Driving personalisation and productivity with AI
Artificial intelligence promises efficiency and better products, but potential risks such as model biases must also be considered
Panellists:
Lawrence Goh, chief operating officer and head of group infrastructure platform services, group technology and operations, UOB
Tancy Tan, chief operating officer, HSBC Singapore
Lim Khiang Tong, group chief operating officer, OCBC
Ben Tan, chief corporate development officer and chief distribution officer, Prudential Singapore
Han Kwee Juan, Singapore country head and acting chief information officer, DBS Bank
BT in your inbox
Start and end each day with the latest news stories and analyses delivered straight to your inbox.
Moderator
Raphael Lim, Correspondent, The Business Times
What are some examples of improved product offerings and/or enhanced customer experiences that AI has delivered so far?
Lawrence Goh: UOB has been using AI and machine learning (ML) across many parts of the bank, with increasing usage over the years. AI has enabled us to create highly personalised interactions with customers at scale.
Our all-in-one app, UOB TMRW, is powered by AI/ML models to serve customers’ banking, financial and lifestyle needs. For instance, we serve real-time personalised insights to customers, such as exclusive offers, deal recommendations or promotions. We also send reminders for credit card payments and suggestions customised to personal spending and saving habits. In the first half of 2023, we served over 44 million insights to around 1.3 million unique customers in Singapore.
We offer robo-advisory services via UOB Asset Management, providing automated and personalised investment advice to customers. AI is also used for fraud detection, anti-money laundering and credit decisioning to keep our customers and the bank safe.
To improve operational efficiency and minimise inconvenience for customers, we deployed AI and geo-analytics to optimise cash replenishment schedules for around 600 ATMs across Singapore, reducing the number of trips required by 25 per cent.
Lim Khiang Tong: OCBC has been at the forefront of adopting AI on a significant scale. Today, more than four million decisions are made with the help of AI in OCBC every day, in processes such as risk management, customer service and sales. This is expected to rise to 10 million in 2025.
An example of how the customer experience has been enhanced is in hyper-personalisation of our mobile banking app. We use AI to trawl through transactions to identify which of our 120 actionable insights may be relevant to the customer, and recommend them to our customers.
We are also one of the first banks in Singapore to deploy generative AI at scale. One of the use cases that we have trialled is the post-sales surveillance review that the bank undertakes. With generative AI to aid in translating speech to text in real time, calls can be automatically summarised and analysed in a much faster way. This enables us to analyse all calls to identify anomalies and opportunities for customer engagement, as opposed to analysing only a fraction of them manually.
Overall, AI and customer experience have become more interlinked than before, given AI’s potential to elevate customer experience in a myriad of ways, including more personalised services and greater convenience and efficiency for customers.
Ben Tan: AI is used in various aspects of insurance operations such as servicing to enhance the overall experience for customers. Prudential Singapore uses an AI talkbot to remind customers to make payments for their premiums so that they remain protected. The talkbot is trained to listen, understand and respond just like a human, allowing us to retain the human aspect in customer interactions while going digital.
We also use AI to automate processes such as verification and digitalisation of information, which helps to reduce turnaround time for customers at the onboarding and servicing stages.
Through this journey, our employees have also benefited from a learning point of view, through greater exposure to AI and by becoming more skilled in using new technology in their work. With the assistance of AI, which brings about increased efficiency, they also have more time to focus their efforts on more complex tasks and strengthen their capabilities in other areas.
Han Kwee Juan: AI has redefined how we interact with retail, wealth and SME (small and medium-sized enterprises) clients. We are now able to do it in a much deeper, proactive and targeted manner, such that clients are equipped with hyper-personalised actionable insights – rather than just broad advice – to improve their financial health.
Today, AI/ML has guided over five million customers towards better investment and financial planning decisions through 45 million nudges every month. Over 100 AI/ML algorithms analyse more than 15,000 customer attributes to ensure that the nudges are relevant, useful and non-intrusive. Those who respond to our AI-powered financial planning nudges have seen an uptick in their financial well-being compared to those who don’t – they save 61 per cent more, invest 200 per cent more, and are four times more insured.
SMEs that bank with us have also benefited from AI and data analytics. DBS is able to provide SMEs with early alerts of credit risks even before problems emerge. With these capabilities, the bank was able to successfully identify over 95 per cent of non-performing SME loans at least three months before the businesses experienced credit stress. Over 80 per cent of identified at-risk borrowers were averted from risk.
Tancy Tan: We have been utilising several technologies including AI, big data, blockchain and cloud to improve operational efficiency and enhance customer experience.
We have been working with AI for many years, with our earliest ML models developed a decade ago. Due to AI’s ability to process and sort large amounts of data quickly, we have been routinely using it to enhance our current banking operations and services. This includes automating repetitive tasks, implementing fraud detection systems, personalising our marketing efforts, and improving customer service. These efforts are aimed at reducing costs, enhancing operational efficiency, and improving customer satisfaction.
On top of using AI to optimise our day-to-day operations, one of the exciting areas in which we have been using AI is creating new customer solutions and enhancing our customers’ banking experience. One example is our AI Markets solution, a client-facing AI chatbot which improves price discovery, client service and distribution using natural language processing.
More recently, there has been a step change due to advancements in generative AI, which has greatly improved the technology’s capabilities and accessibility. As with all new technologies, HSBC’s approach is evolving in lockstep, but we always go back to our core principle of delivering value for our customers and employees responsibly and ethically.
How will new AI developments further transform the industry or allow for different growth opportunities in the coming years?
Ben Tan: Customers today are increasingly looking for hyper-personalisation where they receive the right information at the right time and at the right place. They expect companies to understand them as individuals and make them feel special and cared for.
Advancements in technology such as AI and data analytics will empower insurers to drive hyper-personalisation at scale and create segments of one to better meet each customer’s unique needs and engage them more closely. It will become easier and faster to gather and analyse customer data, improve accuracy of predictions of customer behaviour and recommendations. It can also support the development of new products or services to reach out to new customer segments and help more individuals be better insured.
New developments such as generative AI can also enhance productivity through the automation of routine tasks for employees in domains that require high levels of skills, such as underwriting, actuarial or data analytics. This can shorten the task duration and allow employees to focus on higher-value tasks that are more rewarding.
Tong: The AI landscape is evolving at a rapid pace, especially with the promise of generative AI and how it is touted to be a game-changer across industries. Generative AI is already being used to create new products and services, improve existing ones, and solve complex problems.
For the banking industry, in addition to enhancing customer experience, generative AI can also optimise operational efficiency to reduce cost, improve anti-money-laundering capabilities to reduce risks, and enhance our capabilities to deal with increasingly sophisticated scam incidents.
We have started leveraging generative AI as part of our strategy to uplift the digital capabilities of our workforce. Generative AI allows us to automate routine tasks, freeing up employees’ time to focus on more strategic and creative aspects of the job. This will not only lead to improved productivity but also enhance employee satisfaction and retention.
Key to enabling these is the ability to analyse vast amounts of data in real time, enabling businesses to make informed decisions quickly and improving productivity. This means huge investments in infrastructure, technology and especially, people.
We expect that the ongoing advancements in AI will present a wealth of opportunities for businesses to innovate and tap on, if implemented responsibly.
Tancy Tan: The emergence of AI models such as ChatGPT has brought the opportunities and challenges of AI to the forefront of the world’s attention. In the short term, we see opportunities in three primary areas: AI can enhance operational efficiency by automating repetitive tasks, improving data analysis, and reducing human error. The customer experience can also be improved, as AI-powered chatbots and digital agents can provide personalised support, faster response times, and tailored product recommendations. Risk management and compliance can also be enhanced as AI can analyse vast amounts of data to detect fraud and financial crime, assess credit risk, and help organisations comply with regulations.
In the longer term, we see opportunities in areas such as expanding AI-powered digital agents. We are looking to responsibly and ethically extend digital agents to support customers in navigating challenges, accessing global research, and completing tasks such as onboarding or credit and lending journeys. We also believe that AI can be used to develop advanced analytics and insights by analysing complex data sets to identify trends and provide actionable insights for better decision-making. AI can also be harnessed to develop new, customer-centric financial solutions that cater to evolving customer needs and preferences.
Goh: Some potential development areas and opportunities for AI are in customer service and engagement, regulatory compliance and risk management, as well as financial inclusion and social impact.
Natural language processing and sentiment analysis can help banks better understand customers’ emotions, feedback and expectations to provide more personalised engagement and services. Speed to market can also be improved through AI-assisted software development and improved testing, which allow banks to be more agile and flexible when responding to market changes.
With the ever-changing and complex regulatory requirements, banks can leverage AI to monitor and manage risks such as market, operational or cyber risks, by using predictive analytics and scenario analysis to anticipate potential outcomes and recommend optimal actions.
AI can also help banks reach out to underserved or unbanked populations to provide more affordable and accessible financial solutions such as microfinance or financial literacy. It can also measure and improve banks’ social and environmental impact, such as reducing carbon footprint or promoting diversity and inclusion.
In terms of workforce enhancement and empowerment, AI can complement current the workforce’s roles in data analysis, risk assessment, customer service, compliance and more. UOB has deployed Microsoft 365 Copilot, a generative AI-powered productivity tool, for frontline and back-end employees to improve productivity, collaboration and creativity in their daily work.
Han: The potential of AI is vast, and Asia is one of the fastest growing markets for AI adoption globally. In Singapore, the government’s National AI Strategy envisions AI to transform the country’s digital economy by 2030.
The combination of new capabilities of generative AI on unstructured data and the AI/ML models on structured data can bring new conveniences and insights to customers. At the same time, generative AI can help financial institutions to simplify, automate and increase the level of productivity.
What are the risks from wider use of AI in the financial sector, and how can these be mitigated?
Tong: The wider use of AI in the financial sector brings several risks such as the risk of algorithmic bias, where AI systems might make decisions that unfairly disadvantage certain groups. AI models are trained on data which, if mishandled, can lead to discriminatory outcomes.
We adopt a two-pronged approach to ensure that AI remains fair and ethical – driving internal compliance as well as working with regulatory bodies to define the industry standards. Internally, we have a governance framework in place to manage how we build and deploy AI models. As we increased our AI adoption over the years, we enhanced our governance framework to ensure that our models adhere to the Feat principles (fair, ethical, accountable and transparent).
At the industry level, we are a key member of the Monetary Authority of Singapore’s (MAS) Veritas initiative to ensure that AI algorithms developed are fair and do not discriminate against any groups within society. We have also contributed our open-source codes, which are typically not shared, back to the industry. By doing so, we can drive trust in the adoption of AI technology and foster innovation in Singapore’s FinTech ecosystem.
Ben Tan: AI has swept through various industries like a whirlwind in recent years, and admittedly, its use has been beneficial, but it is not without potential risks. For example, AI and ML can pick up biases from the data that is being used to train the system. This could lead to decisions that may be unfavourable for some groups of people.
This is why it’s important to have a system of checks and balances in place to prevent such risks. Prudential has a set of AI Ethics Principles which set out the standards expected in the design, development and operating of complex applications, as well as AI and ML governance to make sure that these solutions are deployed in a fair, ethical and transparent manner.
When introducing new AI tools to employees, it’s also crucial to ensure that they are aware of the risks and understand how they can best leverage AI. These are some ways to help us mitigate potential risks and deploy AI responsibly to build and sustain the trust of our stakeholders.
Goh: While exploring more and better ways to leverage AI, UOB is aware of its associated risks. For example, AI may introduce or amplify biases or discrimination in the provision of financial services, especially if the data or algorithms used are not transparent, explainable or auditable.
AI may raise ethical dilemmas regarding the accountability and responsibility of AI decisions. It may also create new sources of operational risk, such as human-machine interaction, model risk or algorithmic risk.
To mitigate these risks, we remain committed to adhere to MAS’ Feat framework and will be guided by its principles when deploying AI. We actively collaborate with industry peers and regulators to develop guidelines and standards for the responsible and secured use of AI.
UOB has established robust governance, risk management and internal control mechanisms for the oversight, monitoring and evaluation of AI usage, while ensuring these AI objectives, incentives and outcomes are aligned with the strategic goals and values of the bank.
Tancy Tan: As with most emerging technologies, the opportunities it brings is usually accompanied with some risk. Some of the changing threats facing us and our customers because of generative AI include e-mail and voice phishing – where generative AI can be used to make phishing e-mails and voice messages more believable – or synthetic identity fraud – a type of identity theft in which real and fake personal information is used to create a fictitious identity, which is then used to commit fraud.
We have an established approach in managing such risk to ensure that we are delivering value to our customers and employees in a responsible and ethical manner. We do this by ensuring AI deployment adheres to HSBC’s risk framework and complies with the laws and regulations across markets while simultaneously pursuing innovation. We also ensure deployments align with HSBC’s principles of responsible and ethical use, with a focus on transparency and accountability.
On top of our established frameworks and processes, it is also critical for users of AI to fully understand the technology and its associated benefits and risk. Finally, for AI to fully benefit the industry and consumers, the banking industry together with regulators have to work together to ensure proper governance and accountability in the use of AI.
Han: As the adoption of AI increases in the financial sector, it will result in greater volume of data used for models and higher dependency on various type of AI/ML or generative AI models. This will bring three kinds of risk into focus – data management risk, model risk and unintended consequences on society.
Data management risk is about the right to use data responsibly, and the protection of data. In addition to ensuring that data is kept securely and data quality is good, companies will need to ensure that it is legally and morally permissible to use the data. This is essentially asking the question: Can we use the data?
Models can be biased if trained on a wrong set of data. To avoid such risk, financial institutions need to consider the source of data and the ethical dimensions, together with the firm’s core values, to ensure that the models do not result in loss of trust. Hence, the models need to be assessed against potential bias or exclusion of certain segments of society. This is essentially asking the question: Should we use it?
Finally, strong governance across development is required, particularly for the deployment of more advanced AI-based systems to ensure fair, transparent, explainable and accountable outcomes for the system and reduce unintended consequences on society. This is essentially asking the question: How do we do it?
Hence, use of AI in the financial sector requires a holistic approach, including putting in place the right controls and guardrails to ensure that AI-related risks are adequately addressed.
Copyright SPH Media. All rights reserved.
