Any mention of artificial intelligence (AI) seems to be immediately followed by talk of how the technology is set to transform business as we know it: it's going to make a $15.7 trillion contribution to the global economy by 2020, it's going to boost economies by up to 26 per cent, businesses will be spending $78 billion on AI software by 2025.
With such game-changing numbers attached to AI's future, it's easy to lose sight of the important and useful ways the technology is already being deployed in numerous industries, including payments: analyst group IDC predicts that financial services will spend $5.6 billion on artificial intelligence this year alone.
Identifying fraudulent transactions is perhaps one of the most common uses of AI in the payments industry.
Thanks to the growing volumes of digital payments, along with an increased interest in, and ability to gather data, the payments industry has never held so much information on its users and the transactions they make. However, that rising tide of data also means that signals are at risk of being drowned out by the noise – human agents who may have once manually reviewed data to separate the fraudulent transactions from the merely unusual simply won't be able to keep up with all the information heading their way.
Artificial intelligence and machine learning are increasingly being put to work identifying and preventing fraudulent payments: use of such technologies is predicted to grow at a rate of 40 per cent and be worth $25 billion by 2024.
AI systems can block suspicious transactions by analysing and weighting huge amounts of data around a transaction – from its location to local spending trends to the payee's history – to spot payments that shouldn't go through. And, thanks to the amount of information that's being generated, AI systems can be trained on reliable real-world data, making them not only far more accurate but able to spot emerging trends and future concerns far more quickly than human operators.
While simple algorithm-based risk scoring could previously be used to spot non-legitimate transactions, they were something of a blunt instrument, relying on analysis of previous behaviour to monitor current levels of fraud. AI, however, can learn in real time, meaning the time taken to find and block new types of fraud is much reduced. And, once AI systems have identified fraudulent transactions, they can also be set up to reverse or stop payments at speed – cutting losses for payments companies and frustration for customers.
Another way AI is helping the payments industry is by improving customer service: AI is increasingly taking over tasks once carried out by human customer service agents. While financial services companies may have initially decided to roll out AI-powered chatbots as a way of effectively extending their workforce at relatively low cost and making sure customer service was available 24/7, customers have warmed to chatbots' availability, responsiveness and speed of action and are now happy to accept a first contact with a chatbot rather than a human agent.
AI-powered chatbots are being deployed for a number of roles, including answering customer queries and account management. While chatbots were initially used to answer simple questions – such as what's my balance, where's that transaction from, where do I find information on this topic? – they're now taking on far more complex tasks, including account onboarding and offering tailored financial advice.
By using AI rather than standard algorithms that identify keywords, chatbot systems can adapt to the infinite of variety of customers' language to develop appropriate responses, as well as learn how to deliver them in a way that sounds human rather than mechanical.
Credit scoring is another area where artificial intelligence is making its presence felt. Again, with such vast quantities of data now available on consumer behaviour, AI can build up a more comprehensive picture of an individual. The technology can dig out the relationships hidden deep in the data to predict how someone's creditworthiness will change over time. AI also broadens not only the volume of data that can be taken into account for credit scoring, but also the number of sources – social media, for example.
Further out, AI is likely to streamline how payments companies serve other industries, and their customers. Companies are piloting new ways of forecasting consumer demand for particular products, financial or otherwise. By identifying key characteristics and patterns around customer behaviour, payments companies can predict which services, such as insurance or an overdraft extension, customers may need, but can also provide information to other industries to help forecast changes in consumer demand.