Banco do Brasil Isn't Starting Agentic AI with Chatbots. It's Starting with Relationship Managers.

The most expensive hour in banking is not the fraud alert, the trade settlement, or the loan origination. It's the time a relationship manager spends searching disconnected systems before a client call, assembling context that should already be in one place. Banco do Brasil just pointed its AI investment at that specific hour.

Brazil's largest bank — roughly 75 million customers, R$900 billion in assets, and a distribution network that reaches municipalities where no private bank has a branch — began piloting NICE Copilot this month. The system sits alongside relationship managers in real time: summarizing customer interactions as they happen, surfacing engagement history, and flagging sentiment shifts that a human managing dozens of accounts in parallel would likely miss. The output is not a chatbot. It's better-prepared bankers.

The choice of starting point is deliberate. Brazilian banks have been faster than almost any major market to adopt agentic AI — 52% of global financial services firms are at pilot stage or beyond as of Q1 2026, and Brazil's institutions consistently sit at the higher end of that distribution. But most first deployments went to back-office workflows: AML alert review, document processing, regulatory reporting. Banco do Brasil moving into relationship management signals that the technology is now stable enough to put in front of the front-line, where mistakes have direct consequences for clients and the bank's reputation.

This matters for competitive dynamics across the Brazilian financial system. Banco do Brasil serves a population that private-sector fintechs have struggled to reach at meaningful cost: retirees, rural borrowers, federal employees in small municipalities. If relationship managers at those branches become more effective through AI assistance — closing more loans, catching early signs of financial stress, retaining clients who might otherwise drift to digital-only challengers — the efficiency gap between a state bank with 3,000 branches and a fintech with a clean app closes from an unexpected direction.

The pressure this creates for private-sector institutions is specific: the AI capability race is no longer only between incumbents and challengers. Brazil's federal government controls Banco do Brasil, Caixa, and a majority of the country's credit infrastructure, and is now deploying the same tools as the fintechs — with existing customer relationships, regulatory capital, and distribution that fintechs don't carry. Founders building financial products for Brazil have watched Nubank commoditize the digital checking account from below; they may soon face a state bank competing on advisory quality from above.

The unresolved question is what happens to the data Banco do Brasil's AI accumulates from these conversations — which products customers asked about, what financial anxieties they expressed, which competitors they mentioned. That dataset, built across 75 million customers over years of AI-mediated relationship banking, is worth considerably more than the productivity gains from any single pilot. Whether Banco do Brasil treats it as an operational convenience or a strategic asset will determine what Agentic Open Finance actually means at the scale where most Brazilians live their financial lives.