Computers have reached such a level of sophistication that they can now outperform their human counterparts at some tasks, a fact that has not gone unnoticed in the banking community. But, while smart machines are increasingly being utilised in customer service and data analysis capacities, advocates of these new technologies maintain that they are designed to complement rather than replace the traditional workforce.

They can ‘think’, they can speak and their advice may be better than a human’s. Smart machines are increasingly being used by banks to automate various tasks, although the industry – and the world at large – is nowhere near a science-fiction scenario where artificial intelligence will replace the need for humans altogether.

At one time, artificial intelligence referred to computer programmes that acted similarly to humans, explains Larry Wasserman, a professor in the statistics and machine learning departments at Carnegie Mellon University, Pittsburgh. But there has been a shift in the field, toward programming for specific tasks, such as identifying whether an email is real or spam. This kind of large-scale pattern recognition is already being used across the banking industry, whether it is monitoring unusual spending patterns to detect credit card fraud or writing algorithms to beat the stock market. 

Developing their senses

Banks have been using predictive analytics to anticipate how customers will behave – whether they are likely to repay a loan – as well as scanning the web for sentiment analysis to predict which fee hikes will be acceptable, or unacceptable, to customers.

Nedbank in South Africa uses sentiment analysis to monitor its brand. Eugene Liebenberg, head of decision science modelling at the bank, explains that data from social media sites – such as Facebook, Twitter and LinkedIn – is analysed and could, for example, be used to predict a rise in complaints to the call centre. Also, the bank uses data from its customers’ social media profiles to predict what those specific customers’ needs are.

And, at the consumer end of the spectrum, solutions that reveal patterns to individuals about their behaviour are also being developed. One such example, which plans to launch later this year, is Wallet.AI, which takes information on an individual customer – including their social network activity – and looks for patterns in their financial behaviour.

“The purpose is to show you what you are doing in a way you would not normally see,” says Omar Green, founder and CEO of Wallet.AI. And, unlike humans, Wallet.AI “never gets bored, tired or stale. If you have the smartest, brightest, most well-educated mathematics genius, Wallet.AI could still out-calculate that – it is a machine that only calculates.” 

Cyrille Bataller, European director of Accenture technology labs, the technology research and development arm of consultancy Accenture, describes the broad area of cognitive computing as having three underlying technologies: those that enable computers to sense, comprehend and act. This means, for example, when they sense they can analyse video, when they comprehend they can process spoken language and pages of text and when they act they can drive cars or self-tune systems. 

“The key innovation around cognitive computing in our view is the ability to create expert systems that will answer questions knowledgeably in a domain without having to code lines of base rules,” says Mr Bataller. He explains that 30 years ago it was possible for machines to process complex questions and give answers, but this would have to be coded. By combining natural language processing and machine learning, says Mr Bataller, the computers are now more like bright students, who are given documents to read and they can learn by themselves.  

A life of its own

The most likely application of this technology in the banking industry will be in the form of virtual agents. An avatar may interact with an online customer over instant messenger and answer their queries. Mr Bataller says that when the virtual assistant is unable to answer the query, it hands over the dialogue to a human operator and learns how the human deals with that particular question, and adds this to its evolving body of knowledge. 

There are downsides of this technology, and Mr Bataller points out “they may learn and comprehend and tune in a way that you do not approve or don’t validate”, he says. With something that has been totally coded, the programmer can go in and fix the bugs, but this kind of machine learning has a life of its own. “If the help desk interprets data in an inaccurate way, it may do things that you had not intended,” says Mr Bataller.

Another application of cognitive computing for the banking industry – which is further away from being commonplace – is answering questions in spoken language, drawing on a vast database of knowledge and providing answers or recommendations back. The most notable innovation in this area has been IBM’s Watson, a computer that made its debut in 2011 on the US quiz show Jeopardy. 

Citi was one of the first banks to trial the use of Watson, and the Royal Bank of Canada and Australia’s ANZ have also been exploring how the technology can help them engage with their customers. Watson is an exceptionally clever search engine, through which, rather than manually trawling through reports, databases or spreadsheets, answers can be found quickly.

At DBS Bank in Singapore, relationship managers are using Watson to help them advise their wealthy clients. Olivier Crespin, chief operating officer of consumer banking and wealth management at DBS Bank, says that Watson enables the relationship manager (RM) to amalgamate all the investment literature that is available and apply it to the specific needs of their clients. 

“While wealth management will pioneer these capabilities, we do see many areas across our various businesses in which we can employ Watson... These opportunities could range from customer call centres and credit risk management to trading,” says Mr Crespin. He explains that Watson is not intended to replace humans. “On the contrary, we hope that the successful implementation of this solution would actually increase client interaction as our RMs will be better able to spend more direct time in higher quality engagements as opposed to culling through data, preparing for them.”

Doing the leg work

Tom Austin, vice-president at consultancy Gartner, says that rather than looking at the question of how smart machines are replacing jobs, it is worth considering what jobs will now be made possible, and how existing jobs are being upskilled, because of this technology. He uses the example of Watson being used by some doctors to help diagnose and treat patients. Watson, who has read all the medical journals – something even the best doctors cannot claim to have done – is able to interact with the doctor, ask the doctor to perform more tests and is able to give its recommendations.

Harvey Lewis, research director of analytics at Deloitte, raises questions about the use of this technology: “With knowledge-intensive jobs – doctors, lawyers, consultants – that depend on skills and a deep set of knowledge, the question is how to automate that,” he says. “How much trust are you going to put in a machine rather than a human? What if it means putting trust in something that it has not done before?”

Smart machines are "probabilistic", explains Mr Austin, which means they assess how certain they are of an answer. “The smart machines will not work with simple algorithms that say for given inputs a, b and c the answer is x.” And in the case of virtual agents, and Watson, they are smart enough to know when they don’t know the answer.

Another characteristic of smart machines is that they are narrow in their function, or as Mr Austin puts it: “The smart machine that drives the automobile will not drive the plane.” Mr Wasserman says that even the uses of Watson are narrow: the computer had to train specifically for the Jeopardy game, for example.

Human touch

Another example of a narrow use of a smart machine is Warren, which has been developed by US software engineering company Kenshō. Daniel Nadler, co-founder and CEO of the company, explains that Warren has been developed to help financial professionals find anomalies in asset pricing. Where previously it may have taken days for quantitative analysts to create statistical models, now Warren can do it in a fraction of the time. Warren, says Mr Nadler, is “taking the way humans think about the world and doing things that humans are bad at”.

Financial professionals have always asked what causes prices to move, and what the market impact of a certain event will be. For example, it is possible to ask Warren – in natural language – what the market impact of events in Ukraine will be. Warren is able to scan vast databases and give answers back, in a natural language. Google Ventures, the venture capital arm of digital conglomerate Google, has invested in Kenshō and Warren is currently being piloted at a major investment bank.

Mr Nadler is hesitant about making Warren available to retail investors. There is a danger that retail investors – without the risk controls of an investment bank – would take a pricing anomaly as a reason to buy a particular stock without having an appropriate exit strategy. This would be the equivalent of IBM’s Watson being available to the general public to help them diagnose and treat their own medical conditions.

“The doctor is intermediating the technology. I’m not some instant techno optimist who believes advanced computer systems are going to replace people and people will lose their jobs. Theoretically I want [these machines] to succeed, but I am conservative about this. You are always going to need human professionals intermediating between the computers and the rest of the population. They are just brain extenders,” says Mr Nadler. 

PLEASE ENTER YOUR DETAILS TO WATCH THIS VIDEO

All fields are mandatory

The Banker is a service from the Financial Times. The Financial Times Ltd takes your privacy seriously.

Choose how you want us to contact you.

Invites and Offers from The Banker

Receive exclusive personalised event invitations, carefully curated offers and promotions from The Banker



For more information about how we use your data, please refer to our privacy and cookie policies.

Terms and conditions

Join our community

The Banker on Twitter