How AI Agents and Tokenization Could Transform Cross-Border Payments

AI agents could make cross-border payments more responsive by comparing fees, exchange rates and execution options, then routing and timing transactions according to a user’s instructions. Tokenization could bring payment information, reconciliation and settlement together on programmable platforms. Together, the technologies may reduce friction—but they do not eliminate the need for reliable infrastructure, legal certainty, liquidity and accountable oversight. 17212

How the technologies could change payments

An autonomous payment agent can interpret a user’s objective and carry out steps such as comparing providers or coordinating a payment with available liquidity and contractual terms. The IMF describes these as potential uses, while also noting that agentic payments raise questions about authorization, traceability and responsibility. 172627

Tokenization represents claims on money or other assets in digital form on a programmable ledger. A shared record can reduce the need for separate institutions to reconcile their own records, while programmable rules can automate parts of a transaction. With atomic settlement, the linked parts of an exchange settle together—or not at all. 12

The BIS’s Project Agorá prototype has demonstrated atomic, multi-currency settlement for wholesale cross-border payments using tokenized commercial bank deposits and central bank reserves on a shared platform. That is evidence of technical potential, not proof that one platform is ready to replace existing payment systems across markets. 1213

Why faster payments still depend on sound foundations

Automation cannot fix gaps between payment systems that do not interoperate. Cross-border transactions also span different legal and regulatory frameworks, while institutions need access to the right currencies and enough liquidity to complete payments. The IMF identifies interoperability, governance, settlement and foreign-exchange liquidity among the foundations for cross-border payment innovation. 23

Tokenized platforms could also divide liquidity if assets become confined to closed systems that cannot readily connect with one another. The IMF has identified this risk of fragmentation, alongside the need for clear policy frameworks, legal certainty and safe settlement assets. 2024

Responsibility must be clear when an agent initiates a payment, makes an error or acts outside its intended authority. The IMF’s analysis of agentic payments highlights unresolved legal and liability questions, as well as risks from opaque decisions and unclear authorization trails. 27

How AI could amplify operational and market risks

AI has a dual role in cybersecurity: it may strengthen defenses, but it can also help malicious actors find and exploit vulnerabilities. Risks can spread beyond a single firm when financial institutions depend on shared infrastructure or common service providers. 2139

Automated decisions can create another vulnerability if many systems react similarly to the same conditions. The IMF identifies correlated behavior as a risk from agentic payments and warns that synchronized strategies in markets can contribute to asset correlations and liquidity pressures. 1927

The Bank of England has raised a separate concern about capital markets: growing AI-related debt issuance increases market exposure to developments in the AI sector. That warning describes an area of vulnerability to monitor; it does not mean AI has already caused a bond-market crisis. 33

Practical safeguards for banks and regulators

The policy challenge is to make automation useful without making it unaccountable or brittle. Practical safeguards include:

• Limit what agents can do. Set clear payment permissions, transaction limits and conditions; require additional approval for unusual or high-impact payments. Keep an auditable record of instructions and actions so decisions can be reviewed. These controls address the authorization and traceability concerns identified by the IMF. 27
• Plan for liquidity and settlement failures. Test how payments behave when funding is scarce, a counterparty is unavailable or a connected system is down. Preserve reliable fallback processes rather than assuming that faster settlement alone guarantees resilience. Liquidity, settlement and infrastructure links are core foundations for cross-border payments. 23
• Reduce single-provider exposure. Map dependencies on technology providers and shared infrastructure, and prepare for service interruptions. The Bank of England and IMF both identify common providers or third-party dependencies as potential channels for wider operational risk. 2142
• Keep platforms connected and rules clear. Promote interoperability and establish clear legal responsibility for tokenized claims and agent-initiated actions. That can help avoid isolated pools of liquidity and uncertainty over who is accountable when something goes wrong. 202427
• Retain human judgment where it matters. Use review and escalation for consequential exceptions, and make systems’ decisions sufficiently explainable for people to challenge them. This is especially important given the IMF’s concerns about opacity in agentic payments and the Bank of England’s attention to explainability and transparency in AI models. 2739

The opportunity is not simply to make payments faster. It is to coordinate decisions and settlement more efficiently while preserving competition, clear accountability and the capacity to withstand failures. The technology can help; whether it does so safely depends on the foundations around it.