An Experimental Study on the Implementation of ChatGPT : Investigating Employee Adaptation and Work Performance in a Customer Service Context

Sammanfattning

As generative AI tools like ChatGPT are rapidly introduced into the workplace, there is growing interest in how these technologies affect employee behavior and performance. While many organizations explore AI for customer-facing applications, fewer studies examine how such tools can support internal work processes. This thesis investigates the implementation of a custom GPT-based AI tool in the customer service department of a Swedish financial company. The tool, trained exclusively on internal documentation, was designed to assist six customer service agents over a four-week period. The study combines qualitative and quantitative methods, including pre- and post-implementation interviews, real-time observations, a TAM-based questionnaire, and performance metrics. The Technology Acceptance Model (TAM) provides the primary analytical framework, supported by organizational change theory and motivational perspectives. The results show that participants perceived the GPT as both useful and easy to use, especially for document-based queries. The tool supported faster information retrieval, reduced dependence on colleagues, and enhanced employee confidence. Positive attitudes toward the tool were shaped by both utilitarian benefits and emotional engagement. However, usage patterns varied, and trust in the system was closely linked to content quality and reliability. The findings contribute to a growing body of research on AI adoption in the workplace and offer practical guidance for organizations aiming to introduce similar tools. The study concludes that GPT-based tools are most effective as support systems, enhancing efficiency without replacing the human expertise needed for complex and nuanced tasks.