Checking for fraud, AI rooted out one company’s sexist language

We hear a lot about AI’s potential to take away jobs, but in some cases, AI can do what humans simply cannot. Take the case of a fraud investigation that was tripped after a senior executive from an Australia-based retirement fund provider abruptly resigned and the company’s HR department found mentions of gambling in this person’s emails.

Advisory and investment firm KordaMentha using the e-discovery platformRelativity helped with the analysis of this executive’s emails and other data. Although there wasn’t any evidence of fraud, they did find evidence of sexist language. A deeper dive into the documents looked at how certain phrases were used in context. While the human team was reviewing, the third-party natural-language processing platforms analyzed that data with regards to sentiment, relationships, and entities.

“The analysis we conducted made it clear that the misogynistic language was happening throughout the organization,” said Craig Macaulay, executive director of KordaMentha’s forensic technology team, in a statement. What was more surprising is that the company recently conducted a survey on sexism in their workplace and they “passed with flying colors.”

For privacy reasons, KordaMentha declined to name the company or provide specific instances of the language in the emails. But KordaMentha said it triggered a second investigation of nearly 1.5 million documents. The additional analysis resulted in a handful of employees getting sacked, several others who got issued sanctions, and a plan to implement necessary organizational changes.