Velaw-Revolutionising Compliance with AI
This project is a collaboration between the BFH AIRS team and Velaw AG to develop AI-powered tools that transform Swiss legal and regulatory texts into structured, auditable compliance controls.
Fiche signalétique
- Départements participants Gestion
- Institut(s) Institute for Applied Data Science & Finance
- Unité(s) de recherche Applied Data Science
- Champ thématique stratégique Champ thématique "Transformation numérique centrée sur l'humain"
- Organisation d'encouragement Innosuisse
- Durée (prévue) 01.07.2025 - 21.11.2025
- Direction du projet Prof. Dr. Branka Hadji Misheva
-
Équipe du projet
Prof. Dr. Lucia Gomez Teijeiro
Julius Kooistra - Mots-clés Artificial Intelligence; RegTech; Legal AI; Compliance Automation; Risk Management; Swiss Financial Regulation; Explainable AI; Large Language Models; Multi-Agent Systems; Legal Knowledge Graph; Audit
Situation
Financial institutions face growing regulatory complexity and increasing pressure to demonstrate that compliance processes are transparent, traceable, and auditable. Legal and regulatory requirements are often written in long, complex documents with hierarchical structures, cross-references, and domain-specific terminology. Translating these requirements into operational compliance controls is still largely manual, time-consuming, and difficult to standardize. The Velaw project was established as a collaboration between the Applied AI Research and Solutions (AIRS) team at Bern University of Applied Sciences and Velaw AG. Together, the partners explored how advanced AI methods can support legal document understanding, regulatory risk management, and compliance automation within Velaw’s existing proprietary platform - VELA. The initial focus was on Swiss financial market regulation and on the feasibility of extracting, interpreting, and structuring legal obligations into machine-readable compliance artifacts. The work was positioned as an early-stage feasibility validation at TRL 2-3.
Approche
The project followed an iterative collaboration between the BFH AIRS team and Velaw, combining academic expertise in trustworthy AI with practical RegTech implementation experience. BFH AIRS contributed expertise in explainable AI, multi-agent large language model systems and AI governance, with a strong focus on transparency, traceability, and auditability. A central activity was the conceptual design and prototypical development of a Baseline Compliance System. Namely, an agent was designed to process legal texts and transform them into structured compliance artifacts. The work included legal text extraction, preservation of document hierarchy, cross-reference resolution, and the creation of a structured legal knowledge base. Based on this foundation, the project explored how legal obligations can be linked to regulatory risks, mitigating controls, and structured control questions that can be reviewed, audited, and further developed within Velaw’s RegTech environment.
Résultat
The project delivered a validated proof of concept for automated legal text extraction and structuring. The system is capable of processing Swiss laws and ordinances while preserving the hierarchical structure of legal documents, including chapters, articles, and paragraphs. It also resolves cross-references between provisions, enabling more reliable navigation and interpretation of complex regulatory texts. A structured legal knowledge base was created to represent relevant Swiss financial market regulation in a unified data structure. This knowledge base captures both hierarchical and referential relationships and provides the basis for a scalable legal knowledge graph. Building on this, the AIRS team and Velaw developed a prototype methodology for deriving compliance matrices that connect legal obligations with regulatory risks, mitigating controls, and auditable control questions. The resulting risk and compliance matrix was evaluated by Velaw and assessed as sufficiently robust to support further development.
Perspectives
The collaboration between BFH AIRS and Velaw demonstrated the technical feasibility and practical relevance of applying AI methods to regulatory risk and compliance processes at TRL 3. It also showed the potential of AI-supported compliance systems to improve efficiency, consistency, and auditability in highly regulated domains. Future work will focus on scaling the prototype, strengthening the legal knowledge graph, improving the generation and validation of compliance control questions, and integrating the approach into operational RegTech workflows. The project also clarified important regulatory expectations around transparency, governance, traceability, and auditability, including through exchanges with regulatory experts. These insights will guide the next development phase toward a regulation-ready AI compliance solution for compliance teams, auditors, and supervisory authorities.
