Artificial Intelligence-Driven Personalization in Open Banking: A Systematic Literature Review and Future Research Agenda
TL;DR
No existing review jointly treats AI-driven personalization techniques, the Open Banking data-sharing/regulatory layer, and generative-AI-era personalization as a unified research problem is identified as the field's central open opportunity.
Abstract
Background: Open Banking regulations such as the EU's Second Payment Services Directive (PSD2) have enabled artificial intelligence (AI) systems to personalize financial products and services using API-shared customer data, but research on this topic is fragmented across three largely non-overlapping literatures: personalization effectiveness, explainable AI (XAI) for credit risk, and Open Banking security/regulatory compliance.Purpose: this review synthesizes these literatures, maps the current state of the art, identifies technology trends, and proposes a ranked research agenda. Method: a scoped, rapid systematic review was conducted using the Consensus academic search engine across ten targeted queries spanning personalization, XAI credit scoring, Open Banking security, fraud detection, conversational AI, generative AI/LLMs, and prior systematic reviews; all candidate papers were filtered for direct relevance and independently verified against publisher or arXiv metadata, yielding fifteen papers carried into structured synthesis.Findings: the literature converges on a de facto personalization pipeline (segmentation, recommendation, post-hoc explainability, conversational delivery), with fraud detection and credit scoring the most mature, metrics-driven applications, and generative-AI-based personalization the least empirically validated frontier. Cross-cutting limitations include single-institution datasets, inconsistent personalization-effectiveness metrics, and a near-total separation between the Open Banking security/regulatory literature and the personalization/XAI literature.Gaps and agenda: ten unresolved research gaps are identified and ranked by novelty, practical impact, research difficulty, and publication potential; the highest-priority, most feasible direction is an integrated framework that jointly evaluates personalization effectiveness, explainability/fairness, and Open-Banking-style regulatory compliance, addressed here as a concrete research problem, research questions, objectives, and testable hypotheses.Conclusion: no existing review jointly treats AI-driven personalization techniques, the Open Banking data-sharing/regulatory layer, and generative-AI-era personalization as a unified research problem; this is identified as the field's central open opportunity.