Machine learning applications in banking

The banking industry is in a digital arms race. In 2018 alone, it is estimated that Wholesale banks globally invested approximately US$9.7 billion to enhance their digital banking capabilities in the front office. For retail banks, online and mobile channels have become equally important as their ATMs and branches.

Banks around the world are already realising how investments in digital technologies could benefit customer acquisition and satisfaction. But satisfaction is relative. As leading technology brands, such as Apple, Amazon, or Google, have become the gold standard for digital engagement, many consumers now have a stronger emotional connection with these brands than they have with their banks. If banks want to keep up, they have to manage the digital experience they offer to make these emotional connections, which, ultimately, could translate into sticky interactions and more profitable customers.

Against such a backdrop, this article explores how some of the pioneers in the banking industry are embracing machine learning and AI to strengthen their digital capabilities.


Links to Relevant Books

The rise of the Robo-advisor

Robo-advisors, or robos, are online services that use algorithms to automatically perform many investment tasks done by a human financial advisor. Initially offered by startups, robos are now part of the suite of services offered by major financial institutions. Since they are less expensive than a human advisor, they democratise access to financial advice. Robos can take on customers with little savings since adding one more person wouldn’t cost much more. For example, HSBC recently rolled-out an online platform designed to make wealth advice accessible to existing customers with as little as £1,000 to invest at an initial charge of 0.5 per cent followed by fees of up to 0.46 per cent.

Signing up starts with consumers filling out detailed questionnaires online about their financial goals, risk tolerance and investment timeframes. Robos take the information and use computer algorithms to come up with an asset allocation that fits the customer’s needs. Once the portfolio is created, robos also manage it, doing things like rebalancing the portfolio, executing trades, performing tax-loss harvesting and other actions.


Voice and chat driven virtual assistants are taking control

If you’re a Bank of America customer, you may already know Erica, a virtual financial assistant – or you probably will soon. Erica is billed as the first widely available AI-driven virtual assistant in financial services, and is available for free within the Bank of America mobile app to more than 25 million mobile customers. A recent press release by Bank of America indicates that more than three million customers have already used Erica.

Erica is essentially a financial AI assistant rather than a general personal assistant such as Alexa or Siri. Erica takes certain tasks away from users, doing them itself when guided by voice commands, text, or with gestures. The user can, for example, ask Erica to lock their debit card if it is lost, transfer funds from checking to savings, send money to someone on your behalf, pull up scheduled payments and existing bills, retrieve credit card limits, account numbers, payment confirmations, and more. The actions aren’t limited to controlling the account, but also include retrieving specific data that would otherwise require a lot of hunting.

To gain a better appreciation of what a voice and chat driven financial assistant can do, I will highly recommend watching Erica’s promotional video (follow the “Watch” icon in this link).


AI that executes trades

Around mid-2017, JPMorgan announced a first-of-its-kind AI to execute trades across its global equities algorithms business, after a European trial of the bank’s AI programme showed it was much more efficient than traditional methods of buying and selling. Codenamed LOXM, the job of this AI robot is to execute client orders with maximum speed at the best price, by using lessons it has learnt from billions of past trades – both real and simulated – to tackle problems such as how best to offload big equity stakes without moving market prices. It is worth pointing out though that in many jurisdictions, best execution is a legal mandate that requires brokers to provide the most advantageous order execution for their customers given the prevailing market environment.

And J.P.Morgan is not alone here. Behind the scenes, financial institutions and hedge funds across the world are already in a race to develop and implement AI-based program to make trading more profitable and efficient.


Detecting and preventing fraud

Machine learning techniques have proven to be much better than traditional rules based approach to detecting fraud patterns in financial transactions. I wrote a whole article on this topic, so I will not go through it again. But if you are interested in reading more on this subject, here’s the link to my article on “Machine learning approach to fraud detection” and a second article by PYMTS on how HSBC is using “AI to boost its digital banking immune system”.


The lawyer that is a bot

At JPMorgan, a learning machine is parsing financial deals that once kept legal teams busy for thousands of hours. The program, called COIN, for Contract Intelligence, does the mind-numbing job of interpreting commercial-loan agreements that, until the project went live in mid-2016, consumed 360,000 hours of lawyers’ time annually. The software reviews documents in seconds, is less error-prone and never asks for vacation! According to its designers, COIN has also helped JPMorgan cut down on loan-servicing mistakes, most of which stemmed from human error in interpreting 12,000 new wholesale contracts per year.


Underwriting and credit scoring

Credit in China is now in the hands of a company called Alipay, which uses thousands of consumer data points — including what they purchase, what type of phone they use, what augmented reality games they play, and their friends on social media — to determine a credit score.

Unlike traditional models of underwriting which focus on only a handful of credit attributes, machine learning algorithms can analyse thousands of data points from credit bureau sources and other non-traditional channels such as the Internet and social media, to accurately model credit risk for any consumer. As such, with machine learning, lenders can also assess the credit risk and assign scores to customers with little or no credit history (i.e. the world’s unbanked).


Links to Relevant Books

Leave a Reply

Your email address will not be published. Required fields are marked *