This paper was presented at the ACM International Conference on AI in Finance 2025 in Singapore. The version uploaded here has been further updated for submission to the Electronic Journal of Information Systems in Developing Countries (EJISDC). ABSTRACTWhy do similar AI technologies drive innovation in some African fintech SMEs yet produce stagnation in others? This comparative case study examines two Nigerian fintech firms, XchangeBox (successful adopter) and Payrep MFB (struggling adopter), both using comparable AI tools for credit scoring and risk assessment. Drawing on 24 semi-structured interviews and 45+ internal documents, thematic analysis revealed three paradoxical tensions: efficiency versus exploration, empowerment versus deskilling, and speed versus governance. At XchangeBox, AI functioned as a cognitive collaborator, enabling 42% efficiency gains, three new product launches, and preserved relational capital through structured work rotation, explainable outputs, and embedded governance. At Payrep MFB, AI served as a replacement, achieving 55% processing efficiency but suppressing exploration, eroding manual skills by 34%, and generating unresolved bias complaints. The findings demonstrate that leadership framing, learning culture, employee participation, and governance infrastructure determine whether paradoxical tensions resolve toward innovation or stagnation. This study extends paradox theory to African fintech, contributes comparative evidence to the automation–augmentation debate, and offers actionable guidance for managers and policymakers navigating AI adoption in resource-constrained environments. Keywords: Artificial intelligence; paradox theory; fintech; Nigeria; organizational learning
Oladeji et al. (Wed,) studied this question.