Finance / Banking
Itaú Unibanco: Automating legal document reading and analysis on the IARA AI governance platform
- Company
- Itaú Unibanco
- Country
- Brazil
- Adoption stage
- In operation
- Source published
- Date basis
- The date the source was published. It can differ from the date adoption started.
- How the source was checked
- Read the full source text
The work problem
According to a customer story OpenAI published on October 16, 2026, to operate AI at the scale of a bank in banking, one of the most regulated sectors in Brazil, Itaú had to solve three problems at once: cost control per token, accuracy of responses and application of responsible AI policies. Carlos Eduardo Mazzei, CTO of Itaú, said the information must be preserved and remain within the bank’s context, and the bank must access models in a secure, governed way and understand token-use efficiency for the entire solution to be economically viable for the bank.
Technology and data
Itaú built IARA, a governance and orchestration platform that gives teams controlled access to different models in line with internal policies, and OpenAI is among the model providers available in this environment. The legal department’s solution runs on IARA, automates the reading, analysis and interpretation of documents, and operates on a repository of more than 90 million documents. IARA provides observability, cost control per token and enforcement of responsible AI principles, and these principles apply across the entire lifecycle of solutions, from development to continuous monitoring.
Results
According to the customer story, the legal solution executes more than 2 million tasks per month on more than 90 million documents, with efficiency gains of 20% to 80% in stages of the department’s processes. The customer story says IARA makes it possible to move quickly while maintaining control even in a highly regulated sector, and Mazzei said IARA is a fundamental platform that ensures governance and observability in a highly regulated environment.
Limits and open questions
The customer story does not disclose which process stages and criteria the efficiency gains were measured against, the measurement period, or how people review the results. The figures appear in a customer story published by the vendor OpenAI.
Sources
- How Itaú is building generative AI for its entire customer baseopenai.com, Accessed
Compiled from public sources. These are not results from ATF Works customers.