Soccer's video review system, known as VAR, was built to end arguments about referee calls. Instead it created a new kind of argument: not about what happened on the field, but about the review process itself.
A study that tracked more than 640,000 tweets during Premier League matches found that whenever the video referee intervened in a match, it produced a negative reaction from fans on social media, and this slump lasted 20 minutes on average. Separate research found fans report feeling in the dark due to a lack of communication regarding VAR decisions during matches. The technology got calls more accurate. It did not make people feel the process was fair.
The same pattern shows up in workplaces that add AI to decisions like hiring, performance reviews, or loan approvals. Researchers have a name for it: algorithm aversion, a psychological tendency where people distrust or reject advice or decisions made by algorithms, even when the algorithms outperform human judgment. People more quickly lose confidence in algorithmic than human forecasters after seeing them make the same mistake, even when the algorithm has already beaten a human at the same task. One visible error and the tool loses credibility, no matter how well it performed before.
This is not a theoretical risk. In 2024, Derek Mobley filed an employment discrimination lawsuit against Workday, alleging that their algorithm-based job applicant screening system discriminated against him and other applicants based on race, age and disability. A federal court has since granted preliminary certification, allowing the lawsuit to move forward as a nationwide collective action covering job applicants over 40 who were screened out by the tool since 2020. Separately, a University of Washington study found that massive text embedding models were biased in a resume screening scenario, with the models favoring white-associated names in 85.1% of cases and female-associated names in only 11.1% of cases, using resumes that were otherwise identical.
None of this means AI should stay out of hiring, claims, or review processes. A 2026 workplace survey found that 60% of executives now regularly use AI to support their decisions, and the trend is not slowing down. The mistake is treating AI as a way to remove judgment entirely rather than relocate it.
Some business decisions really are measurement problems: was the invoice paid, does the application meet the stated criteria, did the shipment arrive on time. AI is genuinely good at these and speeds them up without much controversy. Other decisions are judgment problems: was the rejection fair, was the penalty deserved, does this employee deserve a second chance. Those still need a person who can weigh context, not just data.
The real work for any company adding AI to its decisions is sorting its own processes into these two buckets, before a lawsuit or a public complaint forces the sorting to happen the hard way.