B2B Payments

AI: Expense Fraud’s New BFF

Removing friction from the payments experience isn't just about helping consumers spend their own money. FinTech is permeating the enterprise as more managers aim to ease their own employees' payment experience, from business trips to procure-to-pay.

It could be argued that it's easier than ever for an employee to spend corporate cash, and when it comes time to reconcile transactions or request reimbursements, it's also easy to make a mistake – or worse, commit fraud.

As it turns out, expense fraud is quite common, and costly. Research released in 2016 by Captio found that up to 12.8 percent of business travel expenses are potentially fraudulent. Analysts said the most common tactic to commit fraud was submitting old expense reports as new, or expensing transactions that occurred on a weekend.

Separate analysis from JPMorgan released in 2015 found that $1 billion of the $186 billion spent on business travel expenses in the U.S. was fraudulent; collectively, JPMorgan and Certify found that businesses lose a medium sum of more than $30,000 due to T&E fraud.

But, according to Patrick Taylor, CEO of Oversight Systems, it isn't just the financial aspect of T&E fraud that's worrisome.

"The reason you care about fraud isn't necessarily because the dollars are so big for any particular company," he said. "You're talking about a tenth of a percent, a one-hundredth of a percent of employees are committing fraud. It's a relatively small number. The reason you care isn't so much financial."

Instead, the bigger issue is the character of the employees themselves.

Taylor cited research from the Association of Certified Fraud Examiners, which found that more than three-quarters of employees who have committed expense report fraud are also engaged in some other form of fraud within the enterprise.

"They're not the kind of people you want working at your company," said Taylor.

But how can a manager separate out the workers who maliciously steal from their employers, from the ones who made an honest mistake on their expense reports?

According to the executive, it's all about separating "accidents from intent."

And one of the most efficient ways of doing that, he continued, is to identify repeat offenders. The problem with this strategy, however, is that the use of manual auditing solutions would force managers not only to review every single charge on every single expense report, but to somehow keep a record of each employee's past offenses to identify when they repeat a mistake or commit fraud more than once. Taylor said that it's an unreasonable method.

"That big green 'Approve' button is really easy to click, but it's a rather ineffective control," he said. "Most expense reporting systems have some kind of audit rule, and you can flag issues, but they are very black-and-white. Or, managers will use some kind of random sampling – sample 20 percent of expense reports and put eyeballs on them.

"Managers theoretically could find repeat offenders, but if they're clicking the 'Approve' button quickly, what happens most of the time is they're going to miss it," he added.

This is where data comes into play. Technologies like artificial intelligence and machine learning are extremely valuable in not only identifying instances of possible fraud, said Taylor, but in pinpointing individual spend behavior, and keeping track of past errors. The executive said recent innovations in FinTech and payments mean managers can gain the transaction data necessary to make the most use out of AI tools.

"We're detectives looking for clues, and we need to look for a lot of evidence and consider it in aggregate to identify the highest risk activity," he explained. "By looking at your historic data, we can learn a lot about what the norms of activity are, what unusual activity looks like, and the system learns what your corporate policy is."

Access to data, especially level 3 data provided by commercial card transactions, can not only identify fraud, but can also help companies benchmark their spending habits – from corporate travel to procurement to accounts payable – against their peers in other markets, Taylor added.

"The biggest thing we need is data," he said. "Whether it's a virtual card or mobile payment, if we can get a data stream, it gives you so many more clues and more potential places to look."

Behavior Change

Identifying fraud is all well and good, but what's even more crucial to managers and CFOs is the ability to halt repeat offenses before they happen. So while machines may be able to take over human labor of analyzing expense reports, there's a human element of fraud management that remains: behavior change.

"You have to be systematic about what you're going to do with the findings," said Taylor. "First is showing you the particular issue. But then you begin thinking about this from an employee perspective. The way to make things better is by education and changing employees' future behavior."

For Oversight Systems, which recently enhanced its abilities to identify repeat expense fraud offenders, that means automatically sending emails to offending employees - simple and light at first, but automatically bringing in their managers when offenses are repeated.

"If I send a generic email about corporate policy, the employee will blow it off," said Taylor. "If I send one that cites specific places where they did not comply with corporate policy, I've got a whole different effect."

Employees who are aware their transactions are being watched are significantly less likely to commit fraud again, the executive noted.

"The combination of knowing you're watched and understanding what all of the little paper cuts add up to on the expense report means most employees react in the way you want them to," he said. "I'm trying to fix behavior of the future."



The September 2020 Leveraging The Digital Banking Shift Study, PYMNTS examines consumers’ growing use of online and mobile tools to open and manage accounts as well as the factors that are paramount in building and maintaining trust in the current economic environment. The report is based on a survey of nearly 2,200 account-holding U.S. consumers.