As part of its AI First strategy, Itaú has developed two complementary solutions: ia.i, a multimodal conversational layer, and IARA, a governance and orchestration platform. The bank aims to apply generative AI simultaneously to customer-facing services and internal teams in a heavily regulated sector.
What ia.i does
ia.i provides a single environment for text- and voice-activated interactions across checking accounts, credit, cards, insurance and other services, and routes users to specialized human support when needed. The first rollout reached roughly 300,000 customers, and Itaú plans to make the experience available to its entire customer base by the end of 2026.
Itaú emphasizes that the experience should be simple and directly useful rather than technologically showy. Carlos Eduardo Mazzei, Chief Technology Officer of Itaú, says the goal is to deliver something "simple, direct and useful," making the technology almost imperceptible to the user.
Model selection and Customer DNA
Rather than relying on a single model, the bank uses a range of models to cover low-latency, cost-sensitive interactions as well as more complex tasks that require deeper reasoning. Mazzei notes that OpenAI’s models played an important role in delivering this breadth of experience. At the center of the approach is Customer DNA, a layer that captures financial context—needs, habits and life stage—for about 70 million customers. GPT models, including GPT‑5.x versions mentioned in the project, are components that help maintain and enrich this representation under governance and Responsible AI controls.
IARA: governance, observability and cost control
IARA provides controlled access for teams to different models according to internal policies. The platform supplies observability, token-based cost control and enforcement of Responsible AI rules, enabling the bank to scale AI solutions while keeping compliance and economic viability in check.
Mazzei stresses the importance of preserving data within the bank’s context and granting models secure, governed access: "It is essential that this information be preserved, remain within the bank’s context, and access models in a secure, governed way, with an understanding of token-use efficiency so that the entire solution is economically viable for the bank."
Concrete use cases and metrics
One clear use case on IARA is the legal department’s automation: reading, analyzing and interpreting documents. The legal solution operates on a repository of more than 90 million documents and executes over 2 million tasks per month, producing efficiency improvements of 20% to 80% in stages of the department’s processes.
Internally, the same intelligence layer supports specialists, managers and customer-service teams—automating simple tasks and providing time-saving outputs that were previously infeasible, such as meeting summaries, preparation, rapid understanding of a case and generation of alternative plans. Itaú expects three times more time available for customer relationships, up to twice-faster service and up to three times more effective conversations as outcomes of the implementation.
User demand and future features
Research commissioned by Itaú with Grupo Consumoteca found that 80% of Brazilians want to better organize their financial lives and that 65% would like technologies acting as a financial copilot that provide ongoing guidance. These findings support the bank’s push to expand conversational, personalized support.
Planned next-stage capabilities include product simulations and sign-ups, plus recurring financial analyses of income, expenses and habits.
Lessons and principles
The project highlights several operational principles:
- Measure success by simplicity and clear customer benefit rather than technical sophistication.
- Use a range of models to serve both latency- and cost-sensitive use cases and those requiring deeper reasoning.
- Build a governance layer (observability, cost control, Responsible AI) before scaling.
- Treat bias mitigation and cultural alignment as product requirements so responses reflect the institution’s values.
- Apply the same intelligence internally to free up human time for higher-value customer interactions.
- Start with concrete everyday problems and iterate from an initial customer group.
- Monitor token efficiency from day one to ensure economic viability.
- Combine conversational interfaces with navigational and transactional UI elements where appropriate—natural language is not always the best way to perform every action.
Next steps
ia.i and IARA will continue to evolve. Itaú intends to expand the service to its full customer base by the end of 2026 while maintaining regulatory compliance and Responsible AI practices. For Mazzei, this initiative is part of decades of transformation: customer habits and expectations keep changing, and the bank is adapting its technology and processes to meet those shifts.
