How are Financial Institutions Implementing Artificial Intelligence & Machine Learning (ML)?

In our earlier discussions on Elevating Customer Experience, Digital Technologies for Transformation and Open Banking in April 2021, Artificial Intelligence, Machine Learning and Natural Language Processing are terminologies commonly brought up by panelists.

According to AIBP’s 2020 ASEAN Enterprise Innovation Survey, respondents from ASEAN BFSIs listed big data analytics (66%), machine learning (41%), artificial intelligence (40%), and robotic process automation (27%), as the top technologies that they are planning to invest in over the next two to four years. 

“AI is used to complement human interaction, never as a substitute… it is about leveraging the human touch and technology touch.” - Federico Brandi, Chief Marketing Officer, Roojai

“How can AI help improve processes, operations, and [facilitate with] decision making efficiently?” - Yi Chen, Head of Data Innovation at AXA Mandiri

These statements are top of mind for many stakeholders and most of them are working to explore ways that Artificial Intelligence (AI) and Machine Learning (ML) can help them deliver their objectives. As a type of AI, ML uses statistical models and algorithms to analyse and draw inferences from patterns in data, applying that learning without the need for human intervention.

Below are some examples of ML applications across various functions in the BFSI industry:

1. Creating Personalised Experience in Insurance Premiums through Risk Profiles 

Big Data capabilities can track and store as much information about the bank’s customers as needed, providing the most precise and personalised customer experience. Optimising the customer footprint allows banks to leverage the analytical capabilities of ML to detect even the slightest and most subtle tendencies in customer behaviour, which contributes towards a more personalised experience for each individual customer. 

By creating a personalised insurance profile for their customers, insurers in Malaysia are looking to introduce algorithms to apply individual risk scoring for each customer based on the customer’s historical and personal data. Those with better driving records and more driving experience will get cheaper premiums. Hence, insurance agents are not required to manually evaluate each customer’s suitability to a particular premium, thereby increasing efficiency and productivity. This is the second stage of the Phased Liberalisation of the Motor and Fire Tariffs announced by Bank Negara Malaysia in 2016. While this approach raises personal data privacy concerns, a survey conducted by the Center on Global Brand Leadership, Columbia Business School, noted that 75% of participants are willing to share their personal information if this leads to a better and more trusted product or service.

2. Increasing Efficiencies in Customer Assistance via Chatbots

Chatbots designed and implemented by financial institutions are programmed with both ML and Natural Language Processing (NLP). Building and programming bots can be time-consuming initially, but in the long run, can help to save a lot of time and money, and can improve efficiencies by providing fast-paced communication, round-the-clock support and personalised experience for customers through instant access to customer’s data. 

The success in improved customer service is evident in CIMB Bank’s launch of its chat-based virtual assistant, Enhanced Virtual Assistant (EVA), which has managed to generate approximately 130,000 downloads and 300,000 transactions in less than a year. Since the introduction of EVA, the total number of active mobile banking users has increased by 6.3%, and the mobile banking active user rate improved to 41% compared to 38% prior to the launch of the project. By prioritising conversational user interface (UI), CIMB is able to construct a chatbot that is highly valued by customers. 

3. Enhancing Security when Transacting with Image Recognition

Image recognition is a technology capable of identifying objects, places, people and actions in images. It employs ML algorithms that capture, find, store and analyse features in order to verify them with images in a pre-existing database. Image recognition can be used either for enhancing customer experience or security. An example of the former is ID scanning in banks or credit card scanning functionality in payment applications. In these mobile banking applications, the security feature entails either ID or biometric authentication. 

In some instances, there is no need for an ID or even a credit card - you can pay with your face and a camera! Point-of-scale systems at stores in China applied facial recognition technologies, and customers who use services like Alipay or WeChat Pay can enable facial recognition in applications, and pay using connected devices. Wenzhou, a shopping street in China, already allows customers to pay just by showing their faces to the camera of an Alipay device. 

4. Mitigating Fraudulent Transactions by Analysing Real-Time Data 

Taking into consideration hundreds of factors including the device used, customer’s location and transaction history, AI and ML algorithms are capable of processing large amounts of data in real-time. These algorithms function significantly quicker than humans when detecting new and emerging security threats, be it a suspicious transaction that may require additional authentication, hackers stealing credit card information and attempting to make a purchase, or a spike in traffic from an unusual source. McKinsey estimated that losses from such fraud could amount to $44 billion by 2025.

Moreover, since AI and ML algorithms can analyse much larger data points, connections between fraud patterns and entities, the emergence of false positives can be drastically reduced. This means fewer customers will be falsely alarmed for fraudulent concerns and fewer will be falsely rejected for fraudulent transactions, which in turn minimises the labour and time costs associated with allocating staff to review flagged transactions. 

Many banks in ASEAN have been successful in fraud detection and an example would be UOB in Singapore. UOB utilises an anti-financial malware system based on ML that tracks the device its banking services were accessed from and the behavioural biometrics of the user in real-time where any deviation from the transaction patterns of customer accounts would indicate a potential scammer, and thus alert the banks. These have proven effective with UOB being able to recover $6.69 million in a business email impersonation scam in 2020.

Therefore, through the application of AI and machine learning algorithms, financial institutions can analyse a wealth of information from several different data sources and channels in real-time, allowing them to make critical security-related decisions almost instantaneously to prevent fraud, without compromising on the customer experience. Whilst these have been effective in most instances, moving forward, it is imperative that financial institutions collaborate as they harness and enhance their technology to keep up with the changing tactics of scammers, as part of the efforts in stopping fraudulent banking activities. 

To conclude, it is clear that advances in ML have been instrumental in various applications across the BFSI industry. We are excited and stoked to find out more and learn about what ASEAN financial institutions are implementing in this space. 

Stay tuned for our upcoming session on Innovating with AI Technologies and Data Analytics in the BFSI Industry on 22nd June, 1430-1600 (GMT +8).

You can register your interest here.