Imagine being able to scan employee e-mail for signs that top performers might be considering leaving the company, or for indications that unhappy workers are thinking of taking frustrations out on co-workers or the organization. HR executives might appreciate the ability to monitor communications of employees with access to confidential data to ensure that they’re taking proper security precautions when accessing that information.
Sound like Big Brother run amok?
Those are all actions companies are taking today through the use of “linguistic software,” which analyzes vast amounts of electronic communications for words or phrases that suggest employee intent to act in certain ways. The software has capabilities beyond common keyword searching or “rules-based” testing that track down use of words or number combinations in e-mail and text messages.
This text-mining software is used most commonly by financial firms to detect potential fraud or theft of intellectual property in their ranks. But experts say it has growing application for HR and in-house legal teams across industries.
Seeking Probable Cause
Taken together with other factors—the job title of those writing messages, the tone of the communication and where an e-mail or text is being sent—the findings from linguistic software can give companies “probable cause” that message creators are about to take—or have already taken—certain actions they’d like to stop.
“The software uses algorithms to evaluate combinations of words to improve the reliability that e-mail or texts are about a certain topic,” said David Meadows, director of the discovery consulting group at Kroll OnTrack, a legal technologies firm in Eden Prairie, Minn. “When you see certain terms used in combination by people in certain positions, for example, the risk of potential fraud can go up significantly.”
For example, phrases like “have to meet quota,” “let’s take this off line” and “stop by my office” could be signs that employees are under pressure to generate more money or are being secretive. If written with an angry or frustrated tone—“emotive” clues that some software packages can identify—or sent to personal e-mail accounts, it can present a collection of red flags that prompt further scrutiny.
Meadows recently worked on a case involving alleged theft of trade secrets that used linguistic software to help identify perpetrators. Employees in a company were suspected of taking confidential client information with them when they left to work for a competitor, so Meadows and his team were asked to examine the employees’ e-mail and text messages.
“We used e-mail analytics to look at how they were communicating with each other, and we were able to quickly discover they were trying to mask information and sending a lot of messages to personal e-mail accounts, said Meadows.
The software helped determine that employees were adding extraneous text to the beginning of e-mail messages, were using misleading subject lines—for example, “baseball game tonight”—and then burying client data such as buying histories and contact information deep in the body of those e-mail messages.
The software is being used for HR purposes, in some cases after it’s been purchased by IT or other departments for use. Some use it to monitor communications of employees with access to personally identifiable information to ensure that the data isn’t sent through unsecure channels or to protect against potential fraud. Others use it to keep abreast of employees in “pivotal” job roles for indications that they might be unhappy or considering leaving the company.
Vincent Walden, a partner with Ernst & Young's Fraud Investigation and Dispute Service practice, uses linguistic software with clients to analyze communication patterns for risks of employee misconduct. In one case, Walden assisted a stock brokerage firm by analyzing its e-mail and instant messages to conduct a "flight-risk" analysis to identify high-performing employees who may be intending to leave the firm. The company's goal was to intervene and take steps to retain these employees.
Most companies take a “risk-based” approach to using the software, Walden said, meaning that they use it only to monitor some employees.
“If you’re a Fortune 500 company, you might only have 300 to 500 employees with access to personally identifiable information, for example, or who could be in a position to leak other sensitive information,” Walden said. “You don’t want to boil the ocean with the software.”
Historically, organizations haven’t had much incentive to mine and monitor employees’ written communications unless it was as part of an internal investigation. Not only would that be time consuming and expensive, but reviewing large volumes of e-mail using basic key word searching generates a lot of “false positives”—messages that are perceived to be about one thing but are about another.
But improvements in linguistic software have made text-based searching more precise and reliable, according to experts. For example, the software can adjust for the way certain employees talk and the slang that they use.
“Linguistic software just keeps getting better and better, and I think it will become much more pervasive in business as a means to analyze and mine employees’ written communications,” said Alton Sizemore, director of investigations for Foreign Strategic Solutions in Birmingham, Ala., and a former 25-year employee of the FBI.
Sizemore said the software isn’t designed just to monitor senior-level employees suspected of white-collar crime. “You might have a warehouse employee stealing inventory and selling it to someone else, and the software could detect clues about that behavior in the employee’s communications,” he said.
Such employee monitoring raises privacy issues that employment attorneys said companies should address with policies. That’s particularly important for those operating in the European Union, where privacy restrictions are more severe than in the United States.
Teresa Thompson, shareholder and chair of the trade secrets group at Fredrikson & Byron, a law firm in Minneapolis, said most companies using linguistic software have employees sign basic agreements stating they have “no reasonable expectation” of privacy when using company e-mail and text messaging and the corporate network.
But given the more intrusive nature of the software, she suggests that companies take the additional step of having employees sign a separate form stating that they “consent” to the monitoring.
Thompson said that before they deploy the software, companies should understand the privacy laws of each state they operate in and should evaluate the Stored Communications Act and Electronic Communications Privacy Act, which provide protections for data in storage and in transit, respectively.
“The downside to using linguistic software from an HR or legal standpoint is that a company might obtain information it shouldn’t have if they are broadly monitoring communications,” Thompson said. “They might inadvertently obtain employee medical information that employees didn’t want anyone to know about, or they might discover other protected information.” Such discoveries could increase the chances of discrimination lawsuits, she said.
Dave Zielinski is a freelance business journalist in Minneapolis, Minn.