Research: Executives Who Used Gen AI Made Worse Predictions.

The article examines an experiment involving nearly 300 managers and executives who made stock price forecasts for Nvidia, comparing outcomes between those who consulted peers and those who used ChatGPT, a generative AI tool. Results showed that executives using ChatGPT became more optimistic and overconfident in their predictions, ultimately making less accurate forecasts than before, while peer discussions led to more conservative and improved predictions. The study identifies factors such as AI’s reliance on historical trends, authority bias, lack of emotional caution, and the social dynamics of peer groups as explanations for these differences. It recommends combining AI input with human dialogue and critical thinking to mitigate AI-induced biases and improve decision-making in executive contexts.