BBVA is now using ChatGPT to streamline legal queries and marketing processes

In less than a year of use, the ChatGPT Enterprise licenses that BBVA has been handing out to its employees are already succeeding in making the teams more innovative and productive and better at what they do.

A case in point would be the legal department. It has an advisory team of nine lawyers who provide support to the branch managers in Spain by helping to resolve more than 40,000 legal queries received from customers per year. The response process is time-consuming and means consulting a diverse range of documentation, from product specification sheets to numerous internal and external regulations.

To streamline this process, the advisory legal team has created a GPT—a customized wizard within ChatGPT tasked with solving a specific objective—containing all the consultation documents. Following an extensive review and quality control process, this ‘tenth member’ of the team is able to come up with the answers more rapidly, and more extensively. All responses undergo a human review before being sent to the branch managers. As a result, BBVA is now able to respond to a higher number of legal queries from customers within 24 hours, and is also improving the quality of the first response. This also frees up time for the advisory team, which can then spend it on more strategic functions, seeing as though a task that used to take hours can now be completed in a matter of minutes.

ChatGPT has also become a key ally of the Group’s marketing departments. In Colombia, for example, a GPT that helps write creative copy for advertising campaigns has succeeded in reducing the time it takes to create and develop these actions. In parallel, it has made ads more effective by almost doubling the number of times they appear in search engines. To achieve this, the team has set up this customized GPT to work with specific instructions and contextual documents, allowing it to learn which texts performed best in previous campaigns. Armed with this knowledge, the GPT proposes messages that include the keywords most searched by users while at the same time adopting an empathetic tone focused on the usefulness of the product or service they are talking about. As a result, more people are now seeing and clicking on the ads.

“Although these two use cases in particular were born in Spain and Colombia, they are completely transferable to other countries,” says Guillermo Vieira, head of Training and Engagement on BBVA’s Artificial Intelligence Global Adoption team. “One of the strengths of our strategy in generative AI is that we have embraced a global mindset to bring it to the entire Group right from the get-go.”

Thanks to this global approach, a GPT was conceived that is now impacting how data is used at BBVA. The Group operates a Single Data Model, in which it centralizes all the information related to all its departments and operations, thus including financial, sales and marketing, logistical, legal and technological data. This single model contains an immense volume of information, so navigating it is not always quick and easy. However, it is essential for product development, analytics and the bank’s artificial intelligence strategy, as well as for employees throughout the organization.

The team responsible for data governance at BBVA has developed a GPT which, through a conversational interface, helps employees quickly locate not only the type of data they are looking for, but also who owns it and in what other applications it has been used. This ultimately shortens the time it takes to locate the most useful data for each user, which were previously more scattered and harder to locate.

Numerous use cases are also being created to optimize code development. The bank’s software developers are starting to rely on ChatGPT and other tools such as GitHub Copilot to improve the quality of existing code, write code for new features and generate documentation on how they work. It can also be used to port or translate code from one programming language to another, leading to significant time savings when modifying the technologies or platforms that support the bank’s applications and operations. The use of generative AI among developers has helped to improve the code they deliver and shorten task execution time, which positively impacts their productivity.

The key is to create custom GPTs for specific tasks: “Training GPTs for specific tasks is an essential task that requires time and resources up front, but generates a big return because it is a very effective way of bringing out their full potential,” explains Isabel Pérez del Caño, head of Portfolio and Performance on the AI Global Adoption team.

In just seven months, BBVA employees have launched more than 3,000 GPTs, of which over 900 (a number that is growing every day) have been flagged as cases of strategic interest and with upscaling potential. Notably, of these 3,000 or so GPTs, many of them are accessible to any employee with a ChatGPT license, as they may be useful across the entire organization because of the type of task they help solve. To facilitate their localization, the AI Global Adoption team has created a proprietary GPT Store; a library that was also created by a GPT trained to locate those solutions that can be used by any user.

This team manages the use cases generated by users, provides them with technical support and a community of practice where they can exchange information with each other. It also offers workshops and training courses to all employees (whether or not they have a license). To make this happen, the team relies on a network of ‘champions,’ meaning those who oversee all of these activities across all business areas and geographies.

The Global AI Adoption team also analyzes usage metrics to help spot ‘wizards’ or users who, in addition to being very active, drive the adoption of the technology in a more direct way among the wider user community. Together with these champions and wizards, the team is currently conducting an impact assessment of the use cases by identifying their upscaling potential. It also happens to work alongside the legal, compliance and IT security teams to ensure that all uses are safe and responsible. 

Starting in 2025, it will begin to evaluate specific impact metrics based on five criteria: quantity (those use cases where ChatGPT increases the number of tasks executed), improved performance quality, innovation drive, time reduction and direct economic impact (cost savings or increased revenue).

The implementation of ChatGPT is part of BBVA’s broader strategy to explore the impact that generative AI can have on its transformation. This strategy ranges from alliances with external companies (such as this particular agreement with OpenAI), to in-house developments, with the aim of making its organizational processes more efficient, helping its employees do a better job and, ultimately, improving the experience of its customers.