M Booth Automates Expense Reporting with Generative AI

M Booth

M Booth, a marketing and communications agency, automated its error-prone expense reporting process with an AI-powered solution that reduced manual effort, improved accuracy, and enhanced employee experience.

100%

Automated receipt interpretation and categorization

0 manual errors

Minimized mistakes in expense reporting

1 seamless system

Integrated AMEX feed with existing ERP for unified tracking

Problem

Paramount Capital processes 9,000–10,000 loan documents monthly across five clients, with plans to scale to 150–200 clients by 2026.
Existing Amazon Textract solution requires custom training for each document type, limiting flexibility and scalability.
Need for an adaptable, cost-efficient, and scalable document verification process.

Solution

zeb leveraged AWS Generative AI services to develop:

A Generative AI-based document verification system using Amazon Bedrock with Anthropic Claude.
Prompt engineering strategies and confidence scoring to ensure reliable, accurate data extraction and comparison.

Impact

Enabled flexible processing of various document types without custom training, preparing Paramount Capital for large-scale client expansion.
Reduced manual effort and operational overhead associated with document verification.
Improved scalability and potential cost efficiency as document volume increases, ensuring sustainability as the business grows.