AI Document Review
Customer contracts (B2C) may vary with time and currently need human review to validate and confirm data, identify contract details and perform additional check and balances. This takes time, is prone to human error and doesn’t automate data collection for further review (or even other analysis needed). Additionally, depending on contract flow, it may require additional human resources to avoid pilling up contracts (bad customer service).
Challenge
Implementing an AI solution for automating the review and validation of customer contracts in a B2C context presents several challenges that need to be addressed to ensure efficiency, accuracy, and customer satisfaction. These challenges include:
- Variability and Complexity of Contracts
- Accuracy and Reliability
- Integration with Existing Systems
- Data Privacy and Security
- Scalability
- Training Data Availability and Bias
- User Feedback and Continuous Improvement
Approach
Using a custom ETL process:
- 1. Contracts files (pdf/doc) are upload into a common storage
- 2. Using DocAI (text recognition service) files are read and converted into text files
- 3. GenAI (ChatGPT), given a context, analyzes each contract and provide /outputs relevant information like
- 4. Contract details
- 5. Customer details
- 6. Missing information
- 7. Data is stored in a relational database
- 8. PowerBi Analysis
Value Delivered
The traditional approach – using human manual intervention – could add up delays and additional costs, depending on the volume of contracts to be reviewed. A typical HR would take between 10 to 20 minutes to review and collect data for a single contract. On average, one person would be able to review 32 contracts daily. Using AI, the full process was automated, and cost was reduced to 0,04 € per contract reviewed, being able to process 500 contract per hour for a small AI service.
