Automation and Machine Learning in Payment Processing Systems:A Systematic Review

Abstract

The convergence of automation technologies and machine learning (ML) within payment processing systems represents one of the most consequential developments in contemporary financial infrastructure. This systematic review synthesises empirical evidence from 242 peer-reviewed studies, regulatory documents, and high-quality industry reports published between January 2015 and March 2026, following a PRISMA-adapted protocol applied to seven major academic databases and targeted grey literature sources. The review maps the full automation landscape across payment processing sub-domains—including ML-based fraud detection, intelligent transaction routing, robotic process automation (RPA) in compliance and reconciliation, natural language processing (NLP) in dispute management, and large language model (LLM) applications in regulatory compliance—and provides a critical synthesis of quantitative performance evidence. Key findings indicate that: (i) ML ensemble models achieve fraud detection AUC values of 0.97–0.997, outperforming rule-based baselines by 12–22 percentage points; (ii) reinforcement learning routing systems improve authorisation rates by 3–9 percentage points over static routing; (iii) intelligent document processing achieves KYC extraction accuracy exceeding 98%, reducing onboarding time from days to hours; (iv) AML alert precision improves from 1–2% to over 40% with supervised ML, reducing analyst workload by an order of magnitude; and (v) LLM-assisted dispute handling reduces per-case resolution time by 60–75%. The review further identifies eight critical research gaps—including real-time explainability, production-scale federated learning, intersectional algorithmic fairness, and quantum-resistant infrastructure—and proposes a structured research agenda. The sociotechnical implications of automation for payment sector employment, financial inclusion, and algorithmic accountability are examined with reference to the journal’s interdisciplinary mission.