OpenAI case study: how Brazil's Itaú is spreading generative AI across customers and staff

OpenAI published a case study on October 16 describing how Itaú, a major Brazilian bank, is putting generative AI to work for both customers and its own teams. The account is built around interviews with Carlos Eduardo Mazzei, the bank's chief technology officer, who frames the goal as making the technology almost unnoticeable so that the customer benefit, not the model, is what people remember.

The flagship product is ia.i, a conversational layer activated by text or voice that pulls checking accounts, credit, cards, insurance and other services into a single environment, handing customers to human specialists when needed. Itaú says the first phase reached roughly 300,000 customers, and that the bank used that period to learn how customers actually engage with the product.

Model choice was deliberately plural. Mazzei says the bank needed models that could handle cost- and latency-sensitive interactions as well as complex questions requiring reasoning and grounding, and that OpenAI's models operated across that whole range. Itaú's platform therefore routes work across several providers rather than standardising on one.

Behind ia.i sits Customer DNA, a financial context layer that helps the bank understand the needs, habits and life stages of approximately 70 million customers. Under governance and responsible-AI controls, GPT models are among the components that maintain and enrich that representation, making interactions progressively more relevant, according to the case study.

The scaling layer is IARA, an in-house governance and orchestration platform that gives teams controlled access to different models in line with internal policies. It addresses three problems at once: cost control per token, accuracy of responses, and enforcement of responsible-AI rules. A legal-department solution running on IARA works across more than 90 million documents and executes over 2 million tasks a month, with efficiency gains of 20% to 80% at different stages of the process.

The same intelligence layer reaches inward under the bank's AI First strategy, serving relationship managers and customer service teams with pre-meeting summaries, preparation, issue analysis and alternative plans. Itaú's stated expectations are three times more time available for customer relationships, service up to twice as fast, and conversations up to three times more effective.

Itaú treats bias and cultural alignment as product requirements rather than compliance afterthoughts. Mazzei argues that in a heavily regulated sector the governance layer has to exist before scale, because token economics, answer quality and regulatory obligations all become unmanageable otherwise.

What to watch next is the pace of rollout: ia.i is expected to be available to the bank's entire customer base by the end of 2026. Planned capabilities include product simulations and sign-ups, along with recurring analyses of income, expenses and spending habits.