How Itaú is building a Generative AI foundation for 70 million customers

The generative artificial intelligence revolution has moved past being a futuristic promise to become the operational bedrock of the largest financial institutions in Latin America. Itaú Unibanco, Brazil's largest private bank, is leading this transition with an ambitious goal: to bring its new AI-based conversational interface, dubbed ia.i, to its entire base of approximately 70 million customers by the end of 2026. Supported by the proprietary governance platform IARA, the institution has adopted an "AI First" strategy that ranges from end-consumer service to the optimization of critical internal processes.

Creating a close, personalized experience for millions of customers in a highly regulated sector requires more than just implementing large language models (LLMs). It is a matter of infrastructure, cost control, and, above all, a focus on practical utility. The goal is not to showcase the technology, but to make it disappear into the background, delivering direct and memorable solutions to the user.

The challenge of invisibility: AI at the service of simplicity

For Carlos Eduardo Mazzei, CTO of Itaú, the standard set for the bank's use of AI is extremely demanding. Technology must act as the means, never as the end.

"Much of this is about doing something simple, direct, and useful, delivering a direct benefit to the customer and making the technology almost imperceptible," explains Mazzei.

The backdrop to this initiative responds to a real demand in the Brazilian market. A study conducted by Itaú in partnership with the Consumoteca Group revealed that:

• 80% of Brazilians want to better organize their financial lives and seek more effective support for decision-making.
• 65% would like to have technologies that act as a sort of "financial copilot," providing continuous and personalized guidance.

To meet these expectations, the bank needed a solution that went beyond traditional decision-tree-based chatbots, evolving into a true understanding of each user's context.

From isolated service to the ia.i platform: context and background

Itaú's journey with conversational AI did not happen overnight. The foundation was launched in early 2024, when the bank introduced the first capabilities of "Inteligência Itaú" focused on specific financial needs. ia.i represents the maturity stage of this vision, consolidating itself as a complete conversational layer, officially launched in July 2026.

Activated by text or voice, the platform brings together interactions related to the following in a single environment:

• Checking account and invoice inquiries.
• Credit lines and financing.
• Card management and insurance contracting.
• Investment advisory.

ia.i allows for multimodality — operating with text, voice, images, and clicks in the same experience — and is deeply connected to what the bank calls "Customer DNA." This layer of financial context helps the AI understand the needs, habits, and life stage of each of the bank's 70 million customers, progressively enriched by models from the GPT-5.x family by OpenAI, among others.

In its initial phase, ia.i served approximately 300,000 customers, allowing the institution to learn about the real relationship between the consumer and the product. However, growth has been exponential. According to Milton Maluhy, CEO of Itaú Unibanco, in an interview in October 2026, the platform already has 3 million active customers, with a firm goal to reach 100% of the base by the end of the year.

IARA: The governance platform behind the scale

Operating artificial intelligence at the magnitude of a bank requires a robust architecture. The proliferation of language models has led large companies to operate with multiple simultaneous options, such as those from OpenAI, Gemini (Google), and Claude (Anthropic). The risk inherent in this ecosystem is the loss of financial and operational control.

As warned by the bank's technology leaders, the cost of AI can explode without proper management, resembling the challenges faced during the migration to the cloud. To circumvent this, Itaú developed IARA, a proprietary governance and orchestration platform.

IARA is the invisible engine that democratizes the use of AI among the bank's internal teams, solving three central problems simultaneously:

1. Token cost control: Ensuring the economic viability of the project on a massive scale.
2. Response accuracy and hallucination mitigation: Fundamental in financial transactions.
3. Application of Responsible AI policies: Alignment with the bank's culture and mitigation of biases present in frontier models.

"IARA is a fundamental platform because it ensures that we can have governance and observability in a highly regulated environment," highlights the institution's CTO.

The practical results of this architecture are already visible in Itaú's legal department. The platform has automated the reading, analysis, and interpretation of documents. The current repository holds more than 90 million documents, where IARA executes more than 2 million tasks monthly, generating efficiency gains that range between 20% and 80% in process stages.

The "AI First" strategy from the inside out

Itaú's technological implementation extends far beyond mobile app interfaces. The same intelligence that supports ia.i is being made available to specialists, relationship managers, and customer service teams.

The scope of internal use varies from automating trivial tasks to performing highly complex work. Direct benefits include:

• Instant summaries before customer interactions.
• Preparation of meetings and outreach scripts.
• Rapid understanding of specific support problems.
• Development of financial plans and alternatives that previously consumed hours of manual labor.

The bank's internal expectations are bold. Itaú projects generating up to three times more time available for customer relationships, enabling service that is twice as fast, and making conversations up to three times more effective. By giving time back to teams, AI frees employees to focus on what really matters: human connection and value creation.

Impact on the financial sector and the Latin American market

Itaú Unibanco's move sets a new pace for the banking sector in Latin America. Historically, traditional banks have struggled to modernize their legacy architectures and compete on equal footing with the usability offered by fintechs and neobanks. By leveraging advanced Generative AI models in a governed way, Itaú is not only modernizing its infrastructure but redefining the standard for user engagement.

The pioneering effort of placing an AI-based conversational platform at the center of the financial journey — including innovations like Pix operations via WhatsApp and a digital bank for small and medium-sized businesses fully integrated with generative tools — signals to the market that AI has ceased to be an R&D project to become the core of commercial operations.

What to expect next

As the adoption of ia.i advances to cover the 70 million customers projected for the end of 2026, we can expect a few clear trends:

• Predictive hyper-personalization: AI will move beyond just responding to commands and begin to predict customer needs, proactively suggesting investment reallocations or alerting about abnormal spending patterns.
• Expansion of multimodality: Financial transactions will be able to be initiated and completed fluidly through voice commands on smart home assistants or wearables securely integrated into the bank's ecosystem.
• Standardization of LLM Governance: The architecture developed on the IARA platform is expected to inspire other major players in the financial market to build their own multi-platform orchestration layers (OpenAI, Google, Anthropic), seeking to avoid technological "lock-in" and optimizing the cost-per-token ratio according to the complexity of each task.

Frequently Asked Questions

1. What is ia.i and who can use it? ia.i is Itaú's new conversational Artificial Intelligence platform, capable of interacting via text or voice for financial inquiries and transactions. Launched with 300,000 initial users, it is already used by 3 million customers, with progressive expansion to 100% of the base by the end of 2026.

2. Does Itaú's platform use only ChatGPT technology? No. Although the infrastructure extensively uses OpenAI models (such as the GPT-5.x family) for complex reasoning and the creation of "Customer DNA," the bank adopts a multi-model strategy managed by the proprietary IARA platform, choosing the ideal engine according to the cost and latency of each interaction.

3. How does the bank ensure data security and privacy with AI? Through the IARA platform, which acts as a governance and control layer. All interactions occur within the bank's secure environment, under strict Responsible AI policies that ensure information is not leaked and that responses are aligned with financial sector regulations and the institution's values.