At the beginning of January, the 2024 ZIBS Annual Academic Forum, themed Responsibility and Opportunity: Building an ESG-Driven Global Business Ecosystem, was successfully held in the International Campus. The forum delved into cutting-edge topics such as data intelligence, environmental science, fintech, digital innovation, and corporate renewal. It aimed to share the latest research findings, innovative ideas, and practical experiences, offering profound insights and strategies for the sustainable development of global business, technology, and education.
This issue of the ZIBS Insights spotlights the speech delivered by WANG Yiwei, ZIBS Assistant Professor. He discussed generative AI's influence on worker performance. Below is a summary of the key points.
Generative AI and Its Applications
Generative AI, which enables the creation of new content such as text, images, and videos, has become a transformative technology in business operations. Examples like ChatGPT and Gemini illustrate its capability to improve customer service quality and speed while reducing costs. This research focuses on the impact of Generative AI in E-commerce customer service operations, particularly in collaboration with Alibaba's Taobao platform. Generative AI systems learn patterns from large datasets and use that knowledge to create new, often human-like outputs. These systems use advanced neural network architectures, such as transformers, to predict and generate coherent, contextually relevant text. For example, ChatGPT can generate natural language responses, DALL-E can generate images, and StyleGAN can generate videos. These technologies are not only limited to customer service but are also widely used in marketing, software engineering, and product research and development.
Research Objectives and Methodology
The study aimed to quantify the impact of Generative AI on service speed and quality, understand mechanisms underlying its effects, and explore heterogeneous impacts on different user groups. The field experiment involved 6,000 human agents, half of whom had access to a Generative AI assistant. The analysis included metrics like service speed, issue identification time, and customer ratings. The research also detailed the interaction process between customer service agents and customers, including response time, number of messages, and typing time. The experiment setup included over 2.5 million agent-customer interactions and 0.38 million customer ratings. The agents involved in the experiment had less than 12 months of tenure, with most of them starting 4-5 months ago. The treatment group agents had the choice to either follow or disregard the suggestions provided by the GenAI assistant, while the control group agents had no access to the GenAI assistant.
Key Findings and Implications
The results showed significant improvements in service speed and quality. Specifically, customer service agents using Generative AI reduced issue identification time by 32% and chat duration by 4.2%. Customer ratings increased by 5.3%, reflecting higher satisfaction. Interestingly, agents using Generative AI became more engaged, indicating a positive co-production dynamic between human agents and AI. Furthermore, the study found that Generative AI has a more pronounced effect on low-performing employees, narrowing the gap between them and high performers. This finding is significant for understanding and optimizing performance management within organizations. For example, low-performing employees significantly reduced issue identification time and the number of messages in customer interactions, while high-performing employees improved service speed.
Challenges and Future Directions
While Generative AI offers substantial benefits, challenges such as reluctance to adopt AI suggestions and potential over-reliance on AI remain. Future research will explore strategies to enhance trust and efficiency in AI-human collaboration, as well as methods to incentivize and retain top talent. Additionally, the research will focus on the long-term effects of Generative AI on workforce structure and worker welfare, and how to improve Generative AI to accommodate worker adherence behavior. For example, how to increase employee acceptance of AI suggestions through training and incentives, and how to design more effective AI systems to reduce employee resistance.
Conclusion
Generative AI has immense potential to transform customer service operations by improving efficiency and customer satisfaction. This research provides valuable insights into the practical applications and challenges of integrating AI into business processes. As businesses continue to adopt AI technologies, understanding their impact on human performance will be critical to achieving sustainable growth. The study also noted that Generative AI has great potential to increase productivity but also requires attention to its effects on employee motivation and long-term corporate development. For example, how to balance the automation advantages of AI with the creativity of human employees, and how to maintain employee engagement while improving efficiency.
**This article is based on the speech made by WANG Yiwei at 2024 ZIBS Academic Forum. The views and opinions expressed in this article are those of the speaker and do not necessarily reflect the views or positions of ZIBS. Click the read more below to learn more.
WANG Yiwei is an Assistant Professor at ZIBS, and a researcher at Zhejiang University - International Center for Data Science. He obtained his Ph.D. from UC Irvine, Paul Merage School of Business, and his bachelor’s degree from UC Berkeley, department of Industrial Engineering and Operations Research. His research uses field experiments, causal inference, machine learning, and applied economic modelling to study consumer behaviors and operations problems in business operations. His research has been accepted at journals such as Manufacturing & Service Operations Management, Management Science and Service Science. He serves as the Ad Hoc reviewer for Management Science, Operations Research, Journal of Operations Management, Omega, and Decision Analysis.
