October 16, 2026
With its conversational interface ia.i and governance platform IARA, Itaú is putting its “AI First” strategy into practice across both customers and employees in one of Brazil’s most heavily regulated industries.
-
Company size: Enterprise
-
Region: Latin America
-
Industry: Finance
-
Product: API
-
~300,000: Customers reached in phase one of ia.i
-
100%: Targeted customer base coverage by the end of 2026
-
90M+: Documents indexed in the legal solution’s knowledge base
-
20% to 80%: Efficiency gains achieved across various department processes
The challenge of making technology invisible
In a country like Brazil, crafting a warm, personalized experience for tens of millions of customers starts with simplicity, not the model itself.
“The key is to build something simple, straightforward, and practical that delivers tangible value to customers, while keeping the technology itself virtually invisible,” says Carlos Eduardo Mazzei, Chief Technology Officer (CTO) of Itaú.
The standard he describes is exacting: solutions must address real, everyday problems and be direct, simple, and impactful. Technology is merely the enabler, never the main spectacle of the experience.
The market demand is pressing. A study conducted by Itaú in partnership with Grupo Consumoteca revealed that 80% of Brazilians want to better plan their financial lives and seek more effective decision support, while 65% want technology acting as a “financial copilot” to provide ongoing guidance.
“OpenAI’s models have played a critical role in delivering this experience because they operate reliably across such a broad portfolio of solutions.”
—Carlos Eduardo Mazzei, CTO, Itaú
ia.i: Unifying seamless service and advisory in one place
The journey began in 2024, when the bank launched its first Inteligência Itaú features tailored to specific financial needs. ia.i represents the next phase: a complete conversational layer, activated by text or voice, that brings interactions spanning checking accounts, credit, credit cards, insurance, and more into a unified environment—with seamless handoffs to specialized human support whenever needed.
It was precisely the demand for broad operational coverage that dictated the criteria for model selection.
“OpenAI’s models play a critical role in delivering this experience because they perform consistently across such a diverse portfolio of solutions,” Mazzei explains. “From the simplest scenarios requiring cost-effectiveness and low latency, to highly complex use cases demanding sophisticated reasoning and robust grounding to deliver the right answers, they handle it all.”
Deploying a diversified model matrix—rather than a single model—makes it possible to deliver a complete suite of solutions within a single product. At the heart of this experience lies “Customer DNA,” a financial context layer that helps the bank deeply understand the needs, habits, and life stages of approximately 70 million customers. Governed by established frameworks and responsible AI controls, GPT models serve as a core component, continually updating and enriching these profile representations to make interactions increasingly tailored and precise.
The solution is built on three pillars: interactivity, multimodality (blending text, voice, images, and tap interactions into a cohesive experience), and personalization.
Reaching roughly 300,000 customers in phase one allowed the bank to gain deep insight into how customers genuinely engage with the product.
“We were able to observe how customers began using the product more deeply and see how it directly benefits their day-to-day financial lives,” Mazzei says. It spans the entire user journey: from resolving questions about a transaction on a statement to offering investment guidance and easily conducting banking tasks over WhatsApp.
“Many financial products are inherently counterintuitive and require contextual clarity to simplify complexity, making people’s relationship with money much easier.”
—Carlos Eduardo Mazzei, CTO, Itaú
IARA: The backbone powering scale
Behind these intuitive customer-facing experiences lies a critical architectural decision. To run AI smoothly across the entire bank, Itaú built IARA—a governance and orchestration platform that provides teams with controlled access to diverse models in alignment with internal compliance policies. OpenAI is among the model providers available within this environment.
This architectural layer powers solutions across departments such as legal, automating document ingestion, analysis, and interpretation. Running on a repository of over 90 million documents and executing more than 2 million tasks per month, the solution has driven efficiency gains ranging from 20% to 80% across multiple departmental workflows.
Mazzei notes: “It is critical for the bank to ensure that this information is rigorously protected, kept within the bank’s internal environment, and routed to models in a secure, compliant manner—all while managing token efficiency to keep the entire setup economically viable.”
IARA tackles three key challenges at once: cost per token, response accuracy, and operationalizing responsible AI policies. In a heavily regulated industry like banking, this enables rapid innovation while maintaining complete control.
“When we talk about responsible AI at the bank, it is a discipline we approach with extreme diligence,” he says. “We must ensure that when customers interact with the bank, the system’s responses remain consistent with our values and culture.” This also includes mitigating any potential risk of bias in frontier models.
Responsible AI principles are embedded across the entire solution lifecycle—from development to continuous monitoring—designed to ensure security, transparency, ethical compliance, and regulatory adherence as the technology scales.
“IARA is an indispensable foundation, ensuring we maintain rigorous governance and observability within a highly regulated environment.”
—Carlos Eduardo Mazzei, CTO, Itaú
Embracing “AI First” from within
The bank’s “AI First” strategy doesn’t stop at…