As AI agents increasingly handle customer service interactions, it’s natural to ask: How do users respond when a service failure is caused by an AI agent versus a human one?
A study by Yibo Xie, Zelin Tong, and Zhuorong Wu explored just that. Through a series of experiments, where participants were asked to imagine that they had received a bouquet of wilted flowers, the authors found that customers were more dissatisfied when they perceived that the service failure had been caused by human agents than by AI ones.
Why? It seems we hold humans to higher standards. When humans fail, customers are more likely to engage in counterfactual thinking—that is, imagining how things could have gone differently if the agent had acted otherwise. In contrast, when an AI fails, we’re less likely to imagine it could have done better. That is, It’s not that AI is better at handling failure. It’s just that we don’t expect as much from it and, so, we let it off the hook more easily.
But there’s a twist.
Empathy influenced customer satisfaction with the human agent. Namely:
“When the empathy level of customer service is low, customers have a lower level of satisfaction toward human service failure than AI service failure. When the empathy level of customer service is high, customers have a higher level of satisfaction toward human service failure than AI service failure.” (p. 12)
This insight suggests AI agents may be more acceptable in interactions that don’t rely heavily on emotional nuance. But in situations where empathy is required, following a failure, humans may still be the safer bet.
This research is reported in the article “Artificial Intelligence or Human Service, Which Customer Service Failure Is More Unforgivable? A Counterfactual Thinking Perspective”, published in Psychology & Marketing.
Ana Canhoto 