How AI Can Reduce Financial Fraud in Your Business

A close-up view of numbers and letters on a glowing black and blue data screen.

Artificial intelligence is already a force in many businesses, including the financial services industry. Firms are using AI to do everything from answering customer questions via chatbots to enhancing fraud prevention efforts. And it’s a win-win in many ways: 48% of executives believe AI will result in increased revenue and profits, 46% think it will create better customer experiences, and 38% expect to achieve cost savings, according to PwC’s 2019 AI survey.

AI robots can offer companies clear advantages. For example, they’re able to pore over massive volumes of data in a fraction of the time it would take humans to do so. Artificial intelligence can also help financial institutions maintain compliance with U.S. Treasury and SEC regulations, which are often updated to reflect new security requirements. This capability, among other things, can give businesses a competitive edge by reducing financial fraud and increasing efficiencies.

Here are some ways AI can help companies reduce fraud and related costs:

Process Transactions and Apply Rules

Mastercard can handle up to 160 million transactions per hour. Even at a smaller institution, the quantity of transactions—and required oversight—can be daunting. AI allows financial firms to not only manage the transaction flow but also analyze every touchpoint against a rubric of rules, and to do it instantly.

Track Anomalies in Card and Account Management

One of AI’s strengths lies in its ability to analyze data sets in ways that humans can’t. Typical fraud algorithms were simplistic: Was that transaction for a dollar amount that’s significantly more than that person typically spends? Did it happen outside of the customer’s typical geographic area of spending?

AI can go well beyond basic algorithms, pinpointing fraud patterns based on time of year, fraud trends, and user behavior on a variety of metrics. What did this customer's last 10 years of transactions look like? Is this customer exploiting referral codes to get multiple discounts or promotions for themselves? (One Uber rider used the company's promo codes to amass $50,000 in Uber credit.) This analysis is helpful for spotting anomalies in credit card and other account behavior.

Free Up Staff for More In-Depth Work

Although fear persists that AI will result in fewer jobs, the reality is that AI doesn’t have to eliminate jobs—and in many cases creates more need for skilled workers. AI can handle tedious, data-driven work, freeing up people in those positions to focus on other things and to achieve more for the company overall. When asked about the benefits of intelligent technologies in the future, 59% of employees say they’re excited by AI’s ability to reduce repetitive tasks, according to a recent Accenture study.

For instance, if fraud analysts don’t have to spend hours examining transactions, they can focus on other tasks and become more efficient. Employees will still need to oversee the technology, writing programs and tuning chatbot experiences. And in cases where AI handles routine customer care, human employees will have more time and concentration to manage complicated requests or situations that require more empathy.

Keep Rates Down for False Positives

In the same way that AI can analyze mountains of data for fraud patterns, complex machine learning can minimize false positives—customers who are wrongly rejected for possible fraud. For instance, consider a customer who goes on vacation abroad and has their card declined because they don’t typically make purchases in Paris.

False positives are frustrating for customers and they cost companies time and money to handle. AI’s capabilities include screening new clients for potential fraudulent risks and monitoring behavior and transactions for patterns that suggest fraud. AI can “learn” a customer’s evolving behavior over time and adjust rules accordingly, rather than applying a static set of rules to all of a customer’s transactions.

Detect Attacks Sooner

Fraud attacks are growing more advanced as technology evolves. For instance, when companies cracked down on card-not-present credit card fraud, thieves changed tactics to concentrate on account takeover and new account fraud. But AI is growing with the trend, and machines are better at spotting developing patterns before humans, making it easier to identify anomalies and strange behavior in the early stages.

AI does this using both supervised model and unsupervised model machine learning. A supervised model “learns” to identify new instances of fraud based on a large set of data provided by the company – i.e., some transactions are tagged as fraudulent, and others are tagged as non-fraudulent. The more data the company provides, the better trained the model will be. An unsupervised model, on the other hand, essentially teaches itself to detect fraud based on incoming data. This is useful for pinpointing new forms of fraud that a company hasn’t identified yet.

The sooner fraud attacks can be recognized, the sooner they can be stopped or dealt with, and the less money a company has at risk.

Keep Companies Compliant

Compliance requirements are frequently amended, forcing companies to take steps to ensure they’re always adhering to the latest version of financial regulations, including those that affect fraud detection. For instance, companies that provide automatic bank withdrawal payment options will need to make changes to tighten security measures by January of 2020 as required by NACHA, the Electronic Payments Association.

AI can aid in identifying new and applicable rules that apply to a firm and determining whether a company is fulfilling their obligations, helping companies avoid fines and save time overall.

Scale Treasury Efforts

AI will play an increasingly bigger role in tasks like checking balances and automating processes that are repetitive and potentially prone to errors. This will enable businesses to handle more processes and handle them more efficiently. Experts recommend businesses implement these changes slowly, starting with a procedure that’s less involved, and making sure they have a way to check outputs to ensure accuracy.

Bottom Line

AI is here to stay, and companies are taking note: 27% have already implemented AI in multiple areas, according to PwC’s survey, and 35% plan to deploy AI in multiple areas or enterprise-wide. As AI and machine learning continue to advance, the ways in which the technology can detect and block fraudsters will only expand the protections available to financial institutions.

The views expressed by the author are not necessarily those of Fifth Third Bank and are solely the opinions of the author. This article is for informational purposes only. It does not constitute the rendering of legal, accounting, or other professional services by Fifth Third Bank or any of their subsidiaries or affiliates, and are provided without any warranty whatsoever.