How Sicoob Scaled Open-Source AI to Empower 60,000+ Employees

A Monday breakout session at WCUC 2025 gave attendees a behind-the-scenes look at how Sicoob, Brazil’s largest cooperative financial institution, rolled out a secure, enterprise-grade AI solution using open-source large language models like Llama, Mistral, Phy 4 and DeepSeek.

Integrated directly into the organization’s business platform, the tool empowered over 60,000 employees to interact with internal documents in natural language, making it easier to extract insights, generate summaries and support decision-making at scale.

With more than 8.5 million members, over 4,600 branches and a presence in every Brazilian state, Sicoob's success highlights the real-world potential of responsible AI in complex cooperative systems.

"If our members are able to use face recognition, they'll be able to show that they actually or did not authorize transactions. We'll also be able to tell if they've been victims of fraud,” said Edson Lisboa, Head of Systems, Artificial Intelligence and Analytics with SICOOB. "Sometimes we face problems and we use AI to generate reports. This is how we can improve our processes."