By doing so, DBS’s data scientists can shorten the cycle time required to build compliant AI models, and accelerate deployment of the models to improve its operations and decision-making.
“In order to scale or industrialise AI, we had to think about the whole data/machine learning pipeline to reduce the time and effort needed to deploy AI. Our first AI model took almost 15 months to develop. [ALAN empowered us to] bring that down to two months, and we aim to further shorten it to two to three weeks,” says Gupta.
He continues: “With ALAN, our data scientists can quickly deploy reusable assets that are commonly used, such as natural language programming, instead of writing the code [for them from scratch]. This helps reduce the time and effort needed to use AI across the bank.
See also: Unpublished article shouldnt be accessible testing
“For example, if I’m looking to solve a customer attrition problem in mortgage [in a certain bank branch], I can use ALAN to find past use cases on attrition [in other branches] and how it was solved. The platform can show the data types, AI/machine learning techniques and the final AI model used in those cases, as well as the accuracy of the final model. So, this notion of reusability and sharing of best practices means we don’t have to start from a white sheet of paper [to solve most problems].”
To address the ethical concerns around AI, DBS adheres to the Fairness, Ethics, Accountability and Transparency (FEAT) principles by the Monetary Authority of Singapore (MAS). The FEAT principles provide firms with guidance on offering financial products and services on the responsible use of AI and data analytics.
DBS does so through its PURE framework, which stands for purpose, unsurprising, respectful and explainable. “To ensure data is used responsibly, all data [analytics and AI] use cases need to go through the PURE framework. Users need to understand the purpose of the data [they intend to use], ensure the use of that data is unsurprising, and communicate the use of that data to end customers and users respectfully. We’ve created a PURE committee of people across functions to deliberate on some of the cases which might not be clear.”
The people aspect of AI
AI’s role in DBS is to elevate staff productivity and effectiveness instead of replacing them. Having the right talent and equipping employees with data skills are, therefore, key to becoming an AI-fuelled bank.
Recognising this, DBS has been expanding its data science team and upskilling employees to have the right capabilities. Gupta says: “Today, we have nearly 200 data scientists and more than 1,000 data professionals (including data analysts and engineers) in the bank. Most of our hires [for the team] are from tech firms, which shows that we’ve become an attractive company to work for.”
In terms of upskilling, the bank offers not only technical training such as coding but also digital training programmes that teach them how to think digitally or in a more data-driven way. Gupta explains that for the latter, employees are taught the “softer skills” of analytics/AI, such as the type of problems those technologies can help solve, how to ask the right questions to get more accurate answers and how to use data responsibly.
To stay ahead of the latest tech trends, click here for DigitalEdge Section
“It’s not enough to just have data scientists and analytics capabilities. We needed to think from a culture standpoint — starting from the top and cascading down, we’ve to ensure everyone can [and is comfortable to] use data to drive decisions in their work, which is why the upskilling programmes are important,” he says.
He adds that DBS leverages gamification where possible to make technical training on analytics and AI less intimidating for non-technical employees. The DBS x AWS DeepRacer League — launched in collaboration with Amazon Web Services in 2020 — is one example.
Under the programme, DBS employees learnt the basics of AI and machine learning by participating in a series of hands-on online tutorials before putting their new knowledge to the test by programming their own autonomous model race car. Those machine learning models were then uploaded onto a virtual racing environment where employees experimented and iteratively fine-tuned their models as they engaged each other in friendly competition.
Given the continuous innovation in AI, DBS will keep experimenting with new AI-related technologies to improve its operations and delight customers. Gupta says: “Generative AI, for example, can help us unlock most of the unstructured data in the bank, so it may be able to improve our existing use cases or help unlock new use cases that we didn’t think of before. We’re currently running small-scale pilots for generative AI to see if it can help improve revenue and employee productivity as well as other internal use cases such as helping employees grow their careers.”
“We’ve been successful [in our AI journey] mainly because we focused on enabling a data-led culture, change management [and deploying the right technology. The goal isn’t] about having one big [AI or analytics] project but more of making AI pervasive throughout the bank. That way, every decision in the bank is enabled and supported by AI. But it’s not about AI making those decisions [but empowering people to make better decisions] so that the business can run better,” he concludes.
