Abstract
The rapid adoption of AI-driven systems in digital commerce introduces a broader sociotechnical problem: how to scale efficiency without undermining the trust that enables transactions in the first place. In connected-commerce environments such as Zbooni, where business interactions occur through messaging platforms like WhatsApp, trust is not a secondary feature but a core mechanism that drives conversion and customer engagement. My technical and STS projects both address this shared problem from different perspectives. The technical project focuses on improving the efficiency and effectiveness of merchant acquisition through an AI-driven lead qualification system, while the STS research examines how AI-mediated communication reshapes the interactional foundations of trust. Together, these projects highlight a central issue: optimizing systems for scale and automation can conflict with the socially embedded processes through which trust is built. This problem is important because without trust, increased efficiency doesn't translate into meaningful business outcomes, ultimately limiting the effectiveness of AI integration in real-world systems.
My technical research investigated how to reduce inefficiencies in Zbooni's merchant acquisition process by developing a model-driven approach to reduce reliance on manual outreach and low-yield research, with limited visibility into which merchants are most likely to be onboarded. Currently, sales representatives spend a significant amount of time engaging with merchants who are not likely to be onboarded. This creates a misalignment between individual sales efforts and organizational performance goals. To address this, we designed a system that scrapes publicly available data on small businesses and applies a scoring model to rank potential leads. The methodology involved defining stakeholder-informed criteria for “high-quality” leads, constructing a weighted scoring framework, and evaluating how this system could function in a zero-baseline environment without historical data. The results demonstrate that a structured, model-driven approach can significantly reduce time spent on manual prospecting while improving the consistency and transparency of lead evaluation. More broadly, this project shows how systems engineering principles such as requirements definition, stakeholder alignment, and model-based decision-making can be applied to business process optimization in emerging digital markets.
My STS research examined how AI chatbots function as active participants in trust-building interactions within digital commerce environments. Drawing on Goffman's concept of the interaction order and research in human-computer interaction, I argue that trust is not simply a function of system performance but is constructed through socially meaningful cues such as tone, timing, and perceived authenticity. The evidence for this analysis comes from interdisciplinary literature on digital trust, as well as studies demonstrating that users respond to automated systems as if they were social actors. This research illustrates how AI chatbots reshape trust by altering the structure of interaction itself. While automation enables faster and more scalable communication, it can also remove or distort the social signals that customers rely on to evaluate credibility. As a result, users may experience interactions that are efficient but feel “off,” leading to hesitation or reduced trust. This finding reframes AI not as a neutral tool but as an actor within a broader sociotechnical system that actively influences how trust is produced and maintained.
Together, these projects contribute to addressing the broader problem of aligning efficiency with trust in AI-driven systems, but they also reveal important limitations. While the technical system improves the identification and targeting of high-quality leads, it does not fully account for how those leads are engaged once communication begins. Similarly, the STS analysis highlights the importance of interactional trust but does not provide a direct mechanism for integrating these insights into system design. My work demonstrates that solving this problem requires a more integrated approach that combines technical optimization with a deeper understanding of social dynamics. Future research should focus on incorporating trust-aware design principles into AI systems, such as modeling interaction qualities alongside conversion metrics or developing hybrid systems that balance automation with human involvement in key stages of communication. By bridging technical and sociotechnical perspectives, future work can move toward systems that are not only efficient but also capable of sustaining meaningful, trust-based interactions.
I would like to thank my capstone team for their hard work, valuable contributions, and dedication throughout the development of this project. I am thankful for the support and guidance of Professors Wylie, Burkett, and Wayland. I am also grateful to Zbooni for providing the context and motivation for this research, as well as to my friends and classmates for their collaboration and insight during my fourth year at the University of Virginia.