Gain Workplace Efficiences with AWS by Utilising Machine Learning to Automate Manual Processing

Overview

Industry Trading is the leading End of Lease and End of Life Remarketing service provider in Australia. With over 60 years of combined experience working in Asset Finance businesses, the core Management Team boasts extensive knowledge of End of Lease asset processing requirements, the importance of timely, accurate reporting and the severe impact that reporting delays can have.

With business efficiency and timeliness being core components of their business, Industry Trading were looking for ways to speed up internal manual processes without compromising on quality of work. Industry Data were employed to review a time sensitive but manual internal process for ingesting incoming invoices from around the globe. An AWS text recognition and machine learning serverless solution was selected to automate this process, meaning internal resource could be re-directed to more revenue driven tasks, and the time taken to complete the process was signifcantly reduced.

Take a look at the Industry Trading website to learn more about what they do: industrytrading.com

Business Challenges

  • Time consuming and manual internal process to review, classify and store incoming invoices
  • Invoice formats vary greatly by business, industry, and country
  • Incoming document volume varied greatly by month therefore hard to predict required resource
  • Documents contained sensitive information and therefore had to be secured in transit and at rest

“Industry Data are our go-to partner for solving business issues using cloud-based technology. We find that they are attentive, intelligent and it's always a great experience working with them."

--Hayden Shuttleworth, Industry Trading


Project Outcomes

A serverless solution to ingest, store and process invoices was created within AWS achieveing the following benefits:

  • Scalable system means incoming volume is no longer an issue, system runs as and when a new invoice is received and auto scales with demand
  • Process has been almost completely removed from internal resource with only ad-hoc checks completed to ensure accuracy is maintained
  • Documents are encrypted both in transit and at rest to ensure they are secure, and saved in S3 for further processing
  • Job processing time reduced from 3.5 hours per week to minutes, with internal resource spending no longer than 30 minutes per month on quality checks
  • Running costs are kept to an absolute minimum as system only runs as and when required and completely shuts down afterwards

The Industry Data Solution

Industry Data utilised the Textract AWS service to automatically scan and detect text within each invoice received. Set-up focused on the various formats available and the key words that defined each data type, allowing Textract to then identify similar data types when new formats were presented. A Lambda function was built to trigger the service to scan and extract the data from the document upon receipt in an S3 bucket. Post completion a combination of SNS and SQS were used to notify the user and to trigger a second lambda function that formatted the text into a usable file for database load from its location in S3.

The AWS architecture developed for this solution is documented below. For more information on Textract by AWS please visit: Amazon Textract

Architecture diagram for an AWS serverless record matching solution