An artificial intelligence system has, for the first time in Europe, completed a payment inside a live banking network without a human entering the final command. Banco Santander and Mastercard confirmed that they had executed a live end-to-end payment initiated and completed by an AI agent, a software system operating within the bank’s own regulated payments infrastructure.
The move was described by both firms as a milestone in what they call “agentic payments,” where software can act on behalf of customers under set limits and controls.
This was not a simulated experiment. The transaction ran through Santander’s normal payments network using Mastercard Agent Pay, a framework that lets AI agents be registered and treated as participants in the payment flow. The pilot took place under strict security, governance, and compliance rules, and was not open to public use.
The AI agent performed its role inside predefined limits and permissions set by the bank and the customer. The goal was to confirm that an autonomous system could initiate, authorise, and complete a transaction while still meeting the legal and operational guardrails that apply to everyday banking.
Why this AI payment pilot matters
Payments systems are among the most tightly regulated digital services in the world. Any change to how transactions are initiated must still meet authentication rules, fraud protections, and governance standards that financial regulators enforce. That’s why this pilot matters: it embeds an AI actor into a system normally used only by humans.
The transaction was processed through Santander’s live infrastructure rather than a test environment. That means the bank and its partner had to ensure that all compliance checks, security validations, and payment routing worked the same way they would for a normal customer purchase.
Even so, this is still a pilot project. Santander and Mastercard have made it clear that the arrangement is not a commercial service available to customers yet. The objective is to explore how AI agents could one day fit into existing payment flows while keeping the necessary controls intact.
What industry forecasts say
The idea of allowing AI to act autonomously is not limited to payments. Industry analysts have been following the broader shift toward agentic AI systems, software that can complete tasks or make decisions with limited human intervention.
Research and forecast data suggest that this trend is likely to grow in business settings. Gartner, a major technology research firm, forecasts that around 33 % of enterprise software applications will include agentic AI by 2028, up from less than 1 % today. That projection reflects interest among corporate buyers in systems that can perform work on their behalf rather than only assist humans.
Other forecasts align with this view, showing that businesses are increasingly preparing to deploy software agents for routine operations, customer interactions, and workflow automation. These systems are expected to move from early pilots into more common use cases over the next several years.
The Mastercard network itself already reflects the scale of modern digital commerce. Independent reporting notes that Mastercard’s decision-making and fraud-scoring systems work with nearly 160 billion transactions annually across its network, evidence of how vast and complex the environment is where agentic systems might one day operate.
What companies are saying
In its press announcement, Santander highlighted its desire to build a responsible approach to AI payment systems. Matías Sánchez, global head of Cards and Digital Solutions at Santander, said: “Our role is not only to adopt innovation, but to shape it responsibly, embedding security, governance and customer protection by design. As AI agents become part of everyday commerce, building trusted, scalable frameworks will be essential to unlocking their full potential.”
Kelly Devine, President, Europe at Mastercard, described the pilot in terms of continuity rather than change: “With Mastercard Agent Pay, we are applying the same principles that have defined our network for decades — security, interoperability and trust — to a new era of AI-enabled commerce.”
Those comments underscore that neither company is portraying AI payments as already ready for broad use. Instead, they are testing how such capabilities could be governed and scaled safely.
Dogma vs. reality
There is a gap between the buzz around AI and what is operationally feasible today. Agentic AI as a concept promises systems that can act on behalf of users or businesses in real time. But many current applications remain in early stages, and some analyst reports have even warned that a large share of agentic AI projects could be cancelled before they reach production — due to costs, unclear value, or immature technology.
What Santander and Mastercard have shown is that the technical plumbing can work under real-world conditions. But that doesn’t mean consumers can yet unlock AI agents to autonomously pay bills, shop online, or manage subscriptions. Those outcomes will require further testing, regulatory alignment, and robust guardrails for safety, privacy, and fraud prevention.
What enterprise leaders should watch
For business decision-makers, this pilot raises three practical questions:
- Governance and oversight: How will AI agents be controlled so that spending limits, identity checks, and audit trails remain clear?
- Identity and trust: If software can act on behalf of people or companies, how will systems ensure that only authorised actions are taken?
- Risk and liability: Who is responsible when an autonomous agent makes an error or misinterprets instructions?
These are not academic concerns. As enterprise systems begin to support more autonomous tasks, from supplier ordering to subscription payments, organisations will need clear frameworks that define how AI agents are governed, monitored, and held accountable.
The long view for AI-initiated payments
The Santander and Mastercard test is not the finish line for AI-initiated transactions. It is an early step toward understanding how autonomous systems might coexist with regulated financial systems.
The pilot demonstrates that AI systems can be integrated into live payments rails, but only under tightly controlled and monitored conditions. Scaling this to everyday use will require a lot of additional work on controls, security, and compliance.
Still, the fact that a regulated bank and a global payments network have run a successful agent-initiated transaction shows where enterprise experimentation is heading: from pilot programs toward real-world validation. For enterprises planning their own AI strategies, this suggests that action-capable AI may soon move beyond suggestion and automation into governed execution, if done with care and strong oversight.
(Photo by Clay Banks)
See also: Goldman Sachs and Deutsche Bank test agentic AI for trade surveillance
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information.
AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.
