BBVA moved from isolated experiments to making artificial intelligence a way of working across the entire bank. How did they do it without losing control or compromising security? Here I explain it to you, with clear numbers and practical lessons you can apply in any team.
How BBVA took AI from pilot to practice
The strategy was simple in concept but disciplined in execution: don’t leave AI as an isolated technical project, but turn it into a core capability of the bank. In the words of their leadership, the idea was to integrate AI into the business strategy, not push it to the side.
Quick, measurable results:
- Initial rollout to 3,000 employees and a fast rise to 11,000.
- About 3 hours saved per employee per week.
- 83% weekly active use among employees.
- Workflow efficiency improvements of more than 80% in tests.
- Over 20,000 Custom GPTs created internally; roughly 4,000 used frequently.
A concrete example: in Peru, an internal assistant cuts query time from 7.5 minutes to 1 minute — a nearly 80% drop in time per query. That’s not theory; it’s time recovered every day.
Decisions that made the difference
BBVA built its program on three pillars: trust, governance, and structured learning. It wasn’t about stopping innovation, but about giving it a safe channel.
- Visible leadership: 250 leaders, including the CEO and the chairman, received hands-on training. Can you imagine the bank’s president using the same tool you use? That creates legitimacy.
- Governance from day one: security, legal and compliance aligned from the start, which allowed scaling without surprises.
- Safe spaces to experiment: rather than pushing experimentation into the shadows, they brought it onto an approved platform that reduced the risk of ‘shadow AI’.
- Training and democratization: focused training for leaders and controlled access for the workforce.
They also implemented ChatGPT Enterprise and encouraged business areas to create their own custom assistants. The idea was simple: the people who know the process improve it, not a disconnected central team.
What this means for employees and customers
When teams see real value, adoption grows on its own. Some employees joke that if you take the tool away, they’d have to find a new job; that’s how sticky the experience was.
For customers, AI is moving from individual productivity to automating flows, operating systems, and customer-facing channels. One example is Blue, the bank’s digital assistant built with OpenAI models.
Practical lessons for other organizations
- Start at significant scale: piloting with a tiny sample doesn’t always reveal clear signals; scaling quickly gives real evidence.
- Align security, legal and the business from the beginning to accelerate, not block, progress.
- Create a safe space to experiment and bring experimentation into the light.
- Train leaders; their practical use drives cultural adoption.
- Empower those in daily operations to build tools: they know what works.
Does it sound like too much effort? It doesn’t have to be. BBVA shows that with governance and trust, adoption can be accelerated and measured.
BBVA is no longer in the pilot phase: it’s extending AI into products, operations, and customer service, with a plan to keep expanding access to tools and capabilities over time. As the program leads say: once adoption starts, it tends to accelerate.