If You're Using AI for Layoffs, Read This First

Author: David B. Weisenfeld

Date: March 30, 2023

The job market may remain robust but the news about layoffs at big companies has been about as subtle as getting hit by an anvil - Amazon, Google, Microsoft, Salesforce, Disney and IBM are just some of the notable names that have announced mass layoffs in recent months.

But who is making these life-altering decisions about who stays… and who goes? For some employers, it just may be a software program. HR tech companies say algorithms in software remove human biases from layoff decision-making by rapidly crunching numbers on productivity and other metrics. Some of the companies mentioned above, and others, may well have relied on it in making their decisions.

No Foolproof Way to Fully Eliminate Bias

"The motivation is to eliminate human bias," said Littler employment attorney Britney Torres. "But is that actually what's happening? There is a lot of nuance involved."

Jonathan Reynolds, the CEO of Titus Talent Strategies - a recruiting firm that also uses technology to help companies make layoff decisions - does not necessarily disagree. "People decisions still need to be made by people," said Reynolds. "Data shouldn't make the ultimate decision."

A big reason for that, according to Torres, is due to unconscious or implicit bias on the part of programmers themselves which can, in turn, affect the decision-making process.

"We all have biases," acknowledged Reynolds. That's why, he added, it's important with any sort of head count reduction to use data with objective and quantifiable measurements. Reynolds notes that if the system is set up correctly, he can compare people based on their output.

But in the midst of difficult economic times, not all companies make decisions that way.

Plan Ahead

"When tough times come, we scramble to cut costs," said Reynolds. That often leads to cutting employees with the highest salaries during a reduction-in-force. But Reynolds warns against that approach because it may well mean losing your top performers. "Do it by performance, not salaries," he advises. "And don't wait until your lowest moment."

Instead, think about what data you would want to know to find out who needs to go, well in advance of any need for layoffs. According to Reynolds, that means asking pointed questions:

• Is the person doing what you hired them to do?
• What's the number one thing you would measure to decide if a person is performing well?
• How are you measuring good performance?
• Are you comparing people performing the same or similar roles?

Once you've answered those questions, Reynolds suggests, the software can better measure who performed and who didn't.

Monitor the Technology

For her part, Britney Torres said she is getting lots of questions about this type of technology from employers.

In fact, a poll of 300 HR managers late last year by Capterra, a unit of Gartner, found that 98% of HR leaders say they will rely at least partly on software programs or algorithms when deciding whom to let go in case of a layoff.

Numbers like those illustrate why Torres anticipates that the use of these programs will only continue to grow regardless of what happens with the economy.

But whenever using any sort of AI program relating to employment decisions, Torres stressed the importance of closely monitoring it and suggested three key tips:

"It's critical to maintain human involvement," said Torres. "Part of the testing should be whether the technology is decreasing or increasing bias." Without closely monitoring the software , she notes, disparate impact discrimination could result for certain groups.

The good news is the Capterra survey revealed that many in HR still believe in the human element too. Of the 98% whose departments will use software and AI to reduce labor costs, only 47% said they are completely comfortable making layoff decisions based on recommendations from their technology.

Regardless of how sophisticated the algorithms may be that feed AI technology, some leaders may not be wholly comfortable deferring people choices to AI. This provides them with an opportunity to think about how they might create an AI strategy that serves to inform tough employment decisions, such as layoffs, but not make them.

"You should never hide behind technology to make really sensitive people decisions," said Reynolds. "Data can provide support, but don't just press a button."