BBVA
With more than 125,000 employees across multiple countries, BBVA needed to turn generative AI from a limited license pilot into a governed, reusable enterprise capability. The bank also had to manage use-case discovery, legal and compliance review, employee skills, and output quality—not merely count activated accounts.
BBVA first distributed 3,000 ChatGPT Enterprise licenses across organizational levels, allowing frontline employees to build custom GPTs under joint legal, compliance, and IT-security governance. It later expanded access to ChatGPT and Gemini, created a funnel of more than 8,000 active use cases and about 700 strategic candidates, and built an internal network of roughly 750 “Wizards,” formal training, and a 90,000-person community of practice.
Employees closest to each workflow identify opportunities and review outputs. Credit analysts monitor model errors; a nine-lawyer team maintains legal knowledge; legal, compliance, and IT security define usage boundaries; and internal Wizards coach teams and spread working practices.
- Give employees at multiple levels direct access so real workflow opportunities surface
- Publish high-value employee-built GPTs in an internal store
- Separate active experiments from strategically important use cases
- Involve legal, compliance, security, and domain reviewers
- Use champions, training, and community support to sustain adoption
- Enterprise-wide tool access needs a use-case funnel, not only login metrics
- Frontline employees are often best positioned to identify workflow opportunities
- An internal champion network can shorten the organizational learning curve
- Regulated workflows must keep professional judgment and error monitoring with people
BBVA reports that more than half of its workforce uses generative AI weekly. The bank has identified more than 8,000 active use cases, about 700 of strategic importance, and estimates roughly three hours saved per employee per week. Its employee assistant handles more than 34,000 monthly queries over a knowledge base of 2,500+ documents.
- The roughly three hours per week figure is BBVA’s estimate for repetitive-task savings, not an audited per-employee productivity measure.
- The 8,000+ figure covers active use cases; about 700 are considered strategically important or high potential.