Embedding Responsible AI into MLOps Pipelines: Ensuring Fairness, Explainability, and Governance in KYC and FinTech Decisioning

Abstract

Artificial intelligence (AI) has enhanced efficiency, scalability, and personalization in FinTech applications such as Know-Your-Customer (KYC) and credit decisioning. However, reliance on complex models introduces risks related to bias, opacity, and regulatory gaps. This paper presents an integrative review of 25 academic articles published between 2015 and 2025, synthesizing current knowledge on fairness, explainability, and governance (FEG) in MLOps pipelines. The review’s main contribution is a conceptual framework that highlights persistent implementation gaps and the limited operational integration of FEG principles in practice. While fairness interventions, explainability mechanisms, and governance structures can be embedded throughout the MLOps lifecycle, empirical evidence from production-grade deployments remains sparse. Key recommendations include monitoring bias across multiple stages, applying XAI tools such as SHAP, LIME, and counterfactuals, and strengthening governance through automated dashboards and audit trails. Overall, the findings emphasize that responsible AI must function not only as an ethical aspiration but as an operational imperative that fosters transparency and trust among stakeholders.