{"id":135842,"date":"2025-11-15T05:20:08","date_gmt":"2025-11-15T05:20:08","guid":{"rendered":"https:\/\/www.newsbeep.com\/nz\/135842\/"},"modified":"2025-11-15T05:20:08","modified_gmt":"2025-11-15T05:20:08","slug":"ai-is-writing-performance-reviews-what-could-go-wrong","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/nz\/135842\/","title":{"rendered":"AI is writing performance reviews. What could go wrong?"},"content":{"rendered":"\u201cI\u2019d feel pretty bad if an AI just wholesale read everything that I wrote, looked at everything I did, and just assigned me a grade, and then my manager read the script at the end,\u201d Levick said. \u201cThat\u2019s the wrong way to do it, but I would probably also feel bad, honestly, if my manager only has a couple few hours to do this process and they spent all the time painstakingly manually reading the 2% of things they could get through and then wrote a review based on that very small subset.\u201d<\/p>\n
Bosses, like all humans, are prone to bias and faulty memory and may put too much weight on the recent past when reviewing a longer period, said Peter Cappelli, a management professor at the University of Pennsylvania\u2019s Wharton School and director of its Center for Human Resources. Cappelli said AI often can provide a more objective assessment than one you get from a human manager.<\/p>\n
That said, employees may consider a review to have less credibility if they know AI was involved. \u201cI get a reason to discount this and not pay any attention to it because I don\u2019t think it actually came from my boss,\u201d Cappelli said. \u201cIt said I did a good job. Does the boss really think that?\u201d<\/p>\n
Experimental studies have shown that when people are given AI tools, they often offload analysis to the technology and rely less on their own judgment. That can cut both ways in a performance review. Employees may feel processed rather than personally evaluated, deepening cynicism toward a review process many already see as performative, Cappelli said. And because chatbots are known to be sycophantic, an AI-assisted review may be overly positive and fail to adequately address performance issues (one of many reasons why it\u2019s a good idea to check any AI-written material for fairness, tone and accuracy).<\/p>\n
The pros and cons, plus the newness of AI, suggest companies will wrestle with their choices well beyond this performance review season. The decisions may be more clear-cut at companies now explicitly assessing how much or how well employees use AI, which is now part of the review criteria at companies including consulting firm KPMG LLP and e-commerce platform Shopify Inc.<\/p>\n
In his own role, Levick says he maintains responsibility for his opinions on employee performance, but uses AI to inform those opinions by helping him summarise lots of data, such as peer feedback and work employees submitted throughout the year. \u201cIt\u2019s in there in every step, but at the end of the day, the review is fundamentally from me and includes my thoughts as the primary driver,\u201d he said.<\/p>\n
Other companies are experimenting with tools designed to take on a bigger role in the evaluation process. Payroll platform Rippling, for example, is selling a system, called Talent Signal, that assesses new employees\u2019 first 90 days of output, labelling workers \u201chigh potential,\u201d \u201ctypical,\u201d or \u201cpay attention\u201d. The system analyses employees\u2019 performance on tasks, such handling customer queries or writing code, taking into account the complexity of the work. Managers can agree or disagree with a thumbs up or down.<\/p>\n
Some HR experts caution there\u2019s a limit to AI\u2019s usefulness in evaluating performance.<\/p>\n
Nora Jenkins Townson, founder of Bright + Early, an external HR department for tech companies, has been pitched by numerous startups claiming to have \u201cfixed\u201d performance reviews by lightening the load on middle managers and reducing bias, for example by filling out a weekly rating card or logging notes from regular one-on-ones into a system, which at the end of the year would generate something like a performance review.<\/p>\n
\u201cFor them to work well, you already have to have a solid definition of what success looks like, both at a company level and for each job,\u201d she said. \u201cYou still have to sit and do the thinking about what good performance means to you.\u201d<\/p>\n
Listen and subscribe to the Today in Business<\/a> podcast \u2013 the top headlines from the NZ Herald business team summarised and delivered by an artificial intelligence (AI) voice as an easily digestible recap.<\/p>\n","protected":false},"excerpt":{"rendered":"\u201cI\u2019d feel pretty bad if an AI just wholesale read everything that I wrote, looked at everything I…\n","protected":false},"author":2,"featured_media":135843,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[365,2851,28119,4765,9556,363,364,9442,17940,92918,3021,9436,23973,124,14134,92917,4484,3026,18957,29419,5838,9557,185,17534,5812,13102,92919,3025,111,139,69,6777,5204,17146,11596,9361,763,40249,23742,45056,145,6923,14905,1533,40866,64460,25152],"class_list":["post-135842","post","type-post","status-publish","format-standard","has-post-thumbnail","category-artificial-intelligence","tag-ai","tag-also","tag-amp","tag-annual","tag-artificial","tag-artificial-intelligence","tag-artificialintelligence","tag-better","tag-bosses","tag-botwritten","tag-bring","tag-chase","tag-cos","tag-could","tag-decision","tag-dreaded","tag-employees","tag-especially","tag-feedback","tag-go","tag-help","tag-intelligence","tag-is","tag-jpmorgan","tag-make","tag-managers","tag-meaningful","tag-most","tag-new-zealand","tag-newzealand","tag-nz","tag-performance","tag-process","tag-questions","tag-raises","tag-relief","tag-reviews","tag-seeking","tag-stands","tag-tasks","tag-technology","tag-what","tag-whether","tag-will","tag-worse","tag-writing","tag-wrong"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/135842","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/comments?post=135842"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/135842\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media\/135843"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media?parent=135842"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/categories?post=135842"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/tags?post=135842"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}