How We Automated Financial Documents Processing at Scale.

By Blessing Adesiji

October 12, 2022   ·   4 min read

This article will describe how we delivered a document processing solution that helped a distribution company save 97% while ensuring ease of use and transparency. 

Introduction

Using the Mantium platform and AWS Textract, a distribution company for one of the world’s foremost consumer brands saved 97% on the cost of processing a document with no in-house IT or software engineering department. 

Using Mantium, the distributor was able to automatically handle 16,000 invoices per month, freeing up significant time for its employees to complete higher-value work. This article describes how we delivered the document processing solution while ensuring ease of use and transparency.

The business case

Traditionally, accounts payable (AP) departments have individuals responsible for manually processing invoices, sending out payments to vendors, and managing the process for accuracy. Intelligent document processing presents a new world where manual processes can be automated, allowing the AP team to do more with less effort while saving time and money. 

Our customer was interested in automating their AP process, as they were processing invoices with 10 – 16 individuals, who each put in 10 hours of work every week to complete. Mantium helped by building a customized AI workflow to streamline the document process, increase accuracy and security, improve the AP’s team morale, and save costs.

Our Unique Approach & Results

To support the customer’s business case, Mantium provides a pre-configured AI workflow enabling the distributor’s AP team to upload digital copies (PDFs) of invoices and get a processed file with all the desired information. Mantium’s solution leverages AWS Textract to extract and parse the data from the invoice.

AI Workflow with Textract

The workflow involves copying all the invoices from the client’s SharePoint storage to our S3 bucket, from which we can process the files through Textract. With the Textract endpoint, we can extract and parse expense information from invoices on a page-by-page basis. The results are stored in the output file (a CSV file), after which this output file is moved to the client’s preferred storage, in this case, we utilized our custom integration with Microsoft’s O365 Sharepoint Online service.

Simplified Output

There is a lot of information in an invoice; some are not useful to the AP team and lead to long hours of the team manually reviewing thousands of invoices for the most important information. This presents the need for accuracy and prioritization of information in the final results. Mantium’s AI Workflow with Textract integration is able to detect accurate document information, and we filtered through these results to return the information desired in the final output. 

Textract provides a confidence score, a decimal number between 0-100, where a higher value indicates that the system is more confident that the result is accurate. Although it wasn’t implemented in this project, the AP team could manually review low-confidence output with Mantium’s enhanced human-in-the-loop (HITL) capabilities.

In summary, the output CSV file contains all the important fields from invoices after they were parsed, with each row representing information from one document source (PDF). After successfully processing documents, they are all automatically stored into the specified location for easy and consistent access by the AP Team in the systems they already use.

Conclusion

In this project, Mantium provided the best technical solutions using state-of-the-art models to a customer who did not have robust technical departments. Our customer saved 97% in the processing cost per document without an in-house technical department, and their AP team can focus on other accounting tasks without hiring more personnel to process more invoices. 
Visit our AWS Marketplace listing to learn more about the AWS Textract and Mantium integration.

Sign up to join the waitlist for the beta of AI Builder.

Follow us on Twitter @MantiumAI | Follow us on LinkedIn Mantium

ABOUT THE AUTHOR

Blessing Adesiji
With a Bachelor's of Science in Petroleum Engineering, Blessing is a self taught Data Scientist and Software Engineer. He enjoys educating Mantium users on how to build AI applications as a Developer Relations Engineer. When he is not writing code and tutorials, he enjoys doing exercise in the gym, and playing and watching football.

Enjoy what you're reading?

Subscribe to our blog to keep up on the latest news, releases, thought leadership, and more.