IBM and EY empower AI governance to transform institutional trust
As the oldest bank in Latin America, Banco do Brasil has long played a vital role in shaping the region’s financial landscape. With a legacy of innovation and a commitment to serving the public interest, the bank is recognized for pioneering advancements in banking automation, digital services and AI.
In keeping with their tradition of innovation, Banco do Brasil responded to the growing presence of AI not as a distant opportunity, but as an immediate imperative. As a public financial institution with regulatory duties and a public mission, the bank needed a clear strategy to use AI responsibly, transparently and at scale.
With the rapid adoption of AI across critical banking functions, they faced growing challenges related to trust, accountability and regulatory compliance.
To meet growing challenges, Banco do Brasil engaged trusted partners EY and IBM to codevelop an evolution of their governance model. Recognizing the magnitude of this transformation, Banco do Brasil took a bold step by committing to a robust AI governance strategy. EY was brought in to design and support the implementation of comprehensive frameworks tailored to the bank’s institutional context. And IBM provided the technological foundation through the IBM® watsonx.governance® toolkit to deliver automation, transparency and continuous oversight across the AI journey.
EY and IBM have played complementary roles in the evolution of AI governance at Banco do Brasil. While EY led the development of their Generative AI Lifecycle and Continuous AI Monitoring frameworks, IBM operationalized this strategy through watsonx.governance. These frameworks define clear roles and responsibilities throughout the AI lifecycle—from model design to deployment in critical environments—and introduce rigorous evaluation processes for foundational models, such as benchmarking and vulnerability assessments.
IBM has played a critical role in providing continuous supervision, explainability and compliance across traditional AI systems. The platform supports real-time monitoring, proactive alerts and customized metrics—consolidating a unified system of control and trust.
The integration of IBM’s tools with EY’s frameworks is currently being tested in generative AI (gen AI) use cases to help ensure alignment with compliance, performance and risk management requirements.
This synergy between strategic design and advanced technology empowered Banco do Brasil to scale AI ethically and with institutional confidence in a security-rich environment.
With strategic support from EY and IBM, Banco do Brasil now operates under a unified AI governance model. The bank has significantly strengthened their ability to manage AI initiatives with precision, transparency and control—validating every model deployed meets rigorous standards for ethics, security and performance.
Now serving a broad customer base with extensive AI lifecycle governance coverage, the bank is achieving measurable gains in oversight, agility and trust.
Key outcomes include:
This journey is helping position Banco do Brasil as a technology-forward institution—transforming innovation into reliable, purpose-driven solutions with a clear commitment to the future.
Banco do Brasil is the oldest and second-largest bank in Brazil, as well as the second largest in Latin America. They have been one of the four most profitable Brazilian banks since the year 2000 and hold a strong leadership position in retail banking.
EY ,formerly known as Ernst & Young, is among the world’s largest professional services firms and part of the Big Four. Headquartered in London, they operate in more than 150 countries, offering assurance, consulting, strategy, tax and transaction services. Recognized for their diverse teams and expertise, EY holds a leading position in helping organizations transform, manage risk and grow in complex markets.
© Copyright IBM Corporation 2025. IBM, the IBM logo, and watsonx.governance are trademarks of IBM Corp., registered in many jurisdictions worldwide.
Examples presented as illustrative only. Actual results will vary based on client configurations and conditions and, therefore, generally expected results cannot be provided.