AI
April 3, 2018

All too human: Expense report misconduct auditors miss

Josh Anish

Say your human auditor was tasked with analyzing every single one of your expense reports. What would she do? She’d examine line items, and manually cross-reference all receipt information on Google, TripAdvisor, Health Care Provider (HCP) databases, and dozens of other sources to validate their legitimacy. She would also check for mislabeled items, out of policy purchases, and duplicate expenses across employees and reports.

The problem is these manual expense report audits take days, weeks, or months depending on how many employees you have. AppZen’s AI platform takes all of your expense reports as they’re submitted and runs them through the exact same process, in seconds. Not only that: AppZen goes beyond human capability to look at expense behavior across time and employees.

Our internal data shows that manual, human audits consistently overlook the following recurring types of expense report misconduct.

Duplicate charges head most lists when it comes to expense report fraud monitoring, but the reason they’re so difficult for human auditors is that many instances happen over time.

For example, an employee might submit an expense in February and then (accidentally or not) submit it again in April. The human auditing process described above doesn’t even mention going back and combing all previously-approved reports; that would only add to the already time-consuming process. Cunning employees are well aware of this information gap.

Mileage padding is an especially difficult subset of expense reporting for human auditors. It’s a challenge to validate mileage claims on rentals or gasoline, especially if your company lacks GPS tracking capabilities. And while opportunistic employees likely can’t make a fortune padding their mileages, the threat of long-tail leakage for companies is worthy of concern.  

Along those lines, the AppZen AI auditing platform is seeing a lot of misconduct around Uber, Lyft, and traditional taxi services. The common pattern shows employees paying way too much for their rides. In other words, a ride that should cost $20 ends up getting expensed for $100.

Why does this happen? Almost certainly because the employee befriends the driver who then supplies the rider with something  -- drugs, cash, companionship -- in return for an exorbitant tip which is then picked up by the company via the expense report. A human auditor simply doesn’t have the bandwidth to correlate all Uber and taxi expenses with their mileages.

Unused airline tickets are becoming an area of interest for employees looking to defraud companies. The outlines of the scheme are easy to understand: the malfeasance begins when a ticket is booked for a business trip that subsequently gets cancelled or postponed. The employee then books and expenses a second ticket instead of paying much less to cancel or rebook on the first one, thereby handing him a free airline trip on the company’s dime.

Such avian misconduct might be easy to spot at a smaller company where it would be obvious that the employee simply didn’t get on the plane. But of course it presents challenges at a larger organization where the human auditors don’t have the bandwidth to investigate whether every flight expense is accurate.

Artificial intelligence can do the work of hundreds of human auditors by using natural language processing and machine learning to instantaneously scan every available reporting database, in seconds, to cross-check line items and flag instances of employee fraud. This straightforward evolution could save your company millions in T&E spend every year. Get started today.