Nikolai Braiden is a seasoned fintech visionary who has spent years at the intersection of blockchain and digital lending, advocating for a future where technology removes friction from human life. With an extensive background advising startups on the deployment of emerging tech, he has become a leading voice in the evolution of payment systems. His recent focus has shifted toward the transformative potential of agentic commerce, where artificial intelligence moves beyond simple suggestions to executing financial transactions. This conversation explores the groundbreaking collaboration between Santander and Visa, which recently executed the first end-to-end AI-powered payments across five Latin American markets.
We delve into the technical frameworks that allow these digital proxies to operate within strict regulatory bounds, the shift in how financial institutions must govern automated shopping, and the path forward as these systems scale from simple retail purchases to complex financial services.
Recent pilot programs in markets like Brazil and Mexico have successfully tested AI agents purchasing goods like books and chocolates. How were the specific consent parameters established for these agents, and what steps ensure the transactions remain strictly within the consumer’s pre-approved spending limits?
The success of these pilots in Argentina, Brazil, Chile, Mexico, and Uruguay hinged entirely on a controlled consent-driven framework. We established parameters where the consumer explicitly delegates authority to the AI agent, creating a digital “power of attorney” for specific transaction types. These limits are not just suggestions; they are hard-coded into the Visa Intelligent Commerce framework, which ensures that if an agent tries to buy a high-end watch when it was only authorized for a $20 book, the transaction is instantly blocked. It feels remarkably secure because the issuer maintains ultimate control over the credential, allowing for a seamless experience that honors the user’s intent without sacrificing oversight. By testing these real-world purchases of books and chocolates, we proved that the AI can navigate the merchant’s environment while staying within the precise financial boundaries set by the cardholder.
Operating automated commerce across five different jurisdictions requires navigating complex regulatory standards. How does the underlying technical framework manage cross-border compliance, and what specific security protocols prevent these delegated AI payments from being flagged as fraudulent or unauthorized by traditional monitoring systems?
Navigating the diverse regulatory landscapes of five distinct Latin American markets required a sophisticated blueprint for governance and interoperability. The system utilizes the Visa Intelligent Commerce (VIC) infrastructure, which acts as a translator between the AI agent’s actions and the traditional banking network’s security protocols. To prevent these transactions from being flagged as fraud, we use unique identifiers that signal to the bank that this is a “delegated” payment rather than a compromised account. This transparency allows the bank to recognize the AI proxy as a legitimate participant in the ecosystem, adhering to the same supervisory standards as a manual checkout. It is a delicate balance of maintaining strong issuer controls while allowing the AI to move between Argentina and Uruguay with the same level of trust as a human traveler.
With a significant majority of consumers in Latin America already using AI tools to assist their shopping, the transition to fully automated fulfillment seems inevitable. How will the role of financial institutions shift as they move from processing manual checkouts to governing these complex, delegated AI proxies?
We are seeing a profound shift where banks like Santander are moving from being passive payment processors to active governors of digital identities. Research shows that over 70% of Latin American consumers are already integrating AI into their shopping journeys, so the demand for automation is palpable. Financial institutions must now focus on managing the “trust layer,” ensuring that the AI proxies acting on behalf of their customers are verified and secure. This evolution means banks will spend less time worrying about the physical act of swiping a card and more time building robust frameworks for discovery and fulfillment. The goal is to move the needle toward 2026, where the bank’s primary role is to provide the secure infrastructure that validates an AI’s decision-making process before a single cent is moved.
While initial tests focused on low-value retail items, the long-term goal involves high-stakes sectors like travel and wealth management. What technical hurdles must be cleared to scale this infrastructure for complex subscriptions, and how will the validation process change as agents become more autonomous?
The jump from buying a book to managing a multi-leg international travel itinerary or a wealth management portfolio is significant and requires a higher tier of validation. Technical hurdles involve creating more nuanced logic for “conditional consent,” where an AI might be authorized to book a flight only if it falls within certain dates and price points. As agents become more autonomous, the validation process will likely move toward real-time, multi-factor verification where the AI provides a proof-of-intent to the bank. We are building the foundation now with retail items to ensure that the data handling is airtight before we move into the complexities of recurring subscriptions. Scaling will require global banks to adopt a unified underlying infrastructure that can handle the high-stakes liability associated with larger financial commitments.
Establishing a blueprint for interoperability is essential for making AI-driven shopping a global reality. How do these delegated payment frameworks integrate with existing card network standards, and what metrics are used to measure the success of an AI agent’s decision-making process during a transaction?
The beauty of this pilot is that it doesn’t try to reinvent the wheel; instead, it leverages the existing, secure card network standards that Visa has refined over decades. By integrating with these established systems, we ensure that merchants don’t need to overhaul their entire checkout process to accept an AI-initiated payment. We measure success through metrics like transaction completion rates, the accuracy of the agent in following spending limits, and the speed of the consent-to-fulfillment cycle. If an AI can successfully navigate the purchase of chocolates in Brazil using a pre-approved credential without triggering a false-positive fraud alert, we consider that a major win for interoperability. These metrics give us the data necessary to prove to regulators that agentic commerce is not just a futuristic concept, but a scalable reality.
What is your forecast for agentic commerce?
I believe that agentic commerce is poised to become an everyday reality by 2026, fundamentally reshaping the global retail and financial landscape. We will see a rapid transition from simple “discovery” AI to “execution” AI, where the friction of the manual checkout page completely disappears for most routine purchases. As more global banks adopt the governance frameworks we are testing today, consumers will grow to trust these digital proxies as much as they trust their physical wallets. My forecast is that within three years, delegated payments will move beyond Latin America to become a global standard, handling everything from your monthly grocery restock to complex insurance renewals. The era of the “active consumer” is shifting toward the era of the “overseeing consumer,” where our technology does the work and we simply provide the consent.