Across payments, AI is often discussed as a customer-service tool with chatbots, automated replies and faster handling of routine queries. That view feels too narrow for what the market loosely calls PSD3. The wider PSD3/PSR package, now close to adoption, points to something more consequential. It raises the standard for fraud prevention, transparency, user control and service in ways that will push AI deeper into the operating model of payment institutions.
What makes this shift important is that the broader PSD3 framework is not about automation alone. The agreed framework strengthens the quality of payment operations through anti-fraud obligations, better customer protection, more transparency around permissions and access to human customer support. For PSPs and EMIs, that creates a more demanding environment in which customer service, fraud controls and compliance can no longer be treated as isolated functions.
The package raises the bar for operating speed and judgement
Under these new anti-fraud measures, PSPs will have to check that the payee’s name matches the unique identifier, refuse payment orders where there is a discrepancy, offer spending limits and blocking tools, and in some cases reimburse losses from impersonation fraud. The framework also expects PSPs to share fraud-related information, and receiving PSPs may need to freeze suspicious transactions. These adjustments increase the need for faster judgement across large volumes of payment activity, particularly where scam patterns change quickly.
That matters in a market where payment fraud remains material. In 2024, fraudulent payment transactions across the EEA reached €4.2 billion and users bore roughly 85% of credit-transfer fraud losses, largely through authorised push payment (APP) fraud or impersonation scams. In that setting, detecting risk before, during and immediately after authorisation becomes more important than ever, and that is one of the clearest reasons AI is moving closer to the centre of payments operations.
Customer support is no longer the primary use case for AI
Customer service remains an important part of the framework, but it also raises expectations around customer access to human support, not only chatbots. AI can support service teams very well by helping classify cases, surface relevant information and improve response times, but payment disputes, fraud events and account-access issues still require accountable human handling at the right point in the process.
The larger change is that the same technologies now being used in customer support are also being tested and deployed much deeper in the stack. AI is already most frequently used in areas such as customer support, client and transaction profiling, fraud and AML/CFT. In other words, the journey from service to compliance is already under way and PSD3 simply gives it a much stronger operational incentive.
Better models depend on better infrastructure
This is the point where many firms misread the AI opportunity. AI delivers the most value when payment data, customer data, case history and operational workflow can be joined coherently. AI and machine learning applied to large datasets can improve suspicious transaction monitoring in real time, support customer-risk assessment and help compliance teams prioritise higher-risk activity more efficiently. But these tools complement existing systems. They do not replace the need for strong architecture and controls.
For payment institutions, that means infrastructure quality becomes crucial. If transaction events, customer interactions, monitoring alerts, and support histories all sit in different environments, the institution is still operating across fragments systems. In that situation, AI may produce activity, but not always clarity. The firms that will benefit most under PSD3 are likely to be the ones that tighten the connection between payments, support, fraud monitoring and compliance workflows rather than trying to bolt AI onto a fragmented stack.
AI will sit inside a regulated control environment
The conversation around AI in payments cannot focus purely on speed or efficiency. Financial institutions still operate inside heavily regulated environments where customer outcomes and operational accountability remain critical. As AI becomes more deeply embedded into fraud monitoring, onboarding and compliance workflows, payment providers will still need clear internal controls around how decisions are made, reviewed and escalated. From a payments perspective, the practical takeaway is straightforward: AI may accelerate monitoring, prioritisation and analysis, but responsibility still sits with the institution deploying it. Record-keeping, governance and customer outcomes continue to matter, and under PSD3 they matter within a more demanding retail payments environment.
Over the next phase of European payments, the more significant change will happen behind the scenes, where PSPs and EMIs use AI to support fraud prevention, case handling, monitoring and compliance at a level of speed and accuracy that manual models increasingly struggle to maintain on their own. Under PSD3, AI will become less of a front-end feature and more of a component of the modern payment infrastructure.
