Telefónica Tech implements a generative AI virtual assistant for Bizum, achieving 90% accuracy and reducing email support tickets by 30% in just three months.

The integration of generative artificial intelligence into customer service is moving past the experimental phase and into high-stakes financial environments. Telefónica, through its specialized business unit Telefónica Tech, has successfully deployed a new virtual assistant for Bizum, Spain's leading instant mobile payment service. The project serves as a benchmark for how rapidly a large-scale enterprise can transition from legacy, pre-programmed chatbots to fluid, natural language interfaces.

Rapid deployment via the GenAI Platform

One of the most striking aspects of this implementation is the speed of execution. Telefónica Tech completed the development and deployment of the assistant in just three months. This was made possible through the use of the Telefónica Tech GenAI Platform, a framework designed to allow organizations to integrate generative AI capabilities without the traditional long-term development cycles associated with custom software.

Carlos Martínez, Head of Data and AI at Telefónica Tech, emphasized that the goal was to move away from rigid, pre-programmed responses. By utilizing generative AI, the system can handle a broader spectrum of queries with a level of nuance that previous iterations of customer service bots lacked. This shift allows Bizum to provide a more human-like interaction while maintaining the scalability of an automated system.

Quantifying the impact on operational efficiency

The transition to a generative model has yielded immediate, measurable results in terms of operational overhead and user satisfaction. According to data released by the companies, the virtual assistant has achieved a 90% accuracy rate in its responses. This precision is critical in the fintech sector, where incorrect information regarding payments or security can lead to significant user distress and regulatory scrutiny.

Beyond accuracy, the tool has significantly altered the volume of incoming support requests. Bizum has reported a 30% reduction in the number of contact requests sent via email for the resolution of doubts or requests for assistance. By diverting these queries to an automated, 24/7 channel, the company reduces the burden on its human staff, allowing them to focus on complex cases that require manual intervention.

Multi-channel integration and user accessibility

To ensure maximum reach, the assistant is not confined to a single portal. It is fully integrated into the Bizum website and the WhatsApp Business channel. This multi-channel approach meets users where they already spend their time, reducing friction in the support process. The assistant provides continuous support, operating every hour of every day, which is a significant upgrade over traditional business-hour support models.

The technical architecture is described as scalable, utilizing multiple orchestrated assistants. This means the system can route a query to a specific sub-assistant specialized in a particular area, whether it be a general inquiry, a technical incident, or a security concern. To prevent the common AI issue of hallucinations, the system relies on official documentation provided by Bizum as its primary knowledge base, ensuring that the output remains consistent with company policy.

Fighting fraud and managing bank relations

A critical function of the new AI tool is its ability to alert users about potential scams. In an era of increasing social engineering attacks on mobile payment users, having an AI capable of identifying and warning users about fraudulent patterns in real-time is a vital security layer. The assistant is specifically designed to provide fast and specific responses regarding potential scams, acting as a first line of defense for the consumer.

Furthermore, Telefónica has addressed the fragmented nature of the payment ecosystem. Since Bizum operates across various banking institutions, many queries are actually the responsibility of the user's specific bank rather than Bizum itself. To solve this, Telefónica developed a specialized API that queries bank contact details. If a payment-related issue is identified as a bank-level problem, the assistant can automatically provide the correct bank details or even send an email on behalf of the user to the relevant institution, streamlining the resolution process.

At Bizum, we are committed to innovation as a lever to continue improving the experience for all users... It is not innovation for innovation’s sake, but technology with a purpose. — Lorena Poza, Head of Communications, Marketing and Sustainability at Bizum

Strategic implications for the fintech sector

The collaboration between a telecommunications giant and a payment leader highlights a growing trend: the convergence of connectivity and financial services. By leveraging the generative AI capabilities of a tech provider, Bizum has effectively modernized its digital front-end without needing to rebuild its core financial infrastructure. This demonstrates that the primary value of GenAI in the short term is not the replacement of core services, but the optimization of the interface between the service and the end-user.

The success of this project suggests that other financial services providers will likely move toward similar orchestrated AI architectures. The ability to reduce email volume by nearly a third while increasing accuracy to 90% provides a compelling business case for the rapid adoption of GenAI platforms in highly regulated industries.

Global perspective: Lessons for US and UK enterprises

For entrepreneurs and executives in the USA and UK, the Bizum-Telefónica case provides a blueprint for AI implementation in the financial services sector. In the US market, where customer service expectations are exceptionally high and labor costs for support staff are significant, a 30% reduction in manual tickets represents a substantial bottom-line gain.

From a regulatory standpoint, US firms must navigate a patchwork of state-level privacy laws (such as the CCPA in California) and federal guidelines on AI transparency. The Bizum model of using a closed knowledge base—relying on company-provided documentation rather than open-web training—is a strategic move to mitigate the risk of AI hallucinations and ensure compliance with financial accuracy standards. In the UK, where the approach to AI regulation has historically been more pro-innovation and less prescriptive than the EU's AI Act, the focus remains on safety and consumer protection. The integration of fraud detection within the AI assistant aligns closely with the UK's focus on combating authorized push payment (APP) fraud.

The key takeaway for global business leaders is the shift toward orchestrated AI. Rather than a single, monolithic chatbot, the use of multiple specialized assistants managed by a central orchestrator allows for higher precision and better risk management. For companies operating in the US and UK, adopting this modular approach can accelerate deployment while maintaining the strict control required for financial and personal data handling.

FAQ

---

Sources: Larazon, Telefonica, Eleconomista ·