SYMBOL

In the SYMBOL project, researchers at the University of Applied Sciences of the Grisons (FHGR) are exploring the use of neuro-symbolic artificial intelligence for assistants deployed in high-stake environments such as banking and compliance. The goal is to develop systems that meet the highest standards of confidentiality, reliability, and traceability.

Asset managers rely upon combined customer relationship management (CRM) and portfolio management systems (PMS) to meet Swiss regulatory requirements. As these systems expand, the demand to use this data for business intelligence has grown significantly. While existing solutions are too complex, product-specific, and underused due to their steep learning curve, standard LLM-based approaches do not provide the level of reliability required for use in high-stake finance environments.

SYMBOL addresses these issues by building a natural language interface powered by artificial intelligence (AI), which answers queries, creates compliance reports, and enables the extraction of valuable business intelligence. The assistant allows non-technical users to ask natural language questions and receive actionable insights, removing the barrier of technical complexity and significantly enhancing user productivity.

To achieve this goal, the project will overcome the following key challenges:

1. Develop domain-specific tiny LLMs that improve reliability and significantly lowering hardware requirements. This allows confidentiality through deployment on local infrastructure (on-premise) or as an SaaS subscription.
2. Use neurosymbolic AI to ensure the reliability and explainability of LLM results, and enable combining both textual compliance data stored in the CRM with financial data in the PMS.

Parties involved

The project is being implemented by the Swiss Institute for Information Science (SII), in cooperation with Alphasys AG. It is financed by Innosuisse.