Stay in the loop
Sign up for event alerts, as well as insights from our experts.
The global volume of online payments will increase by almost 11% a year between 2015 and 2020 . In emerging Asian markets, payment volumes are expected to grow by as much as 30.9% .
In Europe, the passage of the liberalising second payments directive (PSD2) and the advent of open banking oblige banks to allow regulated fintechs access to their online banking systems. This allows fintechs to develop new payment products for bank customers. And it’s expected to significantly increase the volume of electronic payments within the EU.
As payment volumes increase, payment service providers will no longer be able to rely extensively on manual processes. For everything from fraud detection to lead generation, the industry will rely increasingly on artificial intelligence.
The competition for customers
Propensity modelling is a statistical modelling technique that uses past behaviour to predict a consumer’s — or a consumer segment’s — likelihood to take a certain action. In marketing and customer relationship management, it’s used to forecast the effectiveness of products and promotions, and to model customers’ responses to changes in a company’s business model.
In an era of open banking, we can expect the competition between and among banks and third-party providers (TPPs) to offer new services to consumers to be fierce. Banks, no longer able to rely on a closed relationship with the consumer, will need to react faster and to place more emphasis on customer relationship management. Payment service providers, on the other hand, will need to move fast to establish themselves quickly or risk losing out in the battle for market share.
And this competition is likely to extend beyond simply marketing payment services. The close relationship a TPP has with a consumer and the data that yields, will make the TTP a highly valuable intermediary for other advertisers in other sectors. This is particularly true, as advertising itself moves to an AI-driven and programmatic future. If regulation allows the flexibility, this could be a profitable business extension for many TPPs.
But to realise these benefits, the payments industry will need to be able to run propensity modelling at scale and in close to real time. This will only be possible with the wide-spread use of AI.
AI and customer relationships
There’s little point in using AI to spot sales opportunities based on a customer’s future needs if your relationship with that customer is too poor for you to capitalise on the opportunity. Again, with the volume of transactions and the number of customers set to increase massively over the next few years, TPPs cannot afford to provide most of their customer care manually. It would cost too much to employ people to answer the expected volume of queries and, in any case, using people to answer routine questions is a waste of human capital.
Chatbots are already widely used in the financial industry to answer routine queries. Currently, the role chatbots can play in customer relationship management is limited. They can answer a relatively simple range of queries on a limited number of topics: “What is the balance on my account?”, “How much does this transaction cost?”.
Natural Language Processing (NLP) is a branch of AI based on machine learning. It allows computers to process and accurately understand human speech, learning as they go. There’s a lot of buzz around NLP, because Google, Amazon and other tech giants need improved NLP to power their digital assistants. The incentive for success is high, because whichever company has the most useful and human digital assistant, is likely to win the battle for customers as search goes voice based.
But NLP is also expected to power the next generation of customer service. As available processing power increases and bots are able to more intelligently and flexibly understand human language, the payments industry will shift to a bot-powered frontline in customer service.
Security and authentication
Another key challenge which AI can help to solve, is finding the right balance between security and ease of use. Ideally, merchants and customers want a payment system to be as easy to use as possible. This keeps the barrier to purchase low and allows the widest possible customer base to use the system.
Ideally, a customer would only need to perform a single action — for instance, swipe a fingerprint reader or take a selfie — to authenticate themselves and perform a transaction. The problem is that even biometrics can be compromised. For instance, if the database itself is hacked then a new fingerprint might be associated with your name in the bank’s records.
Until now, the answer has been dual-factor authentication. Consumers are left going through the tedious process of swiping a fingerprint and entering a password or a unique key-phrase. Using AI, however, payment providers will be able to establish robust baselines of what is normal for each customer. When a transaction conforms to this baseline, it’s low risk and can be allowed to proceed with lower standards of authentication. If the transaction stands out as an oddity, then more authentication is required.
Our future tending the machines
Within a relatively short time — no longer than ten years — the payments industry will be almost entirely AI driven. Intelligent systems will perform the onboarding and know-your-customer procedures, run payment backends, continuously scan millions of transactions for patterns that imply fraud and for future industry trends as well as sales and marketing opportunities. When customers interact with payment systems, in the first instance they will almost always do so through an artificial intelligence.
In this system, the role of humans will change. The industry will increasingly need the best AI developers and programmers, and will need to compete with other industries to get them. This won’t come cheap. A recent New York Times report found that the best AI developers are now earning up to $1 million a year.
There will still be a role for analysts and managers, but increasingly the specialists will need to understand and be able to work with AIs. They’ll need to be able to take informed decisions about how to commission their developers to construct and weight AI models, and take responsibility for the consequences their decisions have not just for the business, but for customers, and for society as a whole.
1. World Payments Report, Cap Gemini & BNP Paribas
Explore other insights
You might enjoy these other nuggets of wisdom.