logo

Can finance connect new digital rails without fragmenting the system?

Add The Asian Banker on Google
Discover more trusted banking and financial services insights by adding The Asian Banker as a preferred source on Google.
Can finance connect new digital rails without fragmenting the system?
  • 125

Sibos 2026's opening plenary tied tokenised money, AI and always-on payments to one question: how to link new capabilities to today's global system without weakening liquidity, controls, resilience or trust.

MIAMI — The financial industry is moving from debating individual technologies towards a harder infrastructure problem: how to connect new forms of money, artificial intelligence and always-on services with the systems that already move value across institutions and jurisdictions.

That question ran through the Sibos 2026 opening plenary. Graeme Munro, Chair of Swift's board, described a strategy of maximising the infrastructure the industry relies on today while helping the community build what comes next. Dan Katz, First Deputy Managing Director of the International Monetary Fund, examined how AI could improve existing payments while eventually creating new requirements for an economy in which autonomous agents transact continuously. Citi Group’s Chief Executive Officer Jane Fraser argued that traditional and digital rails cannot be allowed to develop as competing systems.

The common thread was interoperability. The issue is no longer simply whether tokenisation, agentic AI or real-time payments can work individually. It is whether institutions can make them operate together while preserving the trust, resilience and controls on which financial infrastructure depends.

Swift puts interoperability between today's infrastructure and what comes next

Munro said Swift is pursuing two objectives simultaneously: improving today's cross-border payment capabilities and enabling the industry to benefit from tokenisation and always-on finance.

He pointed to Swift's digital ledger as one step towards allowing different forms of value to move across networks with “choice, interoperability, and trust”. He also emphasised the role of richer structured data as the industry progresses with ISO 20022.

The significance of that framing is that Swift is not presenting new digital infrastructure as a clean break from the existing system. The proposition is to connect new forms of value to a network whose established functions still have to remain secure, resilient and available.

Javier Pérez-Tasso, Swift's Chief Executive, returned to the same issue later in the plenary, arguing that interoperability should connect capabilities across jurisdictions and technologies rather than force the industry onto a single platform.

That places the transition problem between two extremes: preserving existing infrastructure unchanged would limit what new technologies can do, while building disconnected new rails would reproduce fragmentation in another form.

AI can improve existing payments before it changes the architecture

Katz argued that the existing payment system was not designed for autonomous agents operating continuously and at scale, but he cautioned against assuming that the immediate answer is to rebuild the financial system around artificial intelligence (AI).

He identified substantial benefits from applying AI to the architecture already in place. Agents could compare fees, exchange rates, services and execution quality and route transactions more efficiently. They could automate complex transactions and machine-to-machine payments, optimise timing around liquidity and contractual obligations, and reduce compliance costs through automated monitoring, reconciliation and reporting.

Katz distinguished agentic systems from conventional programmability. Programmable systems follow explicit instructions within predefined parameters; agents can interpret context and act without requiring a new instruction for each transaction.

Even so, payments still require compliance, dispute resolution and customer support regardless of whether they run on blockchain or another rail. Katz's near-term argument was therefore evolutionary: AI can make the existing payment system work better before the industry resolves the more difficult question of how the system itself may need to change.

Tokenisation adds liquidity and legal questions, not only speed

Katz then turned to the infrastructure that might be required if AI agents become a significant part of economic activity. He said the financial system would need to become more digital, interoperable and programmable, with tokenisation one potentially useful technology because tokenised assets can be programmed and settled rapidly at any hour.

But faster settlement creates its own requirements. Katz said legal frameworks need to establish the nature of tokens, the rights of owners and counterparties and their relationship with underlying real-world assets. Authorities also need to avoid unnecessary barriers between networks.

Atomic settlement can reduce counterparty risk while increasing intraday liquidity requirements because transactions may need to be prefunded. Tokenised commercial-bank money also raises questions about how new forms of private money connect with the central-bank money that anchors the existing banking system.

His argument was therefore not that tokenisation removes existing financial constraints. It changes where some of those constraints appear and increases the importance of legal, liquidity and settlement connections between networks.

Citi brings the infrastructure question back to client liquidity

Fraser connected those system-level questions with what corporate clients are asking banks to solve.

She said globalisation is not disappearing but trade flows are being rerouted as companies adapt supply chains to geopolitical and economic disruption. Against that background, clients want financial infrastructure that is always available, instant, secure, balance-sheet efficient and interoperable.

Fraser said releasing trapped liquidity accounts for more than 70% of the pain points Citi is hearing from clients.

“Our clients want this to be as uncomplicated for them as possible,” she said. “They don't want to be worried about how the plumbing is working.”

She pointed to Citi Token Services as one way the bank is connecting tokenised deposits with existing client accounts and 24/7 clearing. Her broader point was that clients do not want to manage the boundaries between individual rails themselves; they want the bank and the financial system to make those connections work.

Traditional and digital rails cannot become competing islands

Fraser said the industry should not allow a new system and an old system to develop in competition with each other.

“Our clients want to be linking traditional rails to the digital world,” she said.

She argued that digital assets need connections to traditional rails, while established infrastructure also needs access to digital capabilities. Her analogy was the early development of railways with incompatible track gauges: individual networks could function, but fragmentation made movement across systems inefficient.

That connection also changes the discipline required of banks. Fraser rejected the technology-sector maxim of moving fast and breaking things when applied to financial infrastructure.

“Our mandate has to move fast whilst retaining trust,” she said.

Compliance, control systems and the software supporting existing infrastructure therefore have to develop alongside new capabilities rather than follow after them.

Agentic commerce makes identity and control part of the payment problem

The same constraint applies to AI agents that may eventually act rather than merely provide information.

Fraser said banks will need to know whether an agent is genuinely authorised to act for a customer and whether its behaviour remains within intended parameters. She described Citi's ARC control layer as providing observability over agent activity and envisaged agents potentially being supervised by other models and agents as well as people.

That turns authentication and control into part of the infrastructure question. If autonomous agents can initiate payments and economic activity, identity, authorisation, transaction controls and payment execution have to operate together continuously.

Katz had made the corresponding macro-level point: the existing payment system was not built for autonomous agents operating at scale. Fraser's comments illustrated one of the institutional control problems that would have to be solved before that model becomes routine.

The question shifts from invention to connection

The opening plenary did not establish that tokenised finance, agentic commerce or always-on global payments have reached uniform adoption. Katz explicitly identified unresolved legal, liquidity and regulatory questions, while Fraser's comments showed that authentication and controls remain central as AI becomes more autonomous.

What the discussion did establish was a common direction of travel. Swift is trying to connect new digital capabilities with existing cross-border infrastructure. Citi is linking tokenised deposits with existing accounts and clearing. The IMF is examining how AI can improve the payment system already in place even as it considers how that system may eventually need to evolve.

That shifts the business question from whether individual technologies work to whether they can be connected at institutional scale without fragmenting liquidity, controls or trust.

For Sibos 2026, interoperability therefore sits beneath several otherwise separate themes: tokenised money, real-time and 24/7 payments, AI agents, richer data and resilience. Their value to banks and clients increasingly depends on whether the industry can make them work together rather than create another set of isolated rails.

Chat with us WhatsApp