Credit Risk Assistant

Credit Risk Assistant by GFT, developed in collaboration with Oliver Wyman, is a solution that leverages Generative AI (GenAI) to automate the creation of Credit Memos and credit risk reports. The solution enables banks and financial institutions to accelerate credit analysis, improve report quality, and ensure traceability and regulatory compliance, all within a secure and scalable AWS environment.

Key Features & Highlights: End-to-end automation: Extracts, analyzes, and structures data from internal and external sources, documents, and the web. Structured document generation: Produces Credit Memos ready for PDF export and available in multiple languages. Interactive interface: Allows users to refine results, review sources, and regenerate report sections. Traceability and transparency: Each section includes source references, confidence scores, and explanations of gaps. Observability: Integration with platforms like Langfuse for cost, performance, and LLM usage monitoring. Modular, scalable architecture: Deployable on AWS, multi-cloud, or on-premises, using Docker containers and compatible with various LLMs.

Reference Architecture on AWS:

Check attached diagram in this document. Data Ingestion: Amazon S3, AWS Managed Workflows for Apache Airflow. GenAI: AWS Bedrock for LLM generation and embeddings. Database: Amazon RDS/PostgreSQL. Observability: Integration with Langfuse and FastAPI. UI: Web interface for document and report management.

Use Cases:

Credit Memo automation: Reduce turnaround from days to minutes, saving over 50,000 analyst hours annually and up to €4M in costs. Standardization and compliance: Ensures consistency, traceability, and regulatory compliance in all reports. Multi-language support: Operate in international markets and comply with local regulations. Human-in-the-loop: Analysts can review, adjust, and validate results for maximum quality.