People are notoriously biased when it comes to writing performance feedback at work. Now human resources professionals are turning to artificial intelligence to help stamp this out.
More than half (52%) of managers give employees feedback on their personality rather than behavior, according to workplace language guidance company Textio’s Performance Feedback Report.
Beyond that, 39% of managers don’t use clear examples to illustrate their feedback. But it has an impact far beyond just an employee receiving a bad review. It has compounding effects on diversity, equity and inclusion when performance reviews are filled with gender, age and racial biases.
Compared to men, women are seven times more likely to report being described as opinionated, and receive 22% more feedback about their personality than men do. Meanwhile, Black and Latinx people report being described as passionate double the amount as others. White men under 40 years old get the word “brilliant” in performance feedback nearly nine times more than women over 40 years old. The list goes on, showing the importance of performance reviews and choosing the right words that will actually benefit the employee.
The boom in interest around how generative AI can be used at work by helping complete tasks quicker and fight the “blank page syndrome,” has led to HR departments using the tech to write job descriptions, and to help recruit new hires. Now, they want to test its use for improving performance reviews.
“Performance reviews are done kind of inconsistently,” said Tacita Morway, CTO at Textio, which is building its AI for performance review technology. “Many different organizations have many different ways of doing it.”
That inconsistency can lead to employees feeling unclear on what their next steps should be after a performance review. That’s why HR management softwares have begun implementing AI-assisted performance reviews. Textio has 150 organizations using its feature, including customers like Hulu, Spotify, Bloomberg, Duolingo, NBA, T-Mobile and McDonald’s. Meanwhile, Oracle released its new generative AI capabilities within its workforce management software HCM earlier this month.
Both are set out to make reviews more objective, improve consistency and fairness in the review process across teams and organizations, enhance personalized feedback, and address common shortcomings of humans in the review process.
Guy Waterman, global strategy lead and vp of people analytics, HCM technology and innovation at Oracle, said when he introduced this to their customer advisory teams, 89% said they are willing to be early adopters. That’s the highest percentage he’s seen for one of their new products, showing the interest from HR professionals.
“Many of our HR professionals dread performance-review season because it’s a lot of work,” said Waterman. “But we want to get them to focus and say ‘no, this is the opportunity where we can actually guide people and help them make the most significant decisions in their career.’”
Without products made specifically for HR professionals, they might turn to prompting ChatGPT, which would likely not give a helpful response. But with these systems, it already knows the relationship between the manager and the employee and other information around the work they’ve done.
“What we really care about is effective, high quality, bias-free feedback,” said Morway. “We use technology to help you get guidance along the way for the writer. Writing feedback is really hard and trying to understand it at scale, you can’t do it without technology. It’s giving you insights about where the issues are so you know how to go and support people.”
Textio’s product highlights things that managers should take action on. If you have feedback that has soft language, it will prompt you to be more direct instead. It also prompts managers if they aren’t sharing enough examples in the review, to avoid writing based on personalities. If a manager says that the person’s “attitude” is great, the AI bot will prompt them to consider saying the person’s “approach” instead.
“People who told me I was nice in a performance review probably thought they were giving me a happy day and doing me a favor,” said Morway. “Really they were holding me back by not telling me what they wanted to see from me. I was not growing. You don’t see the most of your team or help people bring the most to the table.”
Waterman says that he’s been guilty of just that, with nearly all of his performance reviews being four or five stars. But he now questions the fairness of providing such vague reviews.
Waterman emphasized that performance reviews take a lot of time, which is why implementing AI could be helpful. He has heard countless examples of managers with large teams doing a thorough job of performance reviews for say, the first three people, then running short on time so phoning it in for the remaining ones.
Oracle’s AI assist, which will launch by the end of the year, will help HR professionals use their time in better ways. Instead of looking at a blank page, it provides copy based on what’s in the system already. It’s a little different from Textio’s, which gives suggestions to the text you write yourself first.
“We don’t have to prompt you for the name of the employee again or that you’re looking for a performance review because we can infer that based on where you are in the system,” said Waterman. “We are using a baby bird system where we break it up into different parts of the performance review so that our customers get comfortable.”
Just like with all AI use cases, people worry that the human might be taken out of the picture. It’s one thing that is especially important when talking about people’s job evaluations and providing feedback. In both cases of Textio and Oracle, the AI assist is only supposed to be just that. Once the evaluation is created, it’s up to managers to still have an in-depth conversation with the employee about it.
“There will always be the human piece of management,” said Waterman. “The way we are using generative AI is an opportunity to document more accurately and consistently.”
“It’s becoming a go-to for people,” said Morway. “If they’re not using AI for performance feedback right now, they will be soon because they’re already using it for their job descriptions, their emails, and whatever thing that they’re writing.”