- AI agents move from advice to action – when it comes to payments, the consequences are immediate and can be difficult to reverse.
- Financial institutions have long built trust in knowing who is behind a transaction; Agentic AI adds a layer between the human and the payment that existing controls are not designed to handle.
- The future of trusted transactions will depend on understanding intent, authority and context, as well as establishing strong identity checks.
A small business owner asks an AI assistant to manage a simple task: checking outstanding invoices, checking which suppliers need to be paid, and preparing the next payments. The assistant scans emails, compares due dates, and prepares recommendations for approval.
This could eliminate hours of repetitive work for the business owner.
But for the financial institution that processes these payments, it changes the nature of trust. A payment may still be made from a legitimate account, but the decision behind it may have been designed, prepared or initiated by software acting on behalf of someone else.
This poses major challenges for financial services. It is no longer enough to simply understand who or what is behind a transaction; Institutions must now be able to recognize whether the action reflects genuine human or business intent.
Table of Contents
AI agents have moved from advice to action
This challenge becomes particularly important in payment transactions.
The International Monetary Fund has described agent AI as a development that could shift payments from human-initiated instructions to agent-mediated decisions. Payments are not just another automated workflow. When money moves, the consequences are immediate and can be difficult to reverse. Financial systems have long been based on the assumption that a payment order can be assigned to a person, company or institution with clear authority to act. Agentic AI complicates this assumption by adding another layer between the human and the transaction.
This does not mean that AI agents should be kept away from financial activities. If used carefully, they could help people and businesses complete routine financial tasks more smoothly. But as these systems become more useful, it will be more important to define the conditions under which they can function. The question is not whether AI agents should be used in finance, but rather how institutions can support their use without weakening accountability.
In 2026, Santander and Mastercard demonstrated Europe’s first live payment made by an AI agent in a regulated banking environment. The transaction was completed using pre-authorized customer permissions, tokenized credentials, and existing bank controls, demonstrating that autonomous agents can operate within, rather than outside, established regulatory and security frameworks. Similar initiatives announced by BBVA with Visa and Nordea with Mastercard suggest that financial institutions are increasingly exploring how AI agents can act on behalf of customers while maintaining the governance, authentication and auditability expected of regulated financial services.
Trusted transactions require more than just identity
An agent that can act quickly across multiple systems can also make mistakes quickly. It can be manipulated through false information, compromised instructions, or fraudulent requests that appear legitimate. In a payments environment, the difference between helpful automation and harmful activity can depend on whether institutions can understand intent, authority and context before money is moved.
Today, many financial controls focus on identity. Who is the customer? Is the account legitimate? Does the transaction conform to expected behavior? These questions remain important, but may no longer be enough. When an AI agent acts on behalf of a person or company, institutions also need to understand whether the action reflects a genuine instruction, whether it falls within the agent’s permissible role, and whether there is sufficient evidence to explain why the transaction occurred.
Here, trust in agent financing must be carefully designed. Stronger authentication will be important, but clearer audit trails will also be important. Financial institutions need to know when an AI agent was involved, what it was authorized to do, and whether its actions resulted from a legitimate human or business decision.
Explainability will also become more important. When a transaction is blocked, delayed, or flagged for review, customers and compliance teams need to understand why. When a suspicious transaction is approved, institutions need to understand what signals were missed. Black box decisions are unpleasant in any regulated environment, but in financial services, where decisions can affect people’s access to money, markets and essential services, it becomes a direct issue of trust.
There is also a human dimension within financial institutions. Compliance and fraud teams are already under pressure from faster payments, rising alerts and more sophisticated criminal behavior. Agentic AI could help them identify patterns, summarize cases and prioritize risks. But it should support human judgment rather than replace it.
Regulation continues to evolve alongside these technological advances. In the UK, the Competition and Markets Authority has published guidance clarifying that organizations remain responsible for the actions of AI agents acting on their behalf. Across Europe, the EU AI law tightens transparency, human oversight, governance and record-keeping requirements for higher-risk AI systems. Taken together, these developments point to an emerging model of autonomous compliance where AI agents are expected to operate within robust governance and audit frameworks.
The future of trusted transactions
The World Economic Forum AI playbook for financial services argues that trust, governance and human oversight become critical tests as financial institutions move from experimentation to scaled AI adoption. Agentic AI will make these tests more sophisticated. Institutions must define where autonomy is acceptable, where human consent is still required, and how responsibility is captured when software acts on behalf of a person.
The future of trusted transactions won’t just depend on who someone is. It will also depend on understanding what they intended, what their AI agent was allowed to do, and whether the action can be explained after the fact.
As AI agents become more involved in economic activities, financial trust must evolve with them. The task ahead is not to slow progress, but to ensure that even as AI begins to act, the financial system can still answer one of its most important questions: Should this transaction be trusted?
https://www.weforum.org/stories/artificial-intelligence/how-do-we-regulate-payments-when-its-ai-agents-spending-the-money/
