Does AI Know You’ll Quit?

Computers may now be able to predict if an employee plans to leave. Can—and should—firms use this tool?

It sounds like a dream tool for a boss in an Isaac Asimov science-fiction novel. An artificial intelligence program alerts you that five people on your team of 20 likely will quit over the next six months. That gives you time to determine which of the five you want to keep—and offer them something that will make them stay—and line up potential replacements for the ones you are willing to walk out the door.

That type of tool is now out there. Recently, news reports indicated that several firms have developed AI to predict, with a very high degree of accuracy, which of its employees are planning to leave. These so-called “predictive attrition” programs then suggest what managers should do to engage with the employees.

But what’s the proper use of a tool that knows what the employees are thinking before the actual employees do? Experts say that as advantageous as this technology may be, leaders may have to grapple with potential—and unforeseen—ethical and privacy concerns, says Ron Porter, senior client partner in Korn Ferry’s Global Human Resources Center of Excellence. “It could cross a line in the employee’s mind,” he says.

The quandary is one of many challenges as AI makes headway throughout HR. Many firms have focused on using some sort of AI to find new job candidates outside the company. Predictive attrition programs turn that on its head, using analytics to spot patterns within a company’s own workforce patterns.

Leaders might find that retention-focused AI a boon. An effective AI program could save millions of dollars in retention costs. But employees may find the program to be invasive, Porter says. After all, these algorithms analyze personal data, in part, to predict who might jump ship. Employees might be uncomfortable with their bosses anticipating their departure—especially if, he adds, the thought of leaving hasn’t even yet crossed their mind.

What’s more, how leaders handle the information could lead to some unintended consequences. “The main dangers of using machine learning to predict turnover lies in what you do with the data,” says Jared Shorts, an associate consultant with the Korn Ferry Institute. “An employer needs people with the appropriate educational background and experience to better understand patterns in the results and what to do with the findings.”

For example, a manager who’s indiscreet about why they’re engaging a flight-risk employee might send a signal to that employee that it’s time to start looking. Or leaders may start to view and treat flight-risk employees differently than other team members, investing less time and effort into their development. “There is some risk that the algorithm trying to solve turnover may potentially exacerbate it,” says Mark Royal, senior director for Korn Ferry Advisory.

Then there’s the question of accuracy. Most AI programs use information such as compensation, employee satisfaction scores, performance ratings, tenure, age, and job market demands to predict employee intentions to leave. The more data an algorithm has to analyze, the more accurate its predictions can be. However, that doesn’t mean the science is infallible, experts say. “An employer must realize that no model will perfectly predict voluntary turnover behavior,” Shorts says. “There are too many factors beyond the reach of the company that will impact an employee’s decision to leave.”

That doesn’t mean predictive analytics don’t have a place in HR. When used intentionally, these programs could complement good leaders. But leaders should view these algorithms more as mirrors than flashlights, experts say. In other words, instead of targeting individual employees, organizations could use the AI program to learn more about the conditions and situations that might cause employee turnover—whether due to cultural fit, ineffective leadership styles, misaligned incentives, or another issue.

Once leaders understand the patterns, they then can begin to address them on a structural level and improve the systemic issues that are causing people to leave in the first place. “It could help leaders understand employee turnover on a broad level,” Porter says. “It could be a guide.”