Context: a digital bank under pressure from speed and regulation
C6 Bank operates in one of Brazil's most regulated and competitive environments — the digital banking sector. Launching financial products requires regulatory approval, security testing, compliance reviews, and multiple development cycles. The time between product conception and delivery to customers has historically been measured in months. At the same time, the pace of fintech innovation demands that digital banks compress these cycles without compromising the standards that regulators and customers expect. This is the knot that C6 Bank decided to attack with agentic artificial intelligence.
The challenge: speed without sacrificing control
The most common trap in AI adoption at financial institutions is the temptation to measure success solely by volume of code produced or raw delivery speed. C6 Bank identified this risk before starting. "The key to the success of agentic AI lies in allowing us to gain speed in the product-building process without setting aside the security and governance with which every solution in the financial market must be born," said José Luiz Santana, CISO of C6 Bank. The decision not to separate speed from governance was the starting point — and what differentiates this case from most implementations that never leave the drawing board.
The solution: Kiro as the central agent with a structured ecosystem
The chosen tool was Kiro, an autonomous development assistant integrated with Amazon Web Services technologies. But the technology was only the starting point. C6 Bank's strategy was built around three pillars that Gartner, in a survey published the same week, identified as the differentiators of high performers in AI: team training before scaling, measurement of results at each stage, and treating AI as a value portfolio — not as an isolated project. Instead of deploying Kiro and measuring lines of code generated, the bank prioritized planning phases and trained its teams to work with the agent in a structured way, building an ecosystem where governance and speed coexist.
The results: three cases with verified data
Results were measured across three distinct fronts. In financial product launches, a new product that would have required 8 to 10 development sprints was completed in just one sprint — it is already in production, managing hundreds of thousands of reais and serving thousands of customers. In building a new governed data platform, a scope originally estimated at three months with a full architecture team was delivered in eight days by two architects. In regulatory compliance review, processes that previously consumed weeks were significantly compressed while maintaining the operational controls required by Brazil's Central Bank. The consolidated average across these three fronts resulted in an 88% reduction in delivery time — a figure presented publicly at AWS Summit São Paulo.
The lessons: what others can replicate
The C6 Bank case is not about Kiro — it is about the method. Any organization seeking similar results will need to replicate the logic, not just the tool. The first lesson is that training must precede scaling: teams not trained to work with agents produce inconsistent results regardless of which model or platform they use. The second is that business metrics outperform AI metrics: C6 Bank did not measure tokens generated or inference speed — it measured sprints, delivery days, and products in production. The third is that governance is not a brake — it is an enabler: by maintaining security controls from the start, the bank eliminated the review and correction cycles that typically consume the time "saved" in less disciplined implementations. The 88% result is not the ceiling — it is the starting point for those who begin with the same discipline.
