Should Machines Manage Employees?By Mike Elgan | Posted 2015-04-30 Email Print
Re-Thinking HR: What Every CIO Needs to Know About Tomorrow's Workforce
Algorithms can now identify disgruntled employees by predicting when they will cause problems, create scandals, harm the company's reputation or break the law.
JPMorgan Chase this month unveiled a new initiative to identify so-called "rogue employees." And they're doing it with software.
According to a representative for the company, dozens of facts will be entered into the system on each employee. The facts will include things such as whether workers have violated certain company or industry rules, or whether they've skipped any compliance classes offered by the company. The company is also planning to feed emails, chats and phone transcripts into the system.
Algorithms will crunch the numbers and then predict when employees will cause problems, create scandals, harm the company's reputation or break the law—anything that could result in fines, lawsuits or lost business. JPMorgan Chase is expecting the program to pay for itself many times over through savings from a reduction in lawsuits and fines.
The company is currently testing the system on employees involved directly in trading, but it plans to expand it globally to the asset management and investment banking divisions by next year.
A New and Dark Trend
Obviously, this is predictive analytics applied to human resources. But it's also the beginning of something new and dark: machine-based, automated surveillance of employees. It's a Minority Report-style attempt at pre-crime: stopping employee transgressions before they happen.
Financial institutions are natural early adopters on technology like this because their employees handle a lot of money. Investments happen very fast, it's an intense regulatory environment and IT spending per employee is already astronomical.
The Credit Suisse Group is using algorithms to figure out what factors predict which employees will quit. Their early results are not surprising: Employees who are most at risk of leaving include those with lousy bosses (that is, low-rated managers), workers who don't change jobs within the company and those who work in large teams. When the system identifies someone about to quit, the company's managers develop a plan to retain the employee—a plan that might include a promotion or job change.
But it's not just financial institutions that are taking this approach. Walmart, Box, Micron and other companies also use predictive analytics on their employees.
Google may have been the pioneer in the field of applying predictive analytics to employee behavior. They built their own algorithm more than six years ago to monitor a range of metrics to flag employees on the brink of quitting (and working for a Silicon Valley rival). The company's goal is to know which employees will quit even before the employees do.
Predictive Analytics as a Service
The field of HR predictive analytics is a huge growth area. The HR technology company Workday aims not only to flag employees who are about to quit, but also to provide sound recommendations about how to retain them.
The company uses data such as employee satisfaction surveys, promotions, hiring, salaries and job cuts. It found that interesting and varied work is the best way to retain the best employees.
Workday factors in not only employee behavior, but also Internet job postings and the cost of living in areas where employees live—or could live. The reason for that is brilliant: They've calculated that disgruntled or unhappy employees with the most marketable job skills are the most likely to leave, simply because they have the most options and are in a position to find other work.
The company intends to develop the ability to tell companies which employees will be most productive in the future. It also plans to create an interface for employees so they can see their own data.
Workday customers include Yahoo, Adobe, Symantec, Belkin, VMware and hundreds of others. The company is joined in this field by Ultimate Software Group, VoloMetrix and others.