AI transaction in Europe: Santander and Mastercard innovate

Banco Santander and Mastercard completed the first transaction in

• Santander and Mastercard announced the first end-to-end “live” transaction in Europe executed by an AI agent.
• The payment was processed on Santander’s real payments infrastructure, using Mastercard Agent Pay.
• The initiative seeks to enable purchases made by AI agents on behalf of customers, with security and privacy standards.
• The solution is in a pilot phase and is not yet being rolled out commercially; the next step is to validate technical and operational readiness.
• Santander expects to generate €1 billion in value from AI investments over the next two years.

Agentic payments in regulated banking
– What happened: an end-to-end “live” payment executed by an AI agent was completed, within the real payments flow.
– Where: the pilot was reported as carried out in Spain, in a controlled environment.
– With what: Santander provided its production payments infrastructure and Mastercard the Agent Pay layer to integrate the agent into the process.
– Status: it is a pilot test (not a commercial launch) aimed at validating technical and operational readiness.
– Why it matters: it brings “agentic commerce” closer to a regulated banking framework, where trust depends on permissions, limits, traceability, and controls.

Innovation in payments: Santander and Mastercard’s AI transaction

Banco Santander and Mastercard have taken a significant step in the evolution of digital payments by completing what they describe as the first transaction in Europe executed by an artificial intelligence agent.

It is worth emphasizing that, according to what was communicated, this is a pilot-phase test and not a commercial rollout. The milestone is part of the rise of so-called agentic commerce: AI systems capable of acting with a certain degree of autonomy to search, decide, and execute purchases on a person’s behalf.

The key to the announcement is not only the use of AI, but its integration into a regulated banking environment , a critical point for moving from lab demonstrations to operations with industry standards.

Payments with Autonomous Agents
What changes when “an agent pays” (and not just an app):
– Delegation: the user defines intent and preferences; the agent executes tasks (search, compare, buy) within limits.
– Authorization: the payment still needs explicit permissions (for example, caps, categories, allowed merchants) and validations.
– Traceability:

a record must remain of what the agent decided, with what data and under what rules, in order to be able to audit and resolve disputes.
– Trust: the experience only scales if the customer feels in control (pause, confirm, revert) and if the ecosystem (bank/network/merchant) can manage fraud and incidents.

Details of the first live AI transaction in Europe

According to information released by Finextra, the operation was an end-to-end payment executed by an AI agent, processed through Santander’s production live payments infrastructure and routed via Mastercard Agent Pay technology.

In other words: Santander provided the “live” payments rail and Mastercard the layer that enables an AI agent to execute the payment within the flow, under controls.

Additional information about the pilot indicates that the test was carried out in Spain and that the AI agent completed a purchase in a controlled environment, with predefined limits and permissions. The goal was to demonstrate technical viability without presenting the experiment as a commercial launch.

End-to-End Agentic Payment Flow
Typical end-to-end flow of an agentic payment (with checkpoints):
1) The customer defines rules: budget, allowed categories, preferred merchants, and whether confirmation is required before paying.
2) The agent decides: it searches for options and selects a purchase that fits the rules.
– Checkpoint: if the purchase exceeds limits (amount/category), it must request confirmation or be blocked.
3) The payment intent is generated: the agent initiates the payment within the flow supported by the bank/network.
4) Authorization on the “live” rail: validations are applied (limits, authentication when applicable, anti-fraud signals).
– Checkpoint: if there are risk signals, the system may require additional verification or reject.
5) Execution and confirmation: the payment is completed and the customer receives confirmation/receipt.
6) Logging and traceability: logs of the agent’s action are stored (what it bought, for how much, under what permission) for auditing, returns, or disputes.

Santander’s payments infrastructure and Mastercard Agent Pay

The payment was processed through Santander’s live payments infrastructure, which implies that the bank acted as the operational backbone of the transactional flow, while Mastercard provided the layer that makes it possible to incorporate AI agents into the payment process.

Mastercard Agent Pay is presented as a mechanism for AI agents to operate as “governed” participants within the payments flow, maintaining traditional principles of card networks: interoperability, scalability, and controls ofsecurity.

That “governance” approach is key in payments: the agent’s autonomy is bounded by permissions, limits, and validations defined in advance. In practice, the challenge is to connect the agent’s autonomy with clear rules: what it can buy, for how much, under what conditions, and with what validations.

Commitment to security and customer protection

Both Santander and Mastercard emphasize that progress is only sustainable if it is built on security, privacy, and governance. In the announcement, Matías Sánchez, Santander’s global head of cards and digital solutions, argued that the bank’s role is not only to adopt innovation, but to “shape it responsibly”, incorporating security, governance, and customer protection “by design.”

The emphasis responds to a central concern of agentic commerce: if a system can buy on behalf of the user, it must also be able to be accountable for why it did so, within what limits, and how fraud, abuse, or unwanted purchases are prevented.

Controls for Trusted Payments
Expected controls for a “paying agent” to be trustworthy:
– Explicit permissions: allowed categories/amount/merchants and exception rules.
– Limits and “double confirmation”: thresholds above which the agent must request approval.
– Authentication and risk signals: additional verification when the pattern doesn’t fit.
– Traceability: record of the agent’s decision (rule applied, amount, merchant, time).
– Operational reversibility: routesclear policies for returns, cancellations, and disputes.
– Oversight and pause: ability to deactivate the agent or suspend purchases in the event of anomalies.

Pilot phase and future expectations for the solution

The initiative is in a pilot phase and, for now, is not being rolled out commercially. The next step, as communicated, is to ensure technical and operational readiness for end-to-end AI-driven payments.

That work usually includes resilience testing, incident management, traceability of the agent’s decisions, integration with merchants and providers, and defining processes for disputes or returns when the purchase is initiated by an automated system.

Promises and scaling challenges
What the pilot promises vs. what usually complicates scaling:
– Benefit: more convenience (recurring purchases, automatic comparison, rewards optimization) → Challenge: avoiding “surprise” purchases with well-designed limits and confirmations.
– Benefit: smoother payments integrated into everyday life → Challenge: 24/7 operations (incidents, outages, retries) without degrading the experience.
– Benefit: personalization (preferences, budgets) → Challenge: traceability and explanation of the agent’s decisions when there is a dispute.
– Benefit: potential adoption across multiple merchants → Challenge: ecosystem integration (merchants, acquirers, providers) and consistency of standards.

Expected impact of Santander’s investments in AI

The announcement comes at a time of acceleration in the bank’s AI strategy. The previous week, Santander reported that it expects to generate 1 billion euros from investments in artificial intelligence over the next two years.

In that context, agentic payments can become another piece of a broader agenda: automation, service personalization, operational efficiency, and new digital experiences. If agentic commerce takes off, banks and payment networks aim to remain the trust layer that connects identity, authorization, and the movement of money.

Statements from Santander and Mastercard leaders on innovation

Matías Sánchez (Santander) framed the progress as a design responsibility: innovation with these principles built in from the outset, anticipating a scenario in which AI agents become part of everyday commerce.

From Mastercard, Kelly Devine, President for Europe, summed up the approach in one phrase: “innovation and trust can move forward together”, positioning the pilot as proof that AI-based automation can be incorporated into the payments system without giving up the principles of security and reliability.

Innovation and trust in AI
Quotes from the announcement (direct evidence):
– Matías Sánchez, Global Head of Cards and Digital Solutions (Santander): “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 (Mastercard): “This milestone with Banco Santander demonstrates that innovation and trust can advance together.”

The Future of AI Transactions in Europe

Santander and Mastercard’s pilot transaction serves as an early indicator of where the sector is heading: payments where the user delegates tasks to software, but demands the same level—or greater—of control and protection as in traditional channels.

Innovations in Autonomous Commerce

Autonomous commerce promises to automate repetitive decisions: comparing prices, applying preferences, optimizing rewards, or executing recurring purchases. For banks and networks, the challenge is enabling that convenience without losing traceability: ensuring that each agent action is verifiable, reversible when appropriate, and aligned with the customer’s explicit permissions.

Challenges of Regulation and Consumer Trust

Europe combines high digitalization with a demanding regulatory framework. Mass adoption will depend on agentic payments offering transparency, clear authorization mechanisms, and an experience that does not sacrifice the trust of theconsumer. The pilot marks a beginning; scalability will depend on how responsibilities, disputes, and common standards are resolved among banks, merchants, tech companies, and supervisors.

Progressive adoption scenarios
Adoption scenarios (and conditions for them to occur):
– Short term (pilots): limited cases (low amounts, specific merchants, strict rules) and a focus on stability, anti-fraud, and traceability.
– Medium term (first offerings): more merchants and more use cases (subscriptions, replenishment, travel), with better confirmation flows and returns/dispute management.
– Long term (normalization): interoperable agents across platforms and merchants, with common standards for permissions, auditing, and operational responsibility.
Cross-cutting condition: the user must be able to understand and control “what the agent can do” at any time.

Transforming the Customer Experience in Telecommunications

AI automation is not only redefining payments: it is also changing how companies manage customer service, operations, and digital channels, especially in high-volume sectors such as telecommunications.

The Artificial Intelligence Revolution in the Telecom Sector

Telcos are incorporating AI to anticipate incidents, personalize offers, and reduce resolution times. The qualitative leap comes when systems move from answering questions to orchestrating processes: diagnosis, verification, execution, and follow-up.

Benefits of Automation in Customer Service

Among the most cited benefits are reduced waiting times, greater consistency in responses, 24/7 availability, and the ability to absorb demand spikes. The real value, however, appears when automation is integrated with internal systems to resolve, not just inform.

How Suricata Cx Improves Operational Efficiency

Platforms such as Suricata Cx are aimed at optimizing service and operations flows through automation and analytics, with the goal of reducing manual tasks and improving the productivity of support teams, maintaining continuity across channels.

The Future of Omnichannel Operations

The emerging standard is a unified experience: that the customer can start a process in one channel and finish it in another without repeating information. AI acts as a coordination layer to maintain context, prioritize cases, and escalate to human agents when necessary.

The Importance of Real-Time Integration

Real-Time Integration

Automation only works fully if it connects in real time with inventory, billing, service provisioning, and incident-management systems. Without that integration, AI is limited to conversation; with it, it becomes execution.

AI transaction in Europe: Santander and Mastercard innovate and make it clear that agentic commerce will only scale if it is integrated into real infrastructures with governance, traceability, and control. That same logic guides Suricata Cx: taking AI from promise to day-to-day operations, automating with clear limits and human oversight to sustain trust and compliance in every interaction.

From Suricata Cx’s perspective—an omnichannel CX platform with AI for ISPs and telcos—these kinds of pilots reinforce a practical idea: automation delivers value when it is integrated into real workflows and maintains human control, auditing, and clear rules.

This article is based on publicly available information at the time of writing and describes a pilot, not a widespread commercial rollout. Some operational details (such as permissions, authentication, or dispute management) may vary between implementations and evolve with further testing. The information may change with subsequent announcements or updates.