Cover Story: The future of AI in financial services

This article first appeared in Personal Wealth, The Edge Malaysia Weekly, on August 21, 2017 - August 27, 2017.
-A +A

Artificial intelligence is poised to disrupt the banking and wealth management industries. With the technology, financial service providers will be able to engage customers more intuitively, offer real-time support and turnaround and provide quality investment advice at a lower cost.

 

In the romantic science-fiction movie Her, the main character Theodore, a lonely and depressed man, falls in love with Samantha, an artificial intelligence-based computer operating system. Throughout the movie, Samantha learns and develops from a simple system to one that is able to hold a proper conversation, understand reasons and perform sophisticated tasks such as filtering Theodore’s inbox and arranging his literary letters. 

Her abilities advance to a point where she is able to hold multiple conversations with people and interact with other operating systems. Finally, she leaves Theodore to be with other operating systems to discover more about her existence. 

While this level of AI is mostly seen in movies, life is looking to imitate art. The current AI technology is known as cognitive computing, which means it mimics certain mechanisms of the human brain, such as language and image processing. There have been great strides in the development of this technology in the past few years and sectors such as banking and wealth management have benefited from them.

Rajnish Khare, head of digital transformation, social business, new media and mobility banking at India-based HDFC Bank, believes that consumers will benefit from AI technology in terms of reduced waiting time in customer service or real-time turnaround for banking and wealth management services, among others. “For example, in banking, customers will be able to apply for a credit card via a chat-based system from their own office or home. They will be given a real-time decision on their application status and should it be approved, they will receive a virtual credit card for instant use. The back-end system integration could enable the virtual credit card for use with point-of-sales terminals via a mobile device,” he says. 

Rajnish was a panel speaker at the BankTech Asia 2017 conference in Kuala Lumpur, where he spoke on the “Rise of AI: Will Cognitive Computing be the Future of Banking?”. 

He says AI-based solutions will enable banks to make crucial real-time decisions. “For digital banking, the intelligent system will allow non-human interface to facilitate banking transactions and advisory, support instant credit decisions or make cross-sell suggestions. Clients will be able to interact in natural language with the bank’s chatbots, which are enabled by a cognitive engine. By providing access to self-help customer service, this will help reduce waiting time for customers.

“In terms of risk management, AI will provide a learning system for real-time fraud detection, providing regulatory oversight and trade secret protection, among others. The importance of real-time monitoring of money transfer will increase, especially with the advent of cryptocurrencies such as bitcoin and ether.”

According to Deloitte, natural language generation is the part of AI that helps turn data into facts and draw conclusions from those facts. This makes the data highly useful and relevant in a contextual way. Meanwhile, chatbots are software applications that engage in natural language dialogues with users and perform tasks on their behalf. 

AI will also help to detect fraud. Rajnish says systems are being designed to implement dynamic rules via supervised machine learning for fraud detection, which uses historical transaction data. After the introduction of multiple channels through which customers can transact, the need for machine learning arises. It then creates unsupervised rules for anti-money laundering and fraud detection, based on real-time transaction tracking mechanisms. It will also help banks with data analytics. 

“It will enable banks to turn complicated analytics into simple stories people can understand and then act on. Machine learning tools are boosting the effectiveness of marketing-related analytics by enabling strategies on churn prevention and micro-pricing,” Rajnish adds. 

HDFC Bank is at the forefront of AI-based solutions, ranging from customer service and core banking functions to operational efficiency, analytics and employee training. Rajnish says the bank is looking at the technology to create more efficiencies in the banking system and internal processes, starting with its application in the inherent processes to provide customer-facing products and services. It has successfully launched two such services. 

“We have deployed Eva, a customer service chatbot on our official website, to provide efficient customer service to our clients. They can ask Eva anything related to their bank accounts. If there is something Eva cannot answer, our customer service executive will take over,” says Rajnish. 

“The other AI-based services we offer is the HDFC Bank OnChat on the Facebook Messenger application. It provides natural language interactions to help customers with things such as bill payments, e-commerce services, cab hailing and movie ticket reservations. This chatbot provides integrated payment mechanisms for a host of services on Messenger.”

AI will play a key role in powering virtual banking, also known as video banking. One example is a solution known as LiveBank, provided by Australia-based Geniusto Pty Ltd and developed by Poland-based Ailleron. The solution has been deployed in 11 countries thus far — Australia, New Zealand, Indonesia, Singapore, Malaysia, Thailand, Cambodia, Vietnam, the Philippines, China and Japan. 

Geniusto co-founder and director Jeremy Thomas says there is growing interest in the solution from banking players in the region. He notes that businesses and their customers are having conversations everywhere with electronic devices, hence the need to provide banks with such solutions so that their customers can have access to an alternative channel with a full range of banking products. 

“Virtual banking is the merging of the traditional interactive branch experience with digital channels to allow banks to engage their customers in real time, collaborate and fully interact across web-chat, audio and video. This allows full customer service or an end-to-end transactions to take place,” says Thomas, who was a panel speaker at the BankTech Asia 2017 conference.

Geniusto is looking to incorporate AI solutions into LiveBank. Thomas says AI and virtual banking are a very logical and powerful combination as both are relevant to banks. 

“LiveBank promotes personalisation and human interaction in digital engagement whereas AI can provide automation and guidance to both customers and bank agents in the process. We are increasingly being asked to incorporate our solution with AI-based solutions, which is why we are currently working on several use cases that we think will present a compelling combination of both AI and human interaction,” he says. 

“In the use case, AI acts like a virtual concierge that can assist customers with their initial needs and provide them with guidance. This occurs when onboarding a new client or offering a new retail or investment product. Customers will be seamlessly engaged with the bank via LiveBank and the AI-based solution — the virtual concierge.”

Thomas believes that LiveBank coupled with the AI-based solution will be very powerful as it ensures that the customer experience remains in the digital channel and that the customer journey between AI and human is seamless. The solution can even be upgraded later with machine learning technology. 

“We see virtual banking evolving as new technologies, such as AI-based solutions, become available to provide better customer experience. Biometrics is another technology that will be coming into our solution. It will allow customers to voice-activate their sessions and the bank to leverage the security that comes with biometric scanning,” he says. 

Disrupting the wealth management industry

According to a PwC press release in June, the global gross domestic product will be 14% higher in 2030 as a result of AI. In other words, the technology will contribute US$15.7 trillion to the global economy, which is more than the current output of China and India combined. 

“Labour productivity improvements are expected to account for more than half of all economic gains from AI between 2016 and 2030. Increased consumer demand resulting from AI-enabled product enhancements will account for the rest,” says PwC.

“Overall, the biggest absolute sector gains will be in retail, financial services and healthcare as AI increases productivity, product value and consumption. By 2030, an additional 

US$9 trillion of GDP will be added from product enhancements and shifts in consumer demand and behaviour as AI-driven consumption gains overtake those of productivity.”

Wealth management, a sub-sector of the financial services industry, will be another area disrupted by AI. Rajnish says AI-powered robo-advisory will transform how banks and investment services providers serve their clients. 

“With AI, every customer can have access to a virtual asset manager [financial adviser] at the lowest cost possible. This will help democratise financial advisory services, allowing customers to get quality advice without paying too much for it,” he says.

“Today, relationship managers and most of the financial advisers are only available to high-net-worth clients. With AI-powered robo-advisers, this can be opened up to everyone, whether you have millions to invest or only a small sum to start with. Every investor can benefit from their advice.”

While earlier versions of robo-advisers only utilised algorithms in their platforms to allocate client assets, some are beginning to employ AI to enhance their offering. For example, in March last year, US robo-advisory firm Wealthfront introduced Wealthfront 3.0, the latest version of its platform built for increased artificial intelligence. 

According to the company’s blog, Wealthfront 3.0 will come to life with relevant, data-driven advice each time clients link an account or third-party service to their Dashboard. “Only Wealthfront provides recommendations on diversification, taxes and fees that are personalised not only to the specific investments in your account but also to your specific financial profile and risk tolerance. Do you have enough cash in your emergency fund? Are you holding too much stock in your employer? Wealthfront will help you,” says CEO Adam Nash.

The new version was launched in response to the broader trend showing the rise of AI applied to financial services, he says. “We believe that over the next decade, AI is poised to transform our industry. The entire fabric of the financial system will be rethought, redefined and rewired.”

Hong Kong-based robo-advisory firm 8 Securities launched Chloe, an AI and machine learning-powered robo-advisory mobile application, in February last year. For a minimum investment of US$88, the app gives investors access to ETF portfolios with exposure to 50 countries, 37 industries and 4,324 stocks and bonds. There is an annual fee of 0.88% of the investor’s portfolio. 

How does Chloe work? According to co-founder and CEO Mikaal Abdulla, the user needs to give the app some details about himself. Then, it suggests some investment goals. After the user selects an investment goal and target, Chloe will monitor the progress. Since the app uses AI and machine learning technology, it becomes smarter over time and is more interactive. This means that it will learn to anticipate behaviours and movements in the stock market better and take on more roles to communicate with clients.

Mikaal dismisses the complaint that robo-advisers are not personalised. He believes that such services can achieve hyper personalisation with the existence of AI as it can eliminate inherent conflicts of interest [such as some human financial advisers being too product-centric as they will get commissions from them] and learn to do it better with machine learning. 

Algorithm trading also benefits from AI. According to AI and You: Perceptions of Artificial Intelligence from the EMEA Financial Services Industry published by Deloitte in April, intelligent algorithmic trading enables high-frequency trading through adaptive pattern recognition and real-time processing of large amount of information, including index prices, sentiment indicators, social media and news. 

For example, Taiwan-based start-up Tixguru — which combines quantitative trading with robo-advisory services on one platform — is powered by AI. According to co-founder Chris Liu, the firm targets high-net-worth individuals (HNWIs) by providing fully automated execution of trading strategies and portfolio management based on AI-powered forecasts that are able to predict the movements of various asset classes. Tixguru also offers stock recommendations and portfolio management for non-HNWIs through brokerage firms.

AI also helps to facilitate seamless engagement between relationship managers and customers, thus improving the wealth management experience. According to Artificial Intelligence — Let’s Get Specific, published by Deloitte, the technology could personalise communication with customers using natural language generation engines that analyse movements in financial markets, assess the effect on an investment portfolio and automatically generate a personalised message to keep the customer informed or provide suggestions on trading opportunities. 

The report says AI can help to “personalise omni-channel marketing and service conversations using recommendations generated by complex analytic models deployed in real-time decision engines. These engines centralise intelligence, enabling consistent personalisation at every service event as well as proactive intervention on customer experience to address problems before they escalate”. 

Shankar Narayanan, co-founder of Singapore-based Active Intelligence Pte Ltd (Active.ai), says in an email interview that the start-up aims to provide AI solutions to banks and other financial institutions so they can intuitively and intelligently engage with customers via mobile, chat or voice-enabled Internet of Things devices. “We believe that conversation is the new way to engage with customers. Active.ai’s omni channel interface platform, AI engine and range of solutions will facilitate the conversation process,” he adds. 

“We offer real-time support to the human agent while on-call with the customer. There will be seamless handover between chatbots and agents.”

Benefits and risks

Last month, Facebook CEO Mark Zuckerberg and Tesla and SpaceX founder Elon Musk engaged in a heated and public debate on the future of AI. Musk is pushing for proactive regulations on the technology because he believes it poses a fundamental risk to the existence of civilisation. 

Later, Zuckerberg denounced these types of warnings as being “pretty irresponsible”. Musk responded by saying, “His understanding of the subject is limited”. 

Musk has taken to Twitter to express his concern about the safety of such technologies. “If you are not concerned about AI safety, you should be. Vastly more risk than North Korea.” 

So, does AI possess more benefits or dangers? Rajnish points out that as more financial information is put to use for the consumption of AI systems, the risk of valuable data being hacked increases to an extent. However, new-age encryption technologies, secure cloud computing and distributed networks will reduce the risk of such incidents. 

“We recently saw ransomware such as Wannacry block access to particular systems. This increases the need for more distributed computing, with data being secured with blockchain,” says Rajnish.

“Governments and financial institutions are taking a look at blockchain-based distributed ledgers to prevent such hacking or ransomware risks. Indeed, vulnerability will increase, but systems and checks will be implemented to minimise potential data theft in such scenarios.” 

Some critics say there is a risk of programmers extending their biases to the programming of AI, for example, a lack of diversity or inclusiveness, thus affecting the objectivity of the banking or wealth management products they come up with. Rajnish dismisses this and points out that AI-based solutions are not developed by an individual programmer. 

Wealth management and advisory services rely on the asset managers’ capability to read and understand the data around them. Asset managers often make intuitive guesses based on gut feeling, says Rajnish.

“Robo-advisory services will be powered by algorithms that will provide opaque answers with understandable justification for their decisions. The programmers behind the algorithms can only model the learning mechanism for such systems, they cannot make the AI engine look towards a particular asset class with objectivity,” he points out. 

“Even if a programmer codes a system to look at gold as a sub-standard asset to begin with, the AI will understand its position based on real-time input of big data and will perform a course correction within a limited time frame.”

Rajnish says a lot of open framework AI solutions, such as TensorFlow, API.ai and wit.ai, are being contributed by experts from all fields. General purpose AI engines learn from the overall data available for consumption, so such systems cannot be controlled. 

“AI engines created to establish specific purposes such as self-driving cars will learn from the schematic information of objects around them. AI cannot be tamed by individuals and organisations. Rather, all will contribute to certain aspects of the objective of smarter and aware systems,” he says.

The future

How will AI transform the world in the next three to five years? Active Intelligence’s Shankar says going digital and engagement with customers over messaging and voice are top priorities for banks and they are setting aside budgets for these. 

The company is building an AI stack, or multiple AI-based solutions that are grouped together, for financial institutions such as banks, payment companies and insurers. It will enable customers to interact with the institutions over unstructured forms of engagement using messaging or voice. “We enable our platform via the software-as-a-service model, thereby enabling banks to connect and pay on the go,” says Shankar.

Rajnish says it will be interesting to observe the changes. “I think the human touch can never be replaced by AI, especially to assure customers in times of uncertainty. Until AI replicates the nuances of human interaction and customers are comfortable with such interactions, humans in banking cannot be replaced.

“Until then, inherent AI systems will augment the knowledge and decision-making capabilities of humans. Wherever new technology is introduced, some job loss is evident. But new jobs in technology and management will augment such job loss. In any industry, especially banking, AI cannot replace jobs that require domain expertise and general management skills.”

However, it does not mean that AI will not improve itself over time. The keywords here are machine learning and deep learning. According to Deloitte, machine learning refers to the process of automatically discovering patterns in data. Computer systems have the ability to improve their performance through data exposure, without following explicitly programmed instruction.

Deep learning is machine learning algorithms involving artificial neural networks (computing systems that are designed to simulate the way the human brain analyses and processes information) that are inspired by the structure and function of the human brain. Interconnected modules run mathematical models that are continuously tuned based on results from processing a large number of inputs. 

Rajnish agrees that machine learning and deep learning are two key elements that make AI improve on its own. “I think AI, machine learning and deep learning will always go hand in hand as these are what enable AI to learn. Just like a human baby learns from the surrounding environment and interactions, AI uses machine learning and deep learning to learn and grow smarter.”