feature image

Listen Up: NLP Can Help Detect Fraud Evidence in Business Data

Successfully preventing occupational fraud and fending off external threats generally requires businesses to analyze large volumes of data to uncover possible red flags. This already challenging task is made harder by the fact that approximately 80% to 90% of business data is estimated to be "unstructured" — in the form of emails, voicemails, and social media posts. Unfortunately, most companies lack the right tools and technology to process unstructured data. Instead, they typically rely on random sampling, a hit-or-miss approach that can fail to uncover fraudulent activities and transactions.

Natural Language Processing (NLP) may provide a solution. Still, in its infancy, this form of artificial intelligence can analyze vast amounts of data relatively simply. And, with a machine learning component, the technology is constantly improving.

Sophistication That Simplifies

For simplicity's sake, think of NLP as software that diagrams sentences. Using sophisticated algorithms, it reduces spoken and written sentences to smaller fragments. It isolates the role of words in a sentence and then establishes the words' meaning in the context of the larger sentence. This enables NLP to analyze unstructured data for tone and sentiment and extract signals associated with human behavior. These include signals that an employee plans to or has already committed fraud.

But evidence of fraud isn't the only thing NLP can dig up. By analyzing instant messaging between employees, NLP might identify a worker who's dissatisfied with her job. If you hope to retain the employee, you could use this information to address her concerns and entice her to stay. Or, if you fear the employee's attitude might make her more likely to commit fraud or sabotage, you could watch her closely and limit her access to certain company resources.

Versatile Tool

There are plenty of other ways your company could use NLP. For example, it could help you:

Protect intellectual property (IP)

Ensuring the security of your business's IP is critical. But what if an employee plans to jump ship for a competitor and take nonpublic information about your latest product with him? Before putting in his notice, the worker uses company email to discuss with a friend how valuable he believes your IP would be to other companies. Even though the employee doesn't explicitly say he plans to commit corporate espionage, NLP could flag the email and give you an opportunity to investigate.

Pinpoint internal control weaknesses

Employees often know a company has internal control weaknesses long before management recognizes and corrects them. Employees may even joke about control gaps with their coworkers. By analyzing their written and spoken communications, NLP may give you time to proactively address weaknesses before crooked individuals exploit them.

Stop employee collusion

Sometimes, employees and third parties collude to commit fraud. Take the example of a purchasing department manager who partners with a supplier to inflate invoices and split the proceeds. Their voicemail messages could reveal explicit details about the scheme, provide subtle hints that something is going on, or even just indicate an unusually close relationship. In all three circumstances, NLP might be able to flag these texts as suspicious.

Ensure regulatory compliance

NLP can help uncover potential regulatory violations before they attract government scrutiny. For instance, to comply with anti-bribery and anti-corruption regulations, companies aren't allowed to pay bribes. If you have overseas operations in countries where bribery is common, you may worry that an employee is making under-the-table payments. With NLP, you might look for evidence that he's cozying up to government officials or using questionable tactics to secure business.

Respond to customer feedback

If you receive a large volume of customer complaints, compliments, questions, and other feedback, NLP can help you sift through and separate out the messages that require responses. Such analysis might improve your company's ability to address recurring issues and improve customer satisfaction rates. It also can help you generate additional revenue. For example, say that a business owner complains that one of your products has fallen short of her expectations. You might offer her a replacement, ask her to participate in customer focus groups, and act on her suggestions for improvement. As you build greater trust and loyalty, she's more likely to purchase additional products from you and share her experience with other business owners.

No Substitute

NLP may enable your company to process unstructured data quickly and inexpensively and stop criminal activity before it leads to financial losses. We can help you customize software that will meet your company's complex needs. However, if you suspect fraud is already underway or that your company's internal controls aren't up to snuff, there's no substitute for professional forensic accounting expertise.

About the Author