Over the next few years, AI will bring about game-changing developments for the banking sector, reports Scott Thompson.
“I do not see AI technologies currently adding significant competitive advantages to banks – challenger or traditional,” says Devie Mohan, co-founder and CEO of fintech research company, Burnmark. “There are very few data points of AI use cases in banks to come to a steady conclusion.”
Mohan’s note of caution is refreshing in a technology realm defined by hype, and by inflated expectations of AI’s ability to transform entire sectors, such as banking.
However, over the next five years the majority of banks will adopt AI. This is because, like legal services, the banking sector revolves around enforceable rules and regulations, and patterns of behaviour on which AI can be brought to bear.
The rise of AI will take place in the context of increasing automation within the sector. For example, Mitsubishi UFJ, Bank of America, Merrill Edge, and UBS, are among the many companies pursuing either automated customer service or robotic investment platforms.
As part of these programmes, some are using AI to second-guess customer behaviour and persuade investors to think outside of their comfort zones – in other words, to disrupt some users’ cautious behaviour to banks’ financial advantage.
Neural networks are also helping the financial services sector to detect fraudulent transactions – as previously reported on Internet of Business – and to help with their own compliance activities.
- Read more: HSBC employs AI fraud detection in $2.3bn tech spree
- Read more: NatWest prevented £7m of corporate fraud using machine learning
However, the key difference moving forward will be in where AI implementations occur in the mainstream banking world. The majority of use cases today are in customer-facing areas, such as branch banking, chatbots, customer service, and offer creation, says Mohan.
For example, Lloyds Banking Group has partnered with anti-fraud and call centre authentication specialist, Pindrop, to identify fraudulent phonecalls, using a technology that can identify 147 different features from a voice conversation.
Meanwhile, India’s ICICI Bank is tapping AI for sentiment mapping in its contact centre calls. DBS, Barclays, Santander, Wells Fargo, Caixabank, and Swedbank have also implemented AI for better customer service and support. Meanwhile, Tandem Bank is using cognitive computing to offer personalised insights to its customers.
But moving forward, back office automation via AI will be the biggest opportunity. For example, the Bank of England is currently testing technology from MindBridge to spot abnormalities in financial transactions. And the use of AI in contract analysis and loan approvals is being trialled by Bank of America, ANZ Bank, and JP Morgan, among others.
Contracts, documentation, mortgage approvals, investment advice, and so on, will soon all happen without human intervention,” says Mohan. “The moment the capability comes into existence, all banks will use it with significant cost savings in mind, with no disparity between traditional banks or challengers.”
Case study: OakNorth
New market entrant, OakNorth, was the first UK bank to host its IT systems entirely in the cloud and has also been investing in machine learning and artificial intelligence programmes.
It obtained its UK banking licence in March 2015 and began trading six months later. To date, it has lent £1.5 billion to British businesses without a single default, it claims, directly helping with the creation of over 5,100 new homes and 4,000 new jobs. And it recently reported $10.6 million in profit in its second year of operation.
“The speed at which we have been able to grow our core SME lending business, coupled with the operational agility and smarter credit analysis capabilities that our IT environment offers, have all played a fundamental role in enabling us to achieve profitability this quickly,” says OakNorth co-founder, Joel Perlman.
OakNorth uses data analytics and machine learning to improve processes and credit decisions across the lifecycle of a loan. Its ACORN machine learning platform, which the rest of the bank’s fintech is built on, collects millions of data items on SMEs across various parameters, sectors, and markets, and deploys algorithms to identify which data the bank needs to make more informed credit decisions.
“ACORN’s team of credit analysts and data scientists manages the process, training the machine learning algorithms so that the platform continues to evolve and get smarter,” says Perlman. “This provides us with significant process efficiency gains, data-driven decision making, and smarter credit analysis capabilities.”
As a result, OakNorth says that it is typically able to complete loans in weeks – from first meeting to disbursement of cash – rather than the months it takes larger lenders. “This speed is one of our biggest competitive advantages and a key part of why borrowers choose to come to us, as opposed to their clearing bank,” he explains.
Another benefit is that the bank can be a lot more flexible in terms of structuring loans. For example, if a restaurant owner wants to open several new sites, OakNorth can adjust the covenants to the overall business, rather than link them to each opening.
“At large banks, what tends to happen is that covenants and the growth of the lending facility are tied to every new opening, so every new site has to be profitable before the restaurant can open a new one,” says Perlman. “But what our technology enables us to do is take a holistic view of the company and determine whether existing sites have sufficient capacity to pay debt to fund two or three new openings each year. This flexibility is another key part of what attracts entrepreneurs to our proposition.”
The future’s bright
A report published last year by analyst group IDC revealed that 67 percent of organisations worldwide have already adopted, or plan to adopt, AI in the next five years, with an estimated $58 billion being spent on the technology across different industries over that same period.
“Right now, the most common uses of machine learning in banking are for fraud detection, chatbots, marketing intelligence – including customised product and service recommendations – and automated due diligence and compliance checks,” says Perlman.
“But going forward, I think we’ll see greater investment and innovation in areas such as credit analysis and underwriting, as well as attempts to automate certain processes, such as initial public offerings [an area that Goldman Sachs is investing heavily in], and the analysis of legal documents.”
He adds: “According to a recent study by Accenture Research, companies in the financial services sector that embrace AI could improve profitability by an average of 31 percent by 2035, so it’s not a case of if AI will see mass-adoption by financial services, but when.”
Burnmark’s Mohan also believes that the market is heating up. “We are already seeing that with the use cases we are exploring at Burnmark. The number that use AI has increased substantially since mid-2017,” she says.
Most banks are experimenting with AI for multiple applications across both retail and corporate banking, but the true benefit of AI will be in deploying it to improve the customer experience – without customers realising that it’s being done via a new type of technology.
“Banks need to stop promoting technologies to their customers, which is especially a problem in Asia, and build technology capabilities to meet and exceed customer expectations,” says Mohan. “And the great thing is that the impact of AI is easily measurable, through cost savings and improved customer satisfaction.”
Internet of Business says
“Stop selling technology and focus on the customer experience” is good advice.
The banking sector stands at a crossroads in terms of customer sentiment: memories are still fresh of the 2008-09 crash, the credit crunch, taxpayers’ multibillion-dollar bailouts, and the (ongoing) austerity programmes that followed. Many will also remember the market rigging and fraud within the financial services sector that have occurred since – which were touched on in our report on Mark Carney’s speech on cryptocurrencies earlier this year.
But stepping back, it is interesting how quickly things change in technology, and in banking. Back in 2015, Bank of America published a report on which technologies were most likely to disrupt the sector. It identified three broad trends: the Internet of Things (IoT), the sharing economy, and online services, with all three linked by robotics and automation. AI, however, was not singled out as a prime disruptive force; fast forward to the present, and entire countries and continents are reorienting their economies around the technology.
- Read more: UK launches “£1 billion” Sector Deal for AI | Analysis
- Read more: Europe announces €20 billion AI strategy – UK sidelined | Analysis
That 2015 report said: “We have entered a period of accelerated innovation, made possible by the confluence of many complementary enabling technologies required to change a business model. Think of it like this: big technology innovations often require many smaller technology advances first. This is the ‘building block’ approach to disruption, and we believe the majority of building blocks are now in place.”
The message is clear: disruption ahead. But it stands to reason that the outcome of increased AI and automation in banking – and in any sector – can only be as good or as bad as the thinking and assumptions behind the algorithms. Put simply, applications of the technology will either succeed or fail faster, more cheaply, and more efficiently, which will at least lower banks’ costs and minimise the chances of them losing money or being defrauded themselves.
But that is not the same as mitigating against banks’ risky behaviour, or their appetite for replacing the predictability of compound interest over time with a sector-wide desire to gamble – as the 2008-09 financial crisis amply demonstrated.
Additional reporting: Chris Middleton.