Research: Executives Who Used Gen AI Made Worse Predictions

Harvard Business Review (Jul 1, 2025) by José Parra-Moyano, Patrick Reinmoeller and Karl Schmedders reports an experiment with nearly 300 managers and executives asked to predict Nvidia's one-month stock price after seeing recent prices. Half were allowed to query ChatGPT; half could consult peers. The ChatGPT group became significantly more optimistic and more confident and produced worse forecasts than the peer-discussion group. The authors attribute the gap to AI's authoritative tone and answer-detail producing a strong sense of assurance, unchecked by the social regulation, emotional responsiveness and useful scepticism that made peer discussion more conservative. Direct warning for SEA mid-market boards adopting AI for forecasting.