Next upDPTV HighlightsEpisode 1 · From SF Tech Week to TechCrunch Disrupt
News

Itaú brings governed generative AI to 300,000 banking customers

Itaú says about 300,000 customers can use its ia.i conversational interface, while IARA manages model access, monitoring, token costs and responsible-AI policies.

D
Oct 17, 2026 · 2 min read

Itaú has opened its ia.i conversational banking interface to approximately 300,000 customers in an initial phase. The bank aims to reach its entire customer base by the end of 2026, according to OpenAI in a case study detailing the rollout. The figure reflects customers with access, not active users; Itaú has not disclosed usage, retention or transaction volumes.

The interface sits inside Itaú’s app and accepts text or voice. It covers accounts, credit, cards, insurance and other services, and can connect customers with specialized human support. At launch, Itaú said ia.i could answer questions, help customers understand their finances and the bank’s products, and support Pix payment journeys. Product simulations, sign-ups and recurring analyses of income, expenses and financial habits are planned for later stages.

Itaú had already outlined the launch in its announcement of the ia.i interface. An August earnings release filed with the U.S. Securities and Exchange Commission separately confirmed that ia.i had launched in the Itaú SuperApp as part of the bank’s AI First strategy.

Behind the interface is IARA, Itaú’s centralized governance and orchestration layer. OpenAI says the platform gives internal teams controlled access to different AI models under the bank’s policies, with OpenAI among the available providers. Itaú Chief Technology Officer Carlos Eduardo Mazzei said using a range of models lets the bank handle cost- and latency-sensitive work differently from tasks that require deeper reasoning and grounding.

According to the case study, IARA provides observability across model access and addresses three operational concerns: per-token cost control, response accuracy and enforcement of responsible-AI policies. Itaú says those policies cover development through continuous monitoring, with goals spanning safety, transparency, ethics, regulatory compliance, data protection and reduced bias risk. The bank and OpenAI did not disclose routing rules, token budgets, evaluation thresholds, incident rates or measured cost savings. The opened sources also include no independent audit of those controls.

OpenAI says an IARA-supported legal system operates across more than 90 million documents and runs more than 2 million tasks each month. It attributes efficiency gains ranging from 20% to 80% in parts of the legal department’s processes to Itaú. The opened sources do not independently substantiate those figures.

More news