Can AI make performance reviews less biased?
Workers’ attitudes toward using AI in their jobs may be starting to sour.
They’ve been oversaturated with AI talk, have struggled to learn how to use the tools, and in many cases have found it’s actually adding to their workloads and making them less productive. But there is one area they would like to see more AI use in — their performance reviews.
Over 80% of employees think algorithms could give more accurate and fairer performance reviews than their managers, an October survey of over 3,000 U.S. workers from Gartner found. Another similarly-sized survey from Gartner found nearly 60% of workers think humans are actually more biased than AI when it comes to making decisions around compensation.
It comes as performance metrics are top of mind for employers bringing staff back into offices full-time and for others in hybrid working arrangements. Last week Meta CEO Mark Zuckerberg announced plans to cull 5% of its workforce, targeting cuts toward its so-called “lowest performers.”
While such cuts aren’t entirely uncommon, “too often, poor performance isn’t an employee’s failure but the result of deeper organizational challenges: unclear goals, a lack of alignment, or insufficient communication and manager training,” said Doug Dennerline, Betterworks CEO.
New AI tools have been designed to help make performance management more continuous and transparent, as well as less biased. At least in theory. And they can also take a traditionally time-consuming task off already busy managers’ plates, according to workplace experts.
According to Gartner’s report, HR teams first need to build a strong performance management foundation, regardless of new technology. “With a strong foundation, AI can enhance the process, but without a strong foundation AI will be a distraction at best and a liability at worst,” the report said.
The foundation of strong performance management starts with quality performance data. “AI requires quality inputs in order to produce meaningful outputs. Within performance management, the main input is performance feedback, but organizations often struggle to get employees to provide continuous feedback in performance management systems,” the report said.
Accordingly, HR teams should formalize expectations and processes for gathering performance feedback. They should also create systems to capture “invisible data” — which might include verbal interactions and behaviors. As a result, while AI can free up managers’ time spent on performance reviews, they’ll still have to be involved in the process.
“Managers will still finalize major decisions, as the human in the loop verifying and validating the bots’ recommendations. For more everyday activities such as in-the-moment performance feedback, bots are likely to take on an increasing share of managers’ tasks,” according to the report from Gartner.
Some experts raise concerns about AI eliminating any subjectivity and missing important nuances that help truly define an individual’s performance and contributions at work.
“While AI can help gather objective data, there is so much that should be considered in performance reviews,” said Mamie Jones, an organizational leadership expert.
Jones said performance goals and metrics can be split into two categories: skills goals and behavior goals. “The more nebulous of the two areas is behavioral. AI can assist in gathering skill and project outcome data from HR and development systems, but performance is about human behavior and how employees behave at work, under stress, or in a group setting,” Jones said.
“This requires human interaction and observation to adequately evaluate the employee in these areas,” Jones said. In dynamic teams and project environments in particular, it’s likely to fall short, she said.
The risk with using AI for performance management isn’t so much about accuracy, but quality, said Emily Rose McRae, senior director analyst at Gartner.
“It is going to produce something that is very bland, that is full of business speak and words that don’t have much substance behind them,” McRae said. “It’s not going to have what makes a good performance review. It’s not going to have explicit examples of behavior and actions. It’s not going to have explicit accomplishments listed in there. So a good AI tool is actually going to pull that information from other areas and collect it to really make managers’ lives easier.”