PSD3 fundamentally reframes AI's role in payments from a customer-service efficiency tool into a compliance and fraud-prevention infrastructure component. The regulatory framework, now approaching adoption across Europe, mandates stricter anti-fraud controls—including payee name matching, transaction blocking, spending limits and loss reimbursement for impersonation fraud—whilst simultaneously raising expectations for human support access and transparency around customer permissions. With fraudulent transactions across the EEA reaching €4.2 billion in 2024 and users absorbing roughly 85% of losses, the operational pressure to detect risk before, during and immediately after authorisation has become material. This creates a direct incentive for payment service providers and e-money institutions to embed AI deeper into their operating models, moving it beyond chatbot automation into real-time transaction monitoring, customer-risk assessment and compliance prioritisation. For CX teams accustomed to deploying AI primarily in support workflows, this signals a structural shift: customer service remains important under PSD3, but it is no longer the primary use case driving AI investment in the payments sector.
The critical implication for CX professionals is architectural, not tactical. Firms attempting to bolt AI onto fragmented systems—where transaction data, customer interactions, monitoring alerts and support histories sit in separate environments—will struggle to realise compliance and fraud-prevention gains. The institutions that will benefit most under PSD3 are those that tighten integration between payments, support, fraud monitoring and compliance workflows, treating them as a coherent operational stack rather than isolated functions. This raises a pointed question for teams already managing multiple platforms: if your Zendesk instance, fraud-monitoring tools and compliance systems cannot share context seamlessly, you are operating at a structural disadvantage when regulators expect faster, more accurate decision-making across these domains. The infrastructure quality question becomes as important as the AI model quality itself.
Equally important is the governance reality that underpins this shift. AI may accelerate monitoring and analysis, but responsibility remains with the institution deploying it. Record-keeping, internal controls around decision-making and escalation, and customer outcomes all remain non-negotiable under PSD3's more demanding retail payments environment. This means CX teams cannot treat AI as a black-box efficiency gain; they must understand how decisions are made, reviewed and escalated, particularly where AI recommendations affect customer outcomes in fraud disputes or account access. The transition from reactive, support-focused AI to proactive, compliance-embedded AI requires not just new tools but new governance disciplines—and those disciplines sit at the intersection of customer experience, operations and compliance.
From Customer Service to Compliance: The AI Revolution Under PSD3 The AI Journal