Share this content

How AI can help prevent office fraud

Feb 20th 2018
Share this content

For the past few years, people have been hearing how AI and machine learning will accomplish everything from writing articles like these to curing diseases to driving cars. There is lot of hype, and when there is a lot of hype there is a lot of overblown expectations. AI will likely reshape our lives, but in subtler ways than techno-junkies imagine.

Office fraud prevention is a perfect example of a field where there is not much prestige, but remains vitally important and where AI can make a huge impact. QuickBooks reports that “the typical organization loses a median of 5% of revenues each year due to fraud,” which translates to losses of approximately $3.7 trillion. Furthermore, smaller companies often suffer worse as they are less likely to implement anti-fraud procedures.

Technology and AI cannot stop office fraud by itself. But by analyzing the nature of office fraud, we can understand how AI combined with proper office procedures and human supervision, can play an important role in shutting such crimes down.

Evolving Fraud and Technology

The most important thing to understand about fraud is that it is constantly changing. Crooks, whether inside or outside the office, are constantly looking for new methods to defraud customers and businesses. Accountants can learn about one kind of fraud, implement procedures and warn employees to be on their guard against that fraud, but not be aware of a new kind of fraud until it is too late.

However, machine learning can be constantly improving and detect emerging fraud schemes. A machine learning system can track what is normal business online and what is abnormal. As the system focuses on behaviors and does not blindly follow the rules programmed by an accountant, it also decreases the chances of getting false positives and angering honest workers.

As a basic example, if employees routinely fly and put the costs of the flight onto their expense report, a machine learning system will learn to accept such reports as normal. But if an employee suddenly decides to fly first class, buy gold online, and put that much larger expense on their report, the machine learning system can flag it and alert you. The result is a faster process which more efficiently detect fraud.

Data and Expense Report Fraud

While machine learning can help detect new fraud methods and prevent them from damaging your company, AI can shut down older fraud methods such as expense report fraud. We have all heard of stories where employees “make a mistake” by overestimating their expenses, or even outright lie about what their expenses were. To make matters worse, expense report fraud is difficult to detect. The distinction between mistakes and fraud can be vague and accountants can miss things given the vast amounts of data we have to sift through.

But if there was anything AI were made to do, it was to sift through vast amounts of data and find anomalous results. As TechCrunch notes, machines that are not affected by the sleepless nights of accountants are less likely to miss a dodgy-looking expense. Furthermore, potential fraudsters will be deterred as they know they are more likely to get caught.

 The Role of Humans

Machine learning can detect aberrant and potentially criminal behaviors, and AI can sift through large amounts of data to catch expense report fraud. But as noted above, this does not mean that there will be no role for human accountants.

An important thing to note about AI is that even in fields which they perform better than a human, they can sometimes suddenly fail in catastrophic ways. A classic example of this is Watson, the Jeopardy AI which defeated humans and yet occasionally gave out embarrassing answers. Humans will be needed to cross-check artificial intelligence’s conclusions.   

Furthermore, human accountants will still be needed to set guidelines and parameters. To take the aforementioned example of employees flying, what would happen if all employees suddenly started flying first class and buying alcohol? A machine learning system might assume that is normal behavior, while a human operator could declare that it is not.

This does not mean that AI does not have a role in stopping office fraud. Machine learning systems can detect fraud even before it becomes a serious threat. But it is no substitute for basic anti-fraud policies such as implementing internal controls, setting a zero tolerance policy, and leading by example. AI and humans will work together to build a better future in every imaginable, and so we should not worry about one supplanting the other.

Replies (0)

Please login or register to join the discussion.

There are currently no replies, be the first to post a reply.