BBVA Introduces a Common Framework to Scale AI Agents Across the Bank

BBVA is moving from individual AI projects toward a common framework for deploying agents across the bank. Presented at a meeting with analysts and investors, The Frame is intended to speed development while defining what each agent can access and do. The announcement pairs that shared foundation with plans to widen AI deployment in 2027.

Over the past year, BBVA concentrated on specific uses of AI in customer service, internal processes and employees’ work. Antonio Bravo, its global head of AI Transformation, said those projects revealed recurring needs: reusable components, shared functions and common management requirements. The Frame is the bank’s response to that overlap.

Reusable technology, specific limits

BBVA is building a shared, secure technology environment so teams can develop and deploy agents without starting from scratch each time. Common working processes, roles and responsibilities will sit alongside that foundation. The bank also wants to preserve technological independence, allowing it to work with different AI models, tools and platforms rather than tying the framework to one provider.

• Access: Each agent will have defined permissions for the systems and data it can use.
• Actions: Operational limits will distinguish what an agent can do autonomously from what requires human supervision.
• Measurement: Performance will be evaluated and monitored continuously, with cost and business impact measured systematically.

These controls are intended to operate under a common security, governance and compliance framework. BBVA’s stated goal is to build more agents faster while keeping them repeatable, manageable and subject to control.

Existing uses show the starting point

BBVA says assistants already filter more than half of the conversations its relationship managers receive through the app in Spain. Voice agents successfully resolve close to 90% of interactions in production pilots in Peru and Mexico. Those are results from existing applications; the framework is intended to make creating and managing further agents a shared process.

The 2027 roadmap goes beyond internal tools

At the same presentation, BBVA showed progress on Blue, its planned personal financial assistant. The bank describes it as available around the clock, able to handle multiple forms of interaction and equipped with personalized customer information. BBVA plans a gradual rollout to individuals and businesses in 2027, rather than announcing broad availability now.

The 2027 plans also include more AI tools for retail relationship managers and new capabilities across commercial banking, investment banking and risk. BBVA is extending agents into complaints and financial-crime prevention, too.

In software engineering, global head Carlos Casas set a separate ambition within the strategy: shorten release cycles from months to weeks. Agents would help coordinate and execute development, while people retain control over context, decisions and oversight. That faster release schedule remains an ambition, distinct from the productivity gains BBVA already reports.