A recent study conducted by Harvard, Wharton, and MIT found that using generative AI improved the performance of consultants at Boston Consulting Group (BCG). The study, which is the first of its kind to explore the real usage of generative AI in an enterprise setting, utilized GPT-4, the most powerful large language model currently available. The researchers discovered that consultants who used AI completed tasks more often, more quickly, and at a higher quality compared to those who did not use AI. In particular, the lowest-performing consultants experienced the biggest gains from using generative AI.
The study also highlighted the challenges of deploying AI in business settings. It identified a “jagged technology frontier,” which refers to the difficulty in discerning which tasks can be easily performed by AI and which tasks are beyond its current capabilities. This frontier is constantly shifting as AI capabilities evolve, making it challenging for organizations to determine when and how to deploy AI. The researchers proposed two patterns of AI usage, referred to as “Cyborg” and “Centaur” behaviors, which involve combining the strengths of AI and human work to approach tasks with uncertain AI capabilities.
While generative AI has the potential to greatly improve workforce productivity, the study cautioned against overreliance on AI and highlighted the importance of human judgment. Users can sometimes “fall asleep at the wheel” and outsource their judgment to AI, leading to poorer results. Additionally, the study found that AI can lead to homogenization of ideas, potentially stifling creativity and innovation. To counteract this, the researchers recommended utilizing multiple AI models and increased human involvement to introduce diversity and variability in outputs.
In conclusion, the study reinforces the potential benefits of generative AI in enhancing the performance of consultants. It emphasizes the need for organizations to carefully consider when and how to deploy AI, given the constantly evolving nature of its capabilities. The research also highlights the importance of balancing AI usage with human judgment and creativity to avoid overreliance and homogenization.