Itaú, one of Brazil's biggest banks, has started rolling out ia.i, a chat and voice assistant that handles accounts, credit, cards and insurance in one place. For now it's live for roughly 300,000 customers, a small slice of a base the bank puts at about 70 million.
The account comes from OpenAI's own write-up, built around an interview with Itaú's CTO, Carlos Eduardo Mazzei. So treat the results as the companies' claims. No independent figures on usage or customer satisfaction are in the material we read, and we haven't heard from rival banks, outside analysts or Itaú customers to check them.
What ia.i does
Customers can type or speak, and the assistant is meant to mix text, voice, images and clicks in a single conversation. It can hand people off to human specialists when needed. Examples in the post range from a question about a line on an account statement to investment advice and transactions on WhatsApp.
Mazzei's stated goal is for the technology to be "almost unnoticeable." The post cites a study with Grupo Consumoteca: 80% of Brazilians want to organize their finances better, and 65% would like a kind of financial copilot. The post doesn't link the research itself, so we haven't seen it and can't say how it was run or who paid for it.
Behind it sits what Itaú calls Customer DNA, a layer of financial context about customers' needs, habits and life stages. GPT models help maintain that layer, under governance and responsible AI controls, the post says.
IARA, the control layer
The less visible piece may matter more to other banks. IARA is an internal platform that gives teams controlled access to several model providers under company policy, and OpenAI is one of them. Itaú says it targets three problems at once: cost per token, accuracy and responsible AI rules.
The bank's legal department is the showcase. According to the post, the system works across more than 90 million documents, runs over 2 million tasks a month, and saves 20% to 80% of effort in certain process stages. That range is wide, and the post doesn't say how it was measured.
Mazzei says it's essential that data stays within the bank's context and that models are used efficiently so the whole thing is "economically viable."
What's not yet known
The post gives no date for reaching all customers, no pricing impact and no error rates. It also doesn't say which GPT models handle which tasks. The first phase is described as a learning period.
Nor does it say how ia.i compares with assistants from rivals such as Bradesco or Banco do Brasil, or with those at banks abroad. We haven't verified those comparisons ourselves, so it's unclear what, beyond the IARA control layer, sets Itaú's approach apart.
Watch for the wider rollout timeline. Would you trust a bank's chat assistant with investment advice, or do you still want a person on the line?
Sources
Reported by the WattsUpNext desk from the sources linked below. Spot an error? Tell us at corrections@wattsupnext.com.
The WattsUpNext Brief
Get stories like this in your inbox.
One email each weekday, only the topics you choose. Real news, sourced, no fluff. Unsubscribe in one click.
