Design and Evaluation of an AI-Based Anomaly Detection Framework for Strengthening Accountability in Zambia's Social Cash Transfer Systems

Zambia's Social Cash Transfer (SCT) programme operates across geographically dispersed and institutionally complex delivery settings in which manual identity checks, periodic audits and static rules are insufficient for identifying evolving irregularities. This study designed and evaluated an Artificial Intelligence (AI)-based anomaly detection framework that combines biometric verification, National Registration Card (NRC) optical character recognition, liveness checks, offline processing, server-side fallback, anomaly scoring and human review. A Design Science Research orientation was embedded within a pragmatist mixed-methods case study. Quantitative evidence covered approximately 71,000 beneficiary households processed during the September–October 2024 Emergency Cash Transfer cycle in Kitwe, Ndola, Chililabombwe and Solwezi. Qualitative evidence comprised stakeholder interviews and focus group discussions used to interpret operational, organisational, ethical and social conditions. The framework achieved a weighted on-time verification rate of approximately 95.9% and an offline success rate of approximately 91.0%. It flagged 1,275 suspicious records, of which 959 were confirmed as irregular after review, giving an alert-confirmation proportion of 75.2%. Average verification time fell from 8.5 minutes under manual NRC verification to 3.1 minutes for offline biometric matching, while the reported error rate declined from 3.90% to 0.18%. Device heterogeneity, intermittent connectivity, data-protection concerns, beneficiary trust and limited AI governance remained material constraints. The study concludes that AI can strengthen beneficiary integrity, operational efficiency, auditability and risk-based oversight when it is implemented as transparent decision support rather than as an automatic fraud determination. Keywords—Artificial intelligence, anomaly detection, biometric verification, fraud detection, public-sector information systems, Social Cash Transfer, Zambia.