Technology

BBVA and Visa Test AI Agent Payments, Targeting Bank Cost Savings

In July 2026, BBVA and Visa tested a real payment initiated by an AI agent. What changes in cost, control and competition for banks and fintechs in Mexico.

Pago con agentes de IA: tarjeta y teléfono con datos en escritorio

BBVA and Visa executed a real purchase initiated by an artificial intelligence agent in July 2026, using an issued card and the systems of an active merchant, as part of the Visa Agentic Ready program. The transaction took place in Europe, with tokenization and anti-fraud tools, and remains a test.

The gap between recommending and executing is what separates a chatbot from an agent. According to Expansión, BBVA México's first-quarter 2026 Results Report states that its Blue assistant handled more than 27 million interactions between October 2025 and February 2026, of which 18 million came through the phone center, and that it resolves initial inquiries 4.6 times faster than that channel. The company did not confirm that Blue can make transfers or payments autonomously in the country. In Brazil, Bloomberg Línea reported that Nubank launched Nu Smart Invest, an AI tool that rates portfolios and suggests adjustments without buying assets on its own.

The appeal is about costs. A PwC analysis cited by Expansión estimates that broad AI adoption can improve banks' efficiency ratio by up to 15 percentage points, though this is a projection and not savings already captured. The AI Top Business Leaders México study, by MIT Technology Review en español and Opinno with 25 leaders, found that only four quantified the financial return of their projects. Nu Smart Invest evaluates five pillars (diversification, performance, liquidity, protection and efficiency) and awards up to a thousand points. "Recommendations do not mean a guarantee of performance," said Patricia Whitaker, Nubank's director of Investments. The product does not execute purchases, carries no cost for now and is rolling out gradually in Brazil, where the neobank reports more than 118 million customers.

For a financial operator, the data leaves three questions on the table: which operations require express authorization, who is accountable if the agent makes a mistake, and how much it costs to integrate oversight. Mercedes Bouzas, of Opinno, frames it in Expansión: "Every agent needs an owner, clear limits and a mechanism for a person to step in when necessary."

This article was written with the assistance of artificial intelligence from verified sources and reviewed by a human editor before publication.

Sources