AI, Finance, Fraud Detection, Operation

Using AI To Control Employee Expense Fraud

Darshana Daga

Reimbursable employee expenses are a significant cost for most businesses, and unfortunately, they are often a source of fraud by unscrupulous employees. To combat this, businesses typically rely on auditing employee expense reports manually. As this is a tedious and very labor-intensive process, usually only 10% of all the receipts get audited. Thus businesses have an increased risk of expense report fraud going undetected.

The Certified Fraud Examiner group analyzed over 2,700 global expense fraud cases in 2019 and estimated that the fraud exceeded over $7 billion in losses to the companies.

Shockingly all these companies have some form of manual audits in place, which have just been inadequate to detect all the employee expense fraud. Employees’ unauthorized and extravagant 4-figure lunches and nightclub-bliss disguised as “dinner” often go undetected. Due to a lack of sufficient resources, audits are carried out on only 10% of the receipts on a random basis. This allows many fraudulent, and sometimes even erroneous claims to go undetected. Some erroneous claims are just errors such as duplication or figure misprints during expense report preparation, but the resulting losses are very real.

Some savvy companies are now turning towards Artificial Intelligence (AI) to remedy this situation. There are powerful analytical tools that are powered and backed by AI technology that helps in the smart, rapid, and accurate screening of large volumes of expense report data. The turnaround is fast, and the AI-based modules of the tool can quickly pinpoint suspicious cases and tag them for further assessment by the auditors. The tagged cases are supported by the reason(s) why they have been deemed as worthy of further study by the AI-based tool.

The AI technology is sophisticated and does not merely tag outliers on basis of the amount of claim alone.  The AI modules have the capabilities to do semantic checks on the transactions with reference to taxes, general restaurant bill amounts (using online data), and other relationships. It can also take into consideration changes in the currency and identify potential duplicates, even when some of the details may appear to be different at first glance. And it can do all of this not just for the transactions on the company credit card, but also for the reimbursements claimed by employees who used their own funds for the expenses.

These AI-based modules gather data from internal data sources, online prices of comparable products and services, and other professional databases. Also, it takes into account the company’s policies and procedures. All of these capabilities work together to determine the outliers that need to be tagged for further studies, such as a visit to a strip club that is disguised as a client dinner.

The auditors only have to review the claims which seem suspicious or “anomalies” in AI parlance. A specific expense claim might not be fraudulent, but it seems like an anomaly to the AI-driven fraud detection tool based on its advanced algorithms that rely on all of the types of data as outlined above. After all, it is not the job of the tool to decide with finality whether an activity or expense is fraudulent, but merely to detect anomalies faster and more accurately than any human could, and to present them for further assessment.

These AI-based modules are integrated into the expense reimbursement process, and can automatically read the submitted receipts for analysis. The flagged receipts are shared with the auditors with easy-to-understand explanations as to why they have been selected.

This approach not only reduces the effort of manually auditing expense reports but also speeds up the process of issuing reimbursements to the employees. More and more organizations are adopting these new technologies to save millions of dollars in fraudulent expense claims and to improve employee morale by issuing reimbursements faster.

If you want to learn more about some of the AI-based tools available, please check out our Features pages displaying what NEMESIS has to offer.

Darshana Daga