Machine Learning in Payments – an Overview in Disruptive Times

Daniel Faggella

Daniel Faggella is Head of Research at Emerj. Called upon by the United Nations, World Bank, INTERPOL, and leading enterprises, Daniel is a globally sought-after expert on the competitive strategy implications of AI for business and government leaders.

Machine Learning in Payments - an Overview in Disruptive Times

The coronavirus pandemic has ushered in a new era of digital payments; those who once mailed checks and made purchases in person are now paying their bills electronically and shopping online. As the economy is rattled by the coronavirus, there are some AI startups in the payments space that will succeed and others that will fail. All are pivoting rapidly to eCommerce, if that wasn’t already their focus to begin with.

Financial institutions will be looking for AI companies that can deliver faster, easier, more secure payments so that they can take on more business with less risk in this increasingly digital world. In this article, we explore our AI Opportunity Landscape research on AI innovation across the payments space in the following use-cases:

  • Accessing payment systems
  • Point of sale
  • Using payment data
  • Personalization of products and experiences

We begin our analysis of machine learning in payments with a brief overview and a discussion of how AI can be used to increase access to payment systems.

Machine Learning in Payments

The graphic below highlights the way we like to visualize AI impact in payments, across three discrete application areas:

AI Transformation in Payments Landscape

Below, we’ll explore each of the three application, fleshing out their use-cases in greater detail.

Accessing Payment Systems

There are a number of interface innovations that allow companies to better access payment systems. Voice recognition software is a potential use-case. It hasn’t been used widely for facilitating payments, but as voice-enabled search products become more popular, these applications will become more prevalent in the future. 

As more people opt for no-contact delivery and become more isolated, other conversational interfaces, such as text chatbots, will also likely increase in usage. Companies like 1-800-Flowers have been offering customers the ability to order via chatbot for years. 1-800-Flowers in particular has purportedly seen great success with their chatbot. Shortly after implementing its chatbot in 2016, 70% of customers ordering through the chatbot were new customers, according to CB Insights.

At Emerj, we believe that self-service payment software—that which helps customers serve their own needs and transact with brands—will become more popular with innovation departments once the dust settles and businesses can think about strategy again. 

Companies that are forced to be much leaner will still need to handle customer volumes and will need to consider these kinds of technologies for handling customer concerns. 

Similar applications include those that save customer preferences, such as personal finance and eCommerce applications that prompt a next purchase. These tend to improve customer lifetime value and create a smoother user experience for one buyer journey with a company to the next. 

When it comes to accessing payments, the startup ecosystem is leading the charge. While large credit card companies and established retailers are interested in improving the customer experience, startups and big tech are likely to be responsible for the more radical shifts and changes.

If voice recognition software becomes more popular in response to the pandemic, it will be because of Apple or Google. If text chatbots make a breakthrough, it will be because of some scientific natural language processing breakthrough at Baidu or Microsoft. If a new way of saving user preferences and prompting purchases becomes popular, it will be Amazon or some innovative payments or eCommerce startup that actually makes that happen. 

Point of Sale

During economic crises, payment fraud tends to increase. There is already a large increase in fraud amidst the coronavirus pandemic. During the last recession, chargebacks spiked. Consumers had to cut out expenses and in some cases used chargebacks as a way to do so. They were more likely to claim that a transaction they actually made was fraudulent so they could get their money back via the bank.

As such, during the coronavirus crisis, AI applications that can help detect a fraudulent payment or understand the footprint of a particular customer that has been known to charge back transactions may become more relevant. Payment fraud detection is one of the rare AI use-cases that has an immediate impact on a business in the short-term. At the moment, companies are less focused on moonshot goals and more focused on cutting costs and reducing risk; payment fraud detection will be a valuable application. 

Fraud detection companies like Sift may lose money if its customers go out of business, but its core application will be more valuable now than ever, particularly if the company is able to update its AI algorithms to calibrate for the onslaught of new fraud patterns and the prevalence of friendly fraud that is likely to occur in the economic downturn.

AI-enabled fraud detection vendors generally offer anomaly detection software, a type of machine learning that can detect deviations from the norm (in this case legitimate transactions). Below is a demo of Sift’s interface:

At the point of sale, innovation is much less limited to the startup ecosystem. Companies like Visa and MasterCard have great ability to develop fraud models given the volume of transactions that they process. Massive retailers and financial services organizations can do the same. Innovation in preventing and detecting fraud will be shared by existing established financial services and retail companies as opposed to accessing payments, which will bring much more disruptive changes prompted by the startup ecosystem.

Using Payment Data

Like accessing payments, there are a number of innovative ways to use payment data to improve the user experience or change business models. One use-case for payment data is in credit scoring.

Millennials and those belonging to Generation Z often lack a robust credit history because they don’t borrow nearly as frequently as their parents’ generations. As a result, their credit scores are often either low or nonexistent, which can make it difficult for lenders to know whether or not they are worthy of a loan or line of credit.

Some AI companies in financial services offer AI algorithms that use alternative data sources to determine a loan or credit applicant’s likelihood of replaying their loan or paying their balance on time. These algorithms may use information pertaining to the item the applicant wishes to take out a loan to purchase, such as a car, including its make, model, year, how often the applicant plans to drive the car, and the location of the dealership.

In some cases outside of the US, lenders are using AI to mine an applicant’s social media data and data on their phones (with the user’s consent) to determine their creditworthiness. LenddoEFL explains this use-case in the promotional video below:

Similarly, these algorithms could use an applicant’s history of debit card payments and bank account deposits as an indicator of creditworthiness, namely how often an applicant gets money deposited into their account and how much, as well as the kinds of items they purchase. Lenders might also use an applicant’s purchase history on eCommerce platforms, such as Walmart and Amazon.

This particular AI use-case is nascent due to the potential for breaking fair lending laws. Sometimes these algorithms correlate two seemingly unrelated data points in a way that indicates an applicant’s race or gender, neither of which lenders can use to determine whether or not to approve an applicant for a loan or credit line.

That said, lenders and the AI companies that serve them are likely to innovate with payment data given the history that younger borrowers have buying from eCommerce sites. In doing so, they will be able to offer younger applicants deals on loans and credit files without taking on too much risk.

Such an application may be especially important in the coming few years as lenders deal with the fallout of the coronavirus pandemic. With people out of work and no real end date for true normalcy in sight, it’s likely that people will hold off on making large purchases for the foreseeable future.

They will, however, be making a lot of purchases online compared to in-store. Lenders that can use machine learning to determine the creditworthiness of an applicant using their eCommerce data may be able to win new business as other lenders flounder.

For more on this particular AI use-case, listen to Emerj’s interview with Zest AI’s CTO: How Lenders Can Win More Business with Machine Learning.

Personalization of Products and Experiences

An AI startup ecosystem may develop for taking information from merchants or financial services firms about payments and users and enriching that data to further customize that particular user’s experience based on the behaviors of other users or the preferences of similar users that the startup’s algorithm has been trained on.

This ecosystem will become its own market of third-party data companies similar to Experian. Card-linked marketing is one way that credit card companies are making use of this data already. Card-linked marketing allows advertisers to show targeted ads to customers based on past purchases with their credit or debit card. 

Survival of the Startups

As the pandemic continues, there will be a lot of competition between companies. There is not enough room in the eCommerce ecosystem for all of the AI-enabled payment startups out there, and some will inevitably fade away as others gain market share, even if not revenue amidst the economic downturn. 

The companies and applications that come out on top will be those that can capture value, which may involve pivoting to eCommerce or finding new clients that are likely to survive the crisis. 

Companies selling essentials or doing food delivery, for example, may survive these times. In terms of defense, companies will try to “stop the bleeding,” focusing on risk mitigation and cost-cutting. With regards to payments, fraud detection will win the day, followed by improving customer lifetime value by personalizing customer experiences, improved churn prediction, and making better recommendations.

Emerj for Financial Services Leaders

Leaders at large financial services firms are considering how to innovate in the way they do business with their customers and collect payments. They work with Emerj to learn how they can leverage the AI applications for payment processing and fraud detection from which their competitors are already seeing an ROI.

Emerj AI Opportunity Landscapes provide financial services companies the research and data foundation they need to build AI strategies that will propel companies beyond the coronavirus crisis and win market share with new payment solutions that attract customers. Learn more about Emerj Research Services.

Stay Ahead of the AI Curve

Discover the critical AI trends and applications that separate winners from losers in the future of business.

Sign up for the 'AI Advantage' newsletter:

Stay Ahead of the Machine Learning Curve

Join over 20,000 AI-focused business leaders and receive our latest AI research and trends delivered weekly.

Thanks for subscribing to the Emerj "AI Advantage" newsletter, check your email inbox for confirmation.