PA Images via Getty Images
There are few bigger targets for cyber criminals than credit card companies. With trillions of dollars flowing through their systems, they’re tempting targets. Which is why the U.S. alone had over 270,000 reports of credit card fraud in 2019, double the 2017 rate.
So what’s a credit card company to do?
Use artificial intelligence to sniff out fraud and block it.
“We believe at American Express that we have the world’s largest and most advanced machine learning system in the financial services industry,” American Express’ VP of risk management Anjali Dewan told me recently on the TechFirst podcast. “And these models are ... monitoring 100% of these transactions and returning 8 billion credit and fraud risk decisions in real time.”
American Express provides 114 million credit cards in 106 countries which customers use to buy $1.2 trillion in goods and services annually. So that’s a lot of transactions and a lot of decisions.
The company has been working on AI fraud management for over a decade now, and is just about to release the tenth major model for monitoring and reducing fraud risk.
Watch the interview for this story:
The interesting part is that while AI is hot right now, American Express started using AI back in 2010, a full decade ago, and turned all of its risk management models over to AI in 2015.
That’s seriously early.
If you look at the Gartner hype cycle for artificial intelligence from 2019, for instance, you can see that technologies like AI cloud services, Machine Learning, Insight Engines, and Cognitive Computing were all years from the “peak of inflated expectations,” and therefore even more years removed from “the slope of enlightenment” and the much-longed-for “plateau of productivity.”
The Gartner hype cycle for artificial intelligence in 2019.
The conclusion, therefore, is that American Express was not only an early adopter, but also is well ahead of the curve in actually extracting value from AI. It’s already on the plateau of productivity, in other words.
“100% of our models are AI-powered and it cuts through the customer life cycle ... starting from new account origination, limit assignment, customer management, and fraud detection,” Dewan told me. “If you go by the Nilson Report, you’ll see that American Express for 13 years has come out as the lowest in the fraud space — and not by a little, but by half.”
That’s impressive.
Listen to Anjali Dewan on AI and fraud on the TechFirst podcast
And despite the fact that credit card fraud has gone up since Covid-19, Dewan says that the company’s fraud losses have remained very low.
This is important for multiple reasons.
Not only does fraud cost money when it is executed successfully — costs that are ultimately covered by a credit card company’s customers — but it costs when it fails. Customers who can’t complete purchases when they want to buy something but their card gets flagged for fraud in a false positive and therefore declined get annoyed, and perhaps embarrassed. Over time, that ads up to negative brand impressions, less satisfied customers, and potential customer defection.
Dewan says American Express has improved on both sides.
Anjali Dewan, Vice President, Risk Management at American Express
“Since 2014 when we deployed our AI solution for fraud, our resolution rate, digital resolution rate for fraud has improved by 100%,” Dewan says. “Point of sale disruption has been reduced by 21% since 2014, which is absolutely fantastic for our customers.”
Dewan is a little cagey on the technology hardware and software behind this feat, but did say that Amex is just about to release the tenth major iteration of its global fraud detection model, and uses both generative adversarial networks (GANs) and sequential recurrent neural networks (RNNs) to process risk decisioning.
GANs are important because they can upscale your model’s efficiency quickly: they are a species of machine learning frameworks in which neural networks compete against each other to improve, enabling optimization at machine speed. RNNs, or recurrent neural networks, are important because they don’t just process inputs and move on to other data; they store the results of that data while processing the next set of data, enabling more intelligent and nuanced decisions.
Ultimately, all that back-end technology has front-end customer-facing impact.
Just one example: when you fly to a foreign city and try to make a purchase (assuming we can ever do that again).
“What’s going on in the background is within 15 seconds our machine learning algorithms are initiating a personalized communication with this card member … which could take the form of an email, it could take the form of a push notification, it could take the form of a text,” Dewan says. “And that allows us to real-time communicate with the card member and resolve the fraud concern.”
There’s a lot more data involved, of course.
Dewan says that because American Express is vertically oriented — issuing cards, setting up merchants, and running its own payment network — it has essentially a platform advantage over its competitors: more data on which to base smarter decisions.
According to Dewan, it’s getting more sophisticated all the time.
“If you go back to pre-2014, we used to have 150 models to manage fraud detection across the globe, across countries, across our consumer and commercial portfolios,” she told me. “And the model as it stands today, is a single global model which has a view to the $1.2 trillion flowing through our networks, and a new model that’ll come out next month will have just the freshest data and very powerful new features.”
