Retail Banks Need to Get Smart With Artificial Intelligence

Retail banking is changing at a blistering pace. Online only challenger banks like Monzo have shot from 50,000 to 750,000 customers in little more than a year. Physical bank branches continue to close, meanwhile some 670 have already shut their doors this year, 879 last year, according to data from consumer rights organisation Which. Never, as a result, has the banking world been more in the public eye.

From small fintechs, to multinational tech giants that are increasingly muscling into the payments space, competition is also rife – a challenge compounded for incumbents by regulatory headwinds. Yet the end is far from nigh: emerging technology offering huge opportunities to personalise services, improve efficiency and reduce friction in an era of shrinking margins; a win-win for banks and consumers.

Emerging actors appear at first glance to have a headstart: tech is in their DNA and they don’t sit on any legacy infrastructure. Yet retail banks with decades, sometimes centuries of experience in the sector can and must adapt. To do so, they could draw on the experience of retail, where “segment of one” marketing powered by big data means consumers are increasingly targeted with highly personalised offers; even website landing pages.

Banking is shifting to a similar data-driven, open banking marketplace, which requires a much elastic, robust approach to how applications are incorporated into a broader stack. Delivering on promises, many argue, means making greater use of Artificial Intelligence (AI).

Automatic for the People

IBM’s Eddie Keal is among AI’s advocates. He told Computer Business Review: “In many banks, far too many processes that could easily be automated are still manual, and require a branch visit (think queues) and reams of paperwork. I know this from regular personal experience. Outside of work I am one of the directors of my local rugby club, which regularly gives me fresh – and sometimes painful – insight into life as a business customer.”

He added: “It is time that banks started putting AI to work much more widely; both in apps and in branch: most hurdles to better banking involve branch-based knowledge of rules and skills that can be encoded in AI. They can then be made available to customers as self-service functions, helping guide and reassure customers, improve security and reduce friction. Better yet, AI learns through experience, so it improves with every interaction.”

Generating Better Products

It’s not just customer services processes though that can be improved. What banks can also do is use AI to create products that match lenders with borrowers. Credit risk management in real time is complex, but it is possible and banks have huge scope to harness AI to deliver real insight. The opportunities are significant: global business value derived from AI is projected to total $1.2 trillion (£ in 2018, an increase of 70 percent from 2017, according to Gartner. And online investment platforms like Nutmeg and equity crowdfunding site Seedrs have shown what an appetite there is for retail investment platforms.

Keal adds: “AI could radically change how retail banks match up investors or savers with investment opportunities. You can see the beginning of this with some of the peer-to-peer lenders. These allow customers to choose a risk profile or a return profile before they invest. You can even specify that your money is not to be lent to firms doing animal testing, or rely on fossil fuels. Intelligent systems can help deal with the difficult maths and risk analysis of this sort of saving and investing.”

AI can use correlations and covariance matrices across large data sets and can begin to derive deep understanding from data (as long as it is clean – garbage in, garbage out, after all) for example; determining that a customer has excess cash available and using an AI-generated risk profile to present personalised options to help the customer generate higher income via savings or investments. The emergence of a more open banking infrastructure makes this eminently possible. Why is it not more widely available? An Infrastructure “Shanty Town”?

As Hans Tesselaar, the Executive Director of the Banking Industry Architecture Network (BIAN) puts it: “Too often banks are stitching applications together with little forward planning: the result is like a shanty town, with poorly planned infrastructure, enterprise hygiene – and bad roads, not the gleaming city of the future!”

He added: “This is not a shock: after all, ever application has its own specific scope and boundary and every point-to point connection is unique. When a retail bank’s application portfolio grows, it often ends up with several hundred overlapping systems; adding to or enhancing them becomes an exercise in tracing complex interdependencies, which often become more brittle as they sprawl across a system.”

It is a point echoed by IBM’s Eddie Keal. He said: “Launching new functions into financial services needs more attention than almost any other industry. If an airline has a minor booking issue, they can put an understandably aggrieved customer onto the next flight. For a bank, a minor issue can mean someone else’s money being inaccessible; a real pain point for the customer and a real reputation risk for the bank.”

“Modernisation of core systems has to be progressive — you have to treat it as a journey. The destination is componentised architecture that separates key constructs and their assets from their core transaction engines, but this has to be carefully considered and professionally done. Security, resilience, scale, all demand exceptionally high levels of testing and reassurance. Once that is in place and your house is on solid new foundations, it is a lot easier and less risky for banks to acquire the necessary capabilities they need to support new business models and new functions; from cloud-economics and to sophisticated analytics in the customers' interest.”