Innovation in Risk Management: Three AI-Enabled Solutions for Swiss Banking

September 02, 2025, by Reik Leiterer, data innovation alliance

Did you know the RiskON Event? If not, then it’s about time, as it has developed into one of the most innovative formats in this field in Switzerland.

RiskON is an innovative hackathon organized jointly by University of Zurich – Master of Science UZH ETH in Quantitative Finance, UZH Innovation Hub and the N9 House of Innovation/sminds AG. RiskON aims to promote risk management in Switzerland through collaboration between students, industry experts and financial institutions. Interdisciplinary teams of enthusiastic young master’s students from all Swiss universities and experts will work on real-life challenges in banking risk management. The goal is to co-create and develop concrete solutions, such as proofs of concept (PoC) or minimum viable products (MVP), which can then be further developed in practice. The focus is on current topics such as financial risks, non-financial risks and digital risks.

Driven by advances in technology, banks are encountering increasingly complex and dynamic risk landscapes. In response, a group of Swiss banks defined in 2024 key challenges in three areas, i.e. operational risk classification, internal fraud detection, and regulatory compliance automation. For each challenge, innovative solutions were found that directly support the financial sector’s need for enhanced efficiency, precision, and adaptability – highlighting the power of academic-industry collaboration in delivering practical, forward-looking innovations for the banking sector. Read the full report and learn more about:

• An AI-powered framework for operational risk classification and forecasting: by integrating data cleaning, Large Language Models (LLMs), and Transformer-based time-series analysis, the approach significantly improves classification accuracy and enables timely, comprehensive risk reporting. For banks, this represents a scalable tool to reduce losses and strengthen operational oversight.
  
• Internal fraud detection: addressing this complex topic by the application of an Isolation Forest model to identify anomalous transaction behavior – particularly among relationship managers. Requiring no labeled data and offering a customizable alert threshold, the solution enables compliance teams to prioritize investigations and better allocate resources, enhancing fraud prevention capabilities.
  
• A multilingual compliance tool: based on DistilBERT and designed to automate regulatory document comparison and change detection. Tested on English and German texts, the model demonstrates strong potential to reduce compliance effort and translation costs – addressing a growing challenge for banks operating across jurisdictions.

Collectively, these solutions illustrate how targeted AI applications can modernize risk management, strengthen fraud detection, and streamline compliance processes – while preserving essential human oversight. By improving the agility, reliability, and cost-efficiency of risk practices, Swiss financial institutions can further solidify their position as global leaders in innovation and regulatory excellence. By the way, RiskOn 2025 is underway, addressing challenges such as “Quality Assurance for Client Contact Notes — How can AI help?” or “KYC vs. Adverse Media — Can AI match identities across sources?”. We are looking forward to the results!