As the communications channels used in capital markets trading proliferated and became more digital, Societe Generale identified a need to modernize its approach to trade surveillance. After finding no commercially available software that could meet their needs, the bank decided to build the capability in house, using best of breed technologies. These included natural language processing trained on models to understand the highly specialized language of traders and applied to audio communications as well as email, texting, and documents. The system automatically alerts compliance when it has identified potential abuse.

Societe Generale’s solution does this in 26 languages, across some 30 communications channels, and for 16 countries. The project is impressive in its use of new, often open source technologies, and in terms of its scale.