BOCHK-Loan Document Review System

Designed an AI-powered loan document analysis tool for Bank of China (Hong Kong), enabling automatic comparison between the term sheet and the loan agreement to improve analyst efficiency.

Project Details

SECTOR

Web App

RESPONSIBILITIES

User Research, Information Architecture, Interaction Design, Visual Design

DURATION

5 Months

CLIENT

Bank of China (Hong Kong)

ROLE

Design Lead

YEAR

2024

BOCHK Loan Document Review System Dashboard

Background

The manual comparison of loan documents at BOCHK was creating significant operational challenges:

Time-consuming icon

01 Time-consuming Manual Process

Analysts spend an average of 8 hours comparing a single document (50-100 pages), handle 5-8 files weekly, and need to verify key terms multiple times.

Error icon

02 High Risk of Errors

Each document contains 3-5 important differences that are typically missed, legal terms are easily misinterpreted, and error rates increase by 40% in later sections due to fatigue.

Learning curve icon

03 Steep Learning Curve

New employees require 6-8 months of training before working independently, first-year work needs 100% review, and finding related clauses across documents requires significant experience.

Solution Highlights

Intelligent Document Upload and Auto-Recognition

Automatic Clause Matching: Side-By-Side Comparison View

Intuitive Dashboard - Performance at a Glance

Loan Document Comparison: Before vs After

Loan Document Comparison: Before vs After

Result and Extension Thoughts

After six months of development, the tool now achieves 95% accuracy in real-world use. With a manual review step in place, it fully meets customer requirements. There's still room to improve accuracy further with future AI and machine learning enhancements. As a tool built to boost efficiency, it has successfully delivered on its purpose.

As AI and natural language processing continue to evolve, this tool also has potential for broader application. Industries like law, insurance, real estate, and healthcare all rely heavily on accurate document comparison. It would be valuable to explore how similar solutions could help people handle complex documents more easily across different fields.