AI's expanding role in financial services
Artificial intelligence is increasingly being deployed to automate customer services and manual processes in banking and investing.
The use of artificial intelligence (AI) in finance has thus far been largely focused on two areas of automation: at the front end, often in customer services, and in the back office, to automate manual processes.
The financial services industry is at the forefront of testing the waters of automation in customer service. İşbank, Turkey's largest private bank, for example, has launched an AI platform that supports the same natural language flow that would be provided by a personal, human banker. Credit card companies are considering uses of AI such as for detecting potential fraud. Investment bank UBS uses AI to assess the risk appetite of potential customers with a low minimum investment; other investment bankers hope AI will help them make more accurate forecasts and estimates.
In the back office, robo-accounting systems enable completely autonomous accounting documents processing. AI-based systems can also review contracts, sifting through thousands of commercial loans that would take humans hours. Meanwhile, digital tax assistants — ones that can use AI to help avoid filing errors — have been proposed. (See the sidebar, "What Is AI?", below.)
But AI promises to do much more, and businesses are looking at all sorts of ways they can benefit from the technology.
AI performs customer services
Burak Arik, CEO of Maxitech, the innovation arm of İşbank, has been tasked with digitising more customer services. The bank already offers commercial and personal internet banking as well as cardless cash withdrawals and deposits through a smartphone.
"I've been looking at a lot of startups in Silicon Valley," he said. "We are investigating many companies to look for ways to improve efficiency with AI."
Clinc, the AI platform İşbank is deploying, can answer questions on a range of customer service issues from account balances to finding the nearest ATM, and Maxitech is going to introduce other functionality such as using it to make money transfers and cancel lost cards and order replacements.
"A voice on the end of the app gives you the answers, but you can also text a question and receive a text back," Arik said. "We are using it for two reasons, to save on call centres, but also we want to give customers a better experience. We believe voice interaction is going to be the future."
Himi Khan, vice-president of product management at Michigan-based Clinc, said the system can be deployed in a number of platforms — in mobile apps, at call centres, or even at branches.
"The deployments are dependent on use cases," he said. "It can change the human-in-the-room experience. It understands what we are saying, and it integrates very easily."
Policing finance
AI could also prove very useful to credit card organisations in fraud detection, something that costs the industry over $16 billion a year.
"Neural networks work on inputs and outputs. Input is the description of the problem; output is the answer. For example, the input is the point-of-sale situation, the customer's details; the output is the decision as to whether a sale is a legitimate or a fraudulent transaction," said Michael Wellman, professor of electrical engineering and computer science at the University of Michigan.
By analysing the inputs — data about a person's past transactions, including where they live, where they commonly shop, or even what they buy — a neural network can make accurate predictions to the legitimacy of transactions, he said.
İşbank is already in talks with a company developing AI for fraud detection, Arik said, adding: "We can see it has the potential to make huge savings."
The bank is also considering AI-powered robo-advisers, image recognition, and AI forecasting and estimation software for investment, he added. "AI financial forecasting is a primary focus, I think. It has the potential to be both a critical and very powerful tool for finance organisations."
Seeing into the future
The potential for AI to analyse huge amounts of data and make accurate predictions is never starker than when you consider the world of investments, where predicting the future can mean huge profits.
"Prediction is already very important in a lot of areas related to finance. A lot of finance trading is automated, and this is already based on short-term predictions, so there is no question AI could be extremely useful," Wellman said.
UBS reportedly invested $11 billion in AI between 2010 and 2017, a figure that is set to quadruple by 2020.
The bank already has a robo-adviser that provides automated investment advice based on customers' financial circumstances, and it has unveiled plans to expand a site in southern Switzerland into a centre for AI, analytics, and innovation that will focus on finding specific AI applications.
Credit Suisse has partnered with Spanish data analytics provider RavenPack to create an AI system that comes up with investment strategies by analysing financial news. Meanwhile, JPMorgan has trialled LOXM, an AI system that draws on data from the bank's past trades to execute client orders in Europe's equity markets.
The ability of AI to make real-time decisions based on big data is something finance organisations could benefit from. (For tips on how management can harness AI to use data to help address strategic decisions, see "Making Faster Decisions With AI", FM magazine, Feb. 2019.)
"Most of the processes in the financial world on a strategic level become stale really fast," said Clinc's Khan. "To get the data, then action, it can take hours, days, weeks, months.
"AI can crunch the numbers from these great data lakes and understand what's happening in the market and take immediate actions."
What is AI?
AI in its most advanced iteration is based on machine learning, and one subclass of this is known as "deep learning", where algorithms are used to form neural networks, like the connections that power human brains. These neural networks can analyse and learn from big data — huge swaths of information that no human being could process.
Research by McKinsey Global Institute suggests that AI that allows deep learning can boost value relative to other analytics techniques by 30% to 128%, depending on the industry.
Richard N. Williams is a freelance writer based in the UK. To comment on this article or to suggest an idea for another article, contact Sabine Vollmer, an FM magazine senior editor, at Sabine.Vollmer@aicpa-cima.com.