Understanding customer needs is a central theme in product development since it provides the foundation for creating products that address real challenges that users face. In the agricultural machinery sector, this topic is especially relevant due to the introduction of more complex technologies and the competitive nature of the industry. Farmers today are not only required to adapt to new sustainability standards, but they also face equipment that has become increasingly digitalized and difficult to repair without technical assistance. For companies, this raises the question of how customer needs can be identified in a way that is both accurate and efficient. Currently, integrating artificial intelligence (AI) tools into product development has become a particularly relevant topic. Large Language Models (LLMs) such as ChatGPT and Gemini can generate interview-style responses, creating the possibility that they could partly replace or at least complement traditional research methods. However, little is known about whether such tools are capable of capturing not only surface-level feedback but also the deeper, latent needs of customers. This study addresses this gap by examining to what extent publicly available GenAI tools can replicate insights from semi-structured interviews in the context of the tractor industry. The research uses a qualitative comparative design. Four semi-structured interviews were conducted with five Austrian farmers (two of whom were interviewed together). Based on the demographic information of the farmers, AI-generated interviews were conducted using ChatGPT and Gemini. All transcripts were coded through a step-by-step process combining open and axial coding, followed by thematic analysis. This allowed the results to be compared across data sources, focusing on overlap, divergence, and depth of insights. The findings show that AI is able to determine many surface level customer needs, such as ergonomics and service availability but it lacked contextual richness, emotional nuance, and lived experience that farmers expressed in the interviews. It extends the research by Ronanki et al. by showing that AI’s limitations go beyond feasibility and ambiguity to also include cultural and situational gaps. It also reinforces the relevance of adoption models such as TAM and UTAUT, while suggesting that these frameworks need to account more explicitly for context-specific barriers. In addition, the study provides insights into Kool’s concept of spurious loyalty, showing that while AI can identify loyalty as a theme, only farmer interviews reveal factors such as dealer networks and peer influence that highlight the reasoning behind the spurious loyalty. Finally, it links to the JTBD framework by showing that while AI outlines jobs such as uptime and fuel efficiency, interviews are needed to determine the meaning behind the jobs. Despite this, the results suggest that GenAI tools cannot replace traditional qualitative methods of customer need identification, particularly when latent needs and lived experiences are critical. However, they can still provide value as a low-cost, early-stage tool. AI is useful for identifying broad themes, generating preliminary interview guides, or helping smaller companies explore trends when access to end users is limited. This study highlights the benefit of combining AI-driven approaches with traditional methods to achieve both efficiency and depth, rather than relying on AI alone. In practice, AI can be used when companies need rapid, wide-ranging insights, while interviews remain essential for uncovering depth and context. Future research should expand the sample size, test additional AI tools, and explore further applications of AI in product development.
- Innovation and Product Management
Human Derived vs. AI Derived Insights for Tractor Related Customer Needs
Hallemann, E. R. (Author). 2025
Student thesis: Master's Thesis