Can AI Outperform Human CEOs? Unveiling The Future of Leadership

To understand properly the role of artificial intelligence and its impact on leadership, especially in the corporate sector, we need to understand the fundamentals of generative AI. Similar to other types of AI, generative AI also poses an unsettling influence on human operations. Generative AI, functioning as an initial content creator, has the potential to enhance various roles by boosting productivity, performance, and creativity. For employees in various domains, generative AI can assist in generating first drafts, freeing up more time for them to refine content and explore new solutions. Software developers can then concentrate on tasks like improving code quality under tight deadlines and ensuring compliance with security protocols.

However, these transformations mustn’t occur in isolation. CEOs must recognise the impact AI has on employees’ emotional health and sense of professional identity. In the short term, CEOs must collaborate with their leadership teams and HR leaders to navigate this transformation within their organisations, which entails redefining employees’ roles and responsibilities and making necessary adjustments to operating models.

AI vs. Human CEOs: The Experiment

Generative AI has shown remarkable potential to exceed the strategic decision-making capabilities of human CEOs, particularly in data-intensive areas such as product development and market optimisation. In a simulated experiment focused on the automotive sector, AI models outperformed human counterparts in both market share and profitability. However, they struggled to manage unexpected disruptions, resulting in quicker terminations by the virtual boards. Although AI’s proficiency in analysing intricate data and making rapid iterations could transform corporate strategy, it falls short of possessing the intuition and foresight needed to handle unforeseen black swan events.

The experiment took place from February to July 2024. It involved 344 participants, including undergraduate and graduate students from universities in Central and South Asia, as well as senior executives from a bank in South Asia. They interacted with GPT-4o, a recent large language model (LLM) developed by Open AI. Participants engaged in a gamified simulation meant to mimic the decision-making challenges faced by CEOs, while various metrics were used to assess the quality of their decisions. This simulation served as a coarse-grained digital twin of the U.S. automotive industry, utilising mathematical models grounded in actual data related to car sales, market changes, historical pricing strategies and elasticity, as well as broader factors such as economic trends and the impacts of COVID-19.

Once the human participants finished their turn, the module transferred control to GPT-4o. Subsequently, it evaluated GPT-4o’s performance in comparison to four human participants—the top two students and two executives. The findings were both surprising and thought-provoking, prompting us to reconsider many of our beliefs regarding leadership, strategy, and the possible role of AI in high-level business decision-making.

The Result of the Experimentation

• GPT-4o excelled in its role as CEO, demonstrating outstanding performance across nearly all metrics when compared to leading human contenders. It crafted products with meticulous attention to detail, optimising both captivating automotive features and cost efficiency. Its adept response to market signals left its non-generative AI rivals anxious. At the same time, it generated such significant momentum that it outpaced the highest-achieving student’s market share and profitability by three rounds.
• The large language model (LMM) was ahead in areas such as product design and market optimisation. It could quickly process large data sets and provide valuable insights at a quicker speed than human optimisation.
• The result demonstrated that AI performance was superior in various areas, and it could be a game changer in corporate strategy, especially in areas where large data sets could be processed with superb accuracy and efficiency.
• There was emotional bias, like in human processing, and the decisions were logical and consistent.

Limitation of AI in Corporate Strategy Making

The AI faced challenges with black swan effects, such as unforeseen events like the market downturns seen during the COVID-19 pandemic. The LMM was designed to respond to these unpredictable shocks that could disrupt customer demand, crash price levels, and put pressure on supply chains. The most successful students implemented long-term approaches that included these factors. They steered clear of rigid contracts, reduced inventory risks, and approached growth with caution, maintaining flexibility in the face of changing market conditions. Their clear strategy focused on safeguarding adaptability rather than pursuing aggressive short-term profits.

Artificial intelligence can quickly learn and adapt within a controlled setting, which makes it less suited for handling extreme disruptions that depend on human intuition and foresight. Moreover, GPT-4o excelled in this experiment due to its access to comprehensive data from the simulator. Nonetheless, numerous companies struggle to produce sufficient data regarding speed, amount, accuracy, and diversity. Establishing a strong data infrastructure is crucial before introducing generative AI into high-level decision-making environments.

The Future of Leadership: AI and Human CEOs Working Together

The experiment’s results indicate that the future of CEOs may be characterised by a hybrid approach, where AI and human CEOs work in tandem. AI is particularly adept at handling data-intensive tasks, allowing companies to streamline operations, create superior products, and make quicker, data-informed decisions. Nonetheless, human CEOs will remain essential in areas that demand emotional intelligence, intuition, ethical reasoning, and a focus on long-term goals.

CEOs must thoroughly evaluate the appropriate timing for making such an investment, balancing the potential drawbacks of acting prematurely on a complex project that currently lacks the necessary talent and technology against the risks of being left behind. Presently, generative AI is constrained by its tendency for mistakes and should primarily be deployed in scenarios that can accommodate significant variability. Additionally, CEOs should explore new funding options for data and infrastructure—considering whether budget allocations should be sourced from IT, R&D, or elsewhere—if they conclude that tailored development is both essential and time-sensitive.

CEOs must recognise AI’s influence on employees’ emotional health and professional self-concept. In many cases, enhancements in productivity are mistakenly associated with a decrease in overall workforce numbers, and AI has already heightened anxieties among staff.

Therefore, AI’s effects represent important issues concerning workplace culture and personnel dynamics, necessitating collaboration between CEOs and HR to comprehend the changes in roles. As AI projects are implemented, it’s essential to conduct regular assessments to gauge employee attitudes. Additionally, CEOs should establish a clear change-management strategy to facilitate employees’ acceptance of their new AI colleagues while ensuring that staff maintain their sense of autonomy.

5 Key Questions for CEOs to Adopt Generative AI

1. How will existing roles and responsibilities change because of generative AI?
2. How should CEOs organise different departments for efficient collaboration with AI?
3. What new skills and talents will achieve long-term advantages?
4. Should CEOs upskill, reskill or resize the workforce?
5. How can CEOs encourage a culture where humans and AI collaborate as colleagues?

Conclusion

Corporate Organizations require policies to ensure that employees can safely utilise generative AI, limiting its application to scenarios where it operates within recognised parameters. While fostering experimentation is vital, it’s equally essential to monitor all such activities across the organisation to prevent unauthorised “shadow experiments” that may jeopardise sensitive information. Additionally, these policies should clarify data ownership, implement review procedures to avoid the dissemination of inaccurate or harmful content and safeguard both the business’s and its clients’ proprietary data.

CEOs should look forward to the future and invest in AI technologies, developing robust data infrastructure and a trained workforce. However, AI cannot replace the human touch, and data owners must thoroughly review all content generated by AI.

Ready to embrace AI in your strategy? Contact us!