Skip navigation
A very young Mojave desert sidewinder rattlesnake is seen shortly after dawn near Amboy Crater at Mojave Trails National Monument on August 27, 2017 near Essex, California. Photographer: David McNew/Getty Images

AI Can Now Catch Lies on Your Expense Report

“People don’t have time to look at every expense item,” says AppZen Chief Executive Officer Anant Kale. “We wanted to get AI to do it for them and to find things the human eye might miss.”

(Bloomberg) -- One employee traveling for work checked his dog into a kennel and billed it to his boss as a hotel expense. Another charged yoga classes to the corporate credit card as client entertainment. A third, after racking up a small fortune at a strip club, submitted the expense as a steakhouse business dinner. 

These bogus expenses, which occurred recently at major U.S. companies, have one thing in common: All were exposed by artificial intelligence algorithms that can in a matter of seconds sniff out fraudulent claims and forged receipts that are often undetectable to human auditors—certainly not without hours of tedious labor.

AppZen, an 18-month-old AI accounting startup, has already signed up several big companies, including Amazon.com Inc., International Business Machine Corp., Salesforce.com Inc. and Comcast Corp. and claims to have saved its clients $40 million in fraudulent expenses. AppZen and traditional firms like Oversight Systems say their technology isn’t erasing jobs—so far—but rather freeing up auditors to dig deeper into dubious claims and educate employees about travel and expense policies.

“People don’t have time to look at every expense item,” says AppZen Chief Executive Officer Anant Kale. “We wanted to get AI to do it for them and to find things the human eye might miss.”

U.S. companies, fearing damage to their reputations, are loath to acknowledge publicly how much money they lose each year on fraudulent expenses. But in a report released in April, the Association of Certified Fraud Examiners said it had analyzed  2,700 fraud cases from January 2016 to October 2017 that resulted in losses of $7 billion.

The world’s largest anti-fraud organization found travel and expense embezzlement typically accounts for about 14 percent of employee fraud. It has become easier to fool finance departments thanks to websites such as fakereceipts.us that make it easy to create a bogus paper trail.

For years, forensic accountants like Tiffany Couch, the founder of Acuity Forensics, have had to do the sleuthing one receipt at a time. In one case, she exposed $1.4 million worth of fabricated receipts; in another, Couch outed three auto parts executives who expensed thousands of dollars on a decadent weekend trip to Canada with their wives. But despite such successes, she says the advent of artificial intelligence is long overdue. “It’s an auditor’s worst nightmare to go through expense claim reimbursement,” she says. 

AppZen founder Kale, who has a background in finance and technology, created his firm after discovering how antiquated back office expense systems had become. Only about 20 per cent of claims were being scrutinized and in most cases auditors were just trying to match the amount on a receipt to the expense submitted, he says.

AppZen, which is based in San Jose, California, can audit 100 percent of claims in real-time by running receipts through an algorithm that hunts for duplication, discrepancies or inflated expenses. It reimburses legitimate employee expenses on the same day and kicks back any dubious claims to human auditors for further investigation. The algorithm can compare the average cost of a flight from New York to Chicago against the amount expensed and will flag it if the price seems exorbitant for that day (or if the employee upgraded their flight to first class). It will also sound the alarm if a company listed on a receipt doesn’t exist or if a strip club is masquerading as a steakhouse.

The algorithms have already exposed some creative—and costly—frauds: employees tacking on bottles of vodka to their “work lunch” bill, buying $3,000 worth of Starbucks gift cards and claiming it as “coffee with a contact.” One employee expensed her $900 office farewell party and submitted a claim that contained an animated photograph of her face instead of any receipts—demonstrating how seriously she took the auditors.

A number of Indian tech companies, including Wipro Ltd., have started offering AI-based fraud detection services similar to the one offered by AppZen. Oversight Systems, which is based in Atlanta, began using an early version of AI several years ago and its machine learning technologies now make it possible to scrutinize millions of transactions in real time.

According to Oversight, 30 percent of employee expense claims are risky, wasteful and potentially fraudulent. “You’ll be amazed at what people will try and do,” says Chief Executive Officer Terrence McCrossan.

Will artificial intelligence ever completely replace human auditors? 

Guido van Drunen, a principal in KPMG’s Forensic Advisory Services, believes some lower-level jobs will disappear as more and more companies adopt the technology in the coming years. But he says there’s no way AI can spot all the sneaky ways employees try to defraud their employers.

He recalls being called in after an employee expensed a live python. “Everyone was jumping up and down saying it was a fraud,” he says. Upon further investigation, van Drunen discovered the individual worked in sales and had bought the constrictor snake as a marketing ploy for the launch of a new product called Python.

While the snake could be justified, the employee’s purchase of $1,200 worth of steak as snake food could not. Pythons only consume live prey.

“The python was a legitimate business purchase,” van Drunen says.  “But the steak was for his family barbecues.”

Hide comments

Comments

  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.
Publish