“The Real Battle in RegTech Isn’t AI — It’s Auditability” explores how AI and machine learning are transforming compliance, particularly as financial institutions deal with enormous volumes of communications and increasingly complex regulatory requirements. The discussion argues that building an AI model is no longer the biggest challenge. The real difficulty is ensuring that those models continue to work accurately over time, can be monitored, and can provide evidence for why a particular compliance decision or alert was generated. A key theme is model drift. As data, regulations, customer behaviour and communication channels change, an AI model can gradually become less effective. Rather than creating highly customised models for every customer, the approach discussed favours standardised core models that can be continuously monitored and improved using feedback, performance data and carefully curated training data. The podcast also highlights the needle-in-a-haystack problem facing communications compliance. Potentially problematic behaviour represents only a tiny fraction of the huge volume of legitimate communications. This makes false positives a major challenge. Compliance technology therefore needs to identify genuine risks without overwhelming compliance teams with unnecessary alerts. Another important issue is data quality. AI is only as good as the data used to train it. Smart data selection, labelling, validation and human oversight are therefore critical to building dependable compliance systems. The conversation also considers the role of large language models and generative AI. While LLMs offer significant opportunities, the discussion challenges the assumption that bigger models are automatically better. For highly specific compliance tasks, smaller, purpose-built models can sometimes be more efficient, cost-effective and easier to understand and debug. Ultimately, the podcast makes a strong case that trust will become one of the biggest differentiators in RegTech. Financial institutions don't just need an AI system that says, “this communication looks risky.” They need to understand why it was flagged, how the model reached its conclusion, how reliable the model is, and whether that decision can stand up to regulatory scrutiny.
The next generation of RegTech won't be won by whoever has the most impressive AI. It will be won by whoever can make AI trusted, explainable, measurable and auditable.
This is particularly important as compliance teams move from simply using AI to relying on AI as part of their regulatory decision-making processes. The technology needs to work not only in the laboratory, but consistently and defensibly in the real world.
Host: Mark Walker, Editorial Director at the Fintech Times
Guest: Laksmi Rippe, product marketing manager at Sis ID. Dr. Ruth Wandhöfer, Global Fintech 50 Influencer, Author, Keynote Speaker and Adviser; Women in Fintech Powerlist 2016/17/18 and Senior Leader 2020/21/22/23/24; WIBF Award of Achievement.