AI and chatbots stopping fraud

With the advent of mobile wallets, fast digital solutions and means of payment such as Swish and Apple Pay, more payments are being made than ever before – but the new payment methods open the doors to more cases of fraud. The banks are now using chatbots and AI to counter the threat.

AI-och-chattbottar.jpg

Rikard Lindgren is professor of informatics at the University of Gothenburg, and head of research at the Swedish Center for Digital Innovation. He can see great potential in how AI can be used to combat fraud. 

“Many banks today rely on analysis of digital footprints to identify fraud. Studying a user’s behavior can help you spot inappropriate use.” 

In the world of finance, an increasing amount of focus is being devoted to prediction, where AI is being used to detect non-conforming patterns in a user’s behavior. 

One example is MasterCard, which uses the Decision Intelligence security platform. This collects data from the user’s transactions and flags any behavior that indicates fraud.

AI identifies non-conformances

By learning how a real person makes purchases and performs transfers, it is easier for artificial intelligence systems to detect attempts at fraud. More users means more data, which provides the AI solution with a broader base to study.

One example from the Nordic region is Länsförsäkringar insurance, which is investing in digitalization and digital innovation to generate data about customers and understand them in more depth. 

“By offering benefits to progressive customers, the company is gaining access to data that allows prediction at a level far beyond anything achieved before,” says Rikard Lindgren.

The American company PayPal is another example of how using AI can produce results.  Only 0.32 percent of their profits are lost to fraud, compared to the industry average of 1.32 percent. These may seem like paltry sums considering PayPal handles payments totaling around USD 28 billion per month, but 0.32 percent of this means that fully USD 89.6 million a month are lost to fraudulent activities. 

Banks and finance companies that can reduce the level of fraud therefore have much to gain.

More complex flows in the future

Development is progressing quickly, and Rikard Lindgren believes that we will see more automatically generated information in the future.

“Around ten years from now, we’ll see a broader mix of people and machines. Companies’ understanding of their customers will be based on more innovative and creative data flows,” he says.

According to Rikard, banks will be able to collect data from sources other than purely financial information in the future. By looking at information on social media such as interactions and relationships, a bank will be able to collect new clues and pick out patterns – assuming that people post such data and that it is legal.

“This also applies to combatting fraud.” 

The data a person shares can be used to find signs that things are not as they should be. 

IBM is right at the cutting edge when it comes to helping banks combat fraud. The company has produced a number of solutions that use analyses to make it easier for banks to identify and combat money laundering and the financing of terrorism. Using the data to be found on databases and on users’ cell phones, IBM’s system can display information about users and predict whether they have any criminal intentions.

Chatbots helping

Chatbots can also be used to prevent fraud, although in a different way. A chatbot can be programmed to check for and identify suspicious activities and then notify the customer in question via his/her message service, such as Facebook Messenger or WhatsApp. In this way, a customer can quickly verify suspicious transfers and deal with potential problems before they develop.

A chatbot can also help if a customer does get into difficulties. For example, it can be programmed to notice if a customer is about to input his/her card data or other sensitive information, and then encrypt the conversation online so as to protect the customer's information.