AI Textract solution: How invoice automation led to 15X workforce streamlining and over $10M in annual cost savings
Client Story
Our clients in Logistics, Legal, Retail, and Construction looked for ways to automate and streamline invoice processing with the goal of reducing manual labor and increasing operational efficiency. We offered them a customized tool that automates document data extraction, processing, and analysis – and drives substantial cost efficiencies.
Our solution's widespread adoption and scalability across diverse business operations were crucial. Looking ahead, we successfully implemented this AI-powered solution across more than 10 companies. They generate hundreds of thousands of invoices monthly, and since then, they have also generated higher revenues with our tool.
President
Supply Chain SaaS company
The Challenge
The manual invoicing process presents significant challenges for companies across various industries. Automating this process is crucial for accountants to improve efficiency, reduce errors, and save valuable time for more strategic tasks.
The main challenges are:
- High Volume of Paper Invoices
Companies receive thousands of paper and scanned invoices monthly from various suppliers. - Manual Data Entry
Employees and external accountants manually digitize these invoices, which is labor-intensive and error-prone. - Lack of Standardization
Invoices vary in format, including handwritten annotations, making the digitization process chaotic and complex. - High Error Rate
Human intervention is frequently needed to correct errors, further slowing down the process and reducing accuracy.
Additionally, businesses may face technical challenges, such as integrating the new solution with existing software. Overcoming these issues required collaboration between finance departments and us to ensure a seamless adoption process.
The Solution
Our team developed an AI solution utilizing AWS Textract, a tool that automates the extraction, processing, and analysis of document data.
Technologies used:
- AWS Lambda
- Amazon Textract
- AWS S3
- AWS Simple Storage Service
- AWS API Gateway
Team:
- Senior ESB Engineer
- 2 ESB Engineers
- Senior Java Engineer
- Data Scientist
- Software Architect
The AI solution implementation involves the following steps:
- PDF Acquisition
We received scanned PDF invoices from an email inbox. The Serverless AWS Lambda function periodically checked the inbox for new emails containing invoices. - Preprocessing
The PDFs were preprocessed to convert them into JPEG images. Quality enhancement techniques were applied to improve the readability of the images. - Template-Based Processing
Specific templates were created for each supplier's invoices, as the AI needed guidance to interpret the data correctly. These templates included details such as expected tables and column layouts. - AWS Textract Integration
The preprocessed images were fed into AWS Textract. Textract processed the images and extracted text data based on the predefined templates. The extracted data was evaluated against a quality threshold. If the parsing accuracy was satisfactory, the data was saved as a CSV file and sent to an invoice verification company via FTP. - Human Intervention Loop
If the extracted data did not meet quality standards, a human intervention loop was initiated. A notification was sent to a specific team, who then manually verified and processed the data, saving it as a CSV file. - Table Parsing
Parsing tables was particularly challenging due to overlapping columns and complex layouts. Templates were crucial in guiding the AI to correctly interpret tables and avoid common errors.
Tech stack
- AWS Lambda
- Amazon Textract
- AWS S3
- AWS Storage
- AWS API Gateway
The Results
We completed the project in under 6 months and performed integration into 10+ of our ongoing projects across various sectors.
Here are some of the business outcomes:
- Cost Savings
The solution led to annual savings exceeding $10 million, significantly impacting the companies' bottom line. - Automation Efficiency
By leveraging the solution, the amount of manual labor required for the process was reduced by 15 times. This streamlined workflow resulted in significant time and resource savings for the organization. - Digitization Rate
Almost 95% of invoices are now digitized seamlessly without the need for human intervention. This high level of automation has greatly improved efficiency and accuracy in processing invoices, leading to faster turnaround times and reduced error rates.
President
Supply Chain SaaS company
Let’s discuss your ideas
We have been delivering tailor-made software solutions since 2012 and have successfully completed cases in the supply chain sphere. No matter the complexity, we are prepared to meet your requirements and challenges using the best-suited technologies.
We are ready to involve external subject-matter experts and share our domain expertise. At the same time, we know how to become a seamless part of your team if needed. Get in Touch
