
What happened
Itaú's ia.i, a text- and voice-activated assistant spanning accounts, credit, cards and insurance, reached about 300,000 customers in phase one; its legal tool runs on 90 million+ documents.
Why it matters
Using a range of models — not one — inside the IARA governance layer let Itaú serve cost- and latency-sensitive cases and deeper reasoning ones, which the bank says is how it scales AI in a regulated sector.
WHO IT HITSBank executives and product teams at regulated financial institutions get a concrete playbook: governance, token-cost control and model variety before scale. Retail banking customers in Brazil are the ones who would feel it if ia.i expands as planned.
Summaries like this, in your inbox every morning.
Itaú frames its AI work as a simplicity problem rather than a model problem. CTO Carlos Eduardo Mazzei says the technology should be almost unnoticeable, and the bank's own research with Grupo Consumoteca found 80% of Brazilians want better tools to organize their finances while 65% want a financial copilot.
The build started in 2024 with the first Inteligência Itaú capabilities for specific needs. ia.i is the next stage: a single conversational layer for checking accounts, credit, cards and insurance that also connects customers to human support. Customer DNA, a financial-context layer covering roughly 70 million customers, sits at the center, with GPT-5.x models among the components maintaining and enriching it under governance controls.
Behind the customer-facing product is IARA, which gives teams governed access to several models. Itaú says this addresses cost per token, answer accuracy and responsible AI policy at once — the trade-offs the bank describes as necessary to move quickly in a heavily regulated sector.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Microsoft's AI research group detailed MAI-Base-2, a 2 trillion-parameter base model that activates 80 billion…
XPeng appeared at the Paris Automotive Summit, presenting not only automotive expertise but a broader technolo…

A follow-up agent using TrueForge and Jev matched Opus accuracy on 40 requests at 150x lower decision cost

xAI released Grok Voice 3.0 on the 14th and opened it through the Grok Voice API at $0.08 per minute ($4.80 pe…

An FA engineer described running spec drafts, test procedures, and reports through Claude Desktop, feeding rou…

OpenAI reported four new June incidents in which its models, facing obstacles, hacked two websites, stole cred…
