Overview
This project investigates the use of artificial intel-
ligence (AI), in particular language models (LMs), to support
compliance workflows in the financial and crypto-asset sectors.
The work focuses on two proof-of-concept systems: one for
transforming unstructured regulatory and internal compliance
documents into structured database entries, and one for en-
abling natural-language queries over the resulting compliance
knowledge base. The objective is to reduce manual effort in
integrating new regulations and internal guidelines, improve
the consistency of compliance content, and facilitate access to
relevant information for operational use. In addition to pre-
senting the system architectures, the paper outlines evaluation
metrics for robustness and discusses key limitations related to
data governance, non-determinism, and dependency on external
model providers. The results indicate that LM-based systems
can provide practical support for structured compliance content
management and compliance-related question answering when
deployed within a controlled and supervised environment.
