Using GPU-Optimized Software to Shorten the Feedback Loop in AML and Fraud Models by an Order of Magnitude

At bunq, we’re at the forefront of using AI to make our users’ lives easy. From detecting fraud and money laundering to online marketing, we use artificial intelligence to both improve our product and ensure our users’ safety. For years, we’ve been using AI to detect potentially fraudulent transactions thanks to a transaction monitoring system built by our talented engineers. It dramatically reduces false hits by a factor of 2.5, compared to a “rule-based” approach often used by traditional banks. Moreover, it’s completely scalable (which is important for a growing scale-up!).

While traditional rule-based systems tend to be rigid and lack scalability, generative AI opened possibilities for creating a more dynamic, robust, and intelligent model. Recently we launched Project Finn, our own large language model that sets a new industry standard. Assembling a multidisciplinary team that worked in sprints, we achieved our primary goals in less than a year. It's been a challenging but incredibly rewarding journey. In user support, our AI-powered chatbot manages over 60% of all user tickets. It’s not just an automated response generator — it employs generative AI to understand the context, resolve queries, and even offer additional unsolicited but helpful tips. This nuanced approach has resulted in a marked increase in user satisfaction.

Bunq’s marketing also received a generative AI overhaul. Not only can we analyze consumer behavior more efficiently, but we can also predict future behaviors and tailor marketing campaigns accordingly. The results are in the numbers: higher conversion rates and an uptick in customer engagement metrics.