abaQon Project Report: Efficiency Gains Through GenAI in the Onboarding Process

THE PROJECT

A globally active private bank faced the challenge of making its onboarding process more efficient and future-proof. In particular, manually checking and analysing KYC documents – such as CVs, passport copies, company documents and company websites – was proving to be time-consuming and prone to errors.

The aim of the project was to develop an AI-supported solution that would support relationship managers with KYC tasks, significantly reduce the time required and improve data quality at the same time.

OUR CONTRIBUTION

In close cooperation with the customer, we designed and implemented a workflow-based AI agent that enables the automated processing and analysis of KYC documents.

In the first step, a cloud-based OCR solution (Microsoft Azure) converts image-based documents into text. Large Language Models (LLMs) then extract, structure and validate the relevant information. Based on this, the system generates summaries and suggestions that are made available to the relationship manager for further processing.

Thanks to an iterative development process in close coordination with the KYC departments, the specific requirements could be precisely mapped. Particular attention was paid to seamless integration into existing systems and compliance with the highest data protection and security standards. The solution was successfully rolled out and is now an integral part of the KYC process.

REALISED ADDED VALUE

The implemented GenAI solution reduces the time required for KYC tasks in the onboarding process by up to 80% and at the same time noticeably increases the first-time-right rate. Queries from mid- and back-office functions have been significantly reduced.

Relationship managers are now focussing on the qualitative evaluation of AI suggestions and on customer care – instead of manual document analysis.

Thanks to its scalability, the solution is suitable for several thousand onboardings per year and also forms the technological basis for further AI-based use cases.