Experiment shows AI can perform a CEO's job better than humans, but it struggles in a crisis
AI can outperform CEOs in many areas but struggles with crises.
A recent experiment by Cambridge University researchers tested GPT-4o, OpenAI's advanced language model, placing it in the CEO role during a decision-making game. The study involved college students and senior executives, simulating real-world CEO challenges at a car company. The AI performed better than humans on most metrics including profitability, product design, and optimizing prices, offering an optimal value to consumers without personal biases influencing decisions.
However, the AI struggled with unforeseen black swan events, such as the unpredictable market changes observed during the pandemic. Unlike human participants who effectively managed unexpected customer demand shifts and supply chain collapses, the AI lacked the rapid adaptability needed to prevent termination by a virtual board due to these crises.
Interestingly, executives fared worse than students, illustrating overconfidence in systems valuing long-term thinking as much as ambitious strategies. The findings imply that while AI might not replace CEOs immediately, those not leveraging its capabilities risk being surpassed by competitors who do.