The World's Most Aggressive Agentic AI Adopters Are Brazilian Banks

The consensus assumption is that frontier AI adoption happens in US enterprises first and migrates to emerging markets later. In financial services, the data says the opposite is happening.

Itaú Unibanco — Brazil's largest private bank — is now resolving 70% of its security vulnerabilities through Devin, Cognition's autonomous AI software engineer, without a human in the remediation loop. Cognition named Itaú alongside Goldman Sachs, NASA, and Mercedes-Benz when announcing its $1 billion fundraise at a $26 billion valuation. Not a pilot. Not a proof of concept. Production deployment, running at scale.

The same institution has nearly 2,000 active AI projects across its operations, and became the first major Brazilian bank to launch an AI-only investment advisor — first rolled out to clients in Uniclass and Personnalité segments, on track to expand throughout 2026. The deployment philosophy is systematic, not experimental.

Nubank approaches from a different direction but arrives at the same conclusion. Built AI-first from inception, its credit models run on machine learning across 110 million customers in six countries. The R$45 billion committed to Brazil for 2026 lists AI development explicitly as a core strategic priority — not as a cost center, but as the primary competitive surface.

The structural reason Brazilian banks are moving faster than comparable US institutions isn't talent or technology access. It's competitive intensity combined with data infrastructure. Brazilian banking is extraordinarily concentrated at the top — five institutions control over 80% of assets — and fiercely competed at the application layer, where thousands of fintechs operate on the same Pix and Open Finance rails. Any efficiency gain from AI adoption generates outsized margin improvement in a market where pricing pressure is structurally embedded.

The payment infrastructure underneath this adoption is the amplifier. Pix generates real-time behavioral data at a transaction volume no US payment system currently matches. Open Finance creates a consented data-sharing layer that converts that behavioral data into actionable credit and risk signals. When Itaú's AI security agent operates, it draws on a data environment US bank equivalents simply don't have.

In enterprise AI adoption, the institutions that reach production before peers accumulate data advantages that compound every quarter. The Brazilian banks running agentic systems in production today are building leads that will be structurally difficult to close — in the country whose payment infrastructure gives AI models the richest behavioral data in the world.