Agentic Payments are moving from Theory to an Operating Model

By Koen Vanderhoydonk, CEO at The Connector.

A Brussels roundtable with fifteen senior industry leaders suggests the future of payments will be shaped less by rails than by trust, liability, and the ability to prove who authorised what.

There are moments in financial services when a new idea arrives with just enough maturity to become impossible to dismiss, but not yet enough structure to become easy. Agentic payments now sit squarely in that territory. They are no longer a futuristic thought experiment for innovation teams. They are becoming a live strategic question for banks, payment networks, technology providers, regulators, and corporates that want to understand what autonomous commerce will demand of the systems beneath it.

That was the backdrop to a recent Brussels roundtable, where fifteen senior voices from banking, payments, technology, and regulation gathered under the Chatham House Rule to test where agentic payments are really heading. The value of the discussion was not in personalities or quotes, but in the shape of the consensus: where the room aligned, where it split, and what those fault lines reveal about the industry’s next phase.

The signal from Brussels

The first point of convergence was timing. Most participants expected consumer-facing AI agents to enter the market within five years, with a corresponding Know Your Agent control layer emerging not far behind. That is a material shift. Once software can search, decide, initiate, and complete payment within defined parameters, the debate moves beyond product novelty and into institutional readiness.

The second signal was regulatory. The sharper discussion in the room was not whether agentic payments are technically feasible, but which rulebook will shape them most decisively.

The Payment Services Regulation was widely seen as more important than PSD3 in defining the commercial environment ahead, suggesting that governance may move from a supporting role to a leading actor rather quickly.

And then there was the lingering forecast. By 2030, participants saw agents driving more than one in ten European consumer payments, even without consensus on who would own the rails. Card schemes had their believers, big tech had its advocates, and banks and stablecoin networks each had their case. The lack of agreement was revealing in itself: the field is open, and the competitive map is still being drawn.

The real issue is liability

The most telling point from the Brussels discussion was not volume, speed, or convenience. It was a liability. When an agent makes the wrong payment, responsibility will not sit neatly with one party; it will be shared across the customer, the agent provider, and the bank.

That insight goes to the heart of why agentic payments deserve senior attention. In conventional payments, responsibility can already be complex. In agentic payments, the chain becomes denser, faster, and more distributed. The institutions best placed to compete will not simply be those that enable autonomous action, but those that can clearly and quickly prove who authorised what, under which mandate, and within which controls.

From advice to action

The white paper’s core contribution is that it treats agentic payments not as a feature, but as a structural shift. For years, AI in financial services mainly watched and advised. It scored fraud, flagged anomalies, ranked risk, and recommended action. A person or deterministic system still made the payment.

Agentic payments break that pattern. An agent can hold a goal, break it into tasks, call the right tools, adapt as conditions change, and execute payment within defined boundaries. A procurement agent may source suppliers and settle invoices. A travel agent may rebook and pay inside budget. A machine agent may purchase compute, data, or digital services in tiny increments at speeds no human could manage efficiently.

That is why the paper argues that payments have not merely grown more digital. They have changed form.

From rails to trust

One of the paper’s sharpest arguments is that the industry contest is shifting away from rails and toward trust. The settlement mechanism still matters, of course, but less than the ability to establish intent, verify identity, enforce limits, and maintain accountability after the fact. In that world, trust is not a soft concept. It becomes architecture.

The paper maps this through a three-layer model: intent and orchestration, authorisation and control, and settlement and finality. That separation matters because autonomy can only scale if freedom at one layer is constrained by certainty at another. The agent may reason. The control layer must check. The settlement layer must complete with the same finality expected of any regulated payment flow.

The work beneath the vision

This is also where the white paper becomes refreshingly practical. It does not pretend that a clever interface or capable model is enough. It points to the harder work beneath the vision: enterprise integration, policy controls, reconciliation, observability, and the operational discipline required to move from pilot to service.

That distinction matters more than many transformation narratives admit. A pilot can be watched closely. A service has to run inside the institution’s risk, compliance, and operational perimeter at scale. Agentic payments will not be won by demos. They will be won by infrastructure.

Why the roundtables matter

That is precisely why these roundtables matter. They are not side conversations orbiting a trend. They are part of the process through which the industry decides how autonomous payments will be governed, deployed, and trusted. The Brussels discussion made one thing plain: the future is not waiting for perfect consensus.

More senior roundtables will follow across the Benelux later this year. For leaders who want a serious view of where agentic payments are heading, and an opportunity to help shape the standards, assumptions, and strategic choices around them, this is the right time to be in the room.