BACKGROUND
While considerable investment had been made to streamline the processing and conversion of paper documents to digital files, there is still much room for improvement. Our client, a leading National Title insurer estimated that their current data extraction and scanning effectiveness was no better than 60 percent. All remaining documents had to be manually configured or modified. What follows is a data extraction case study explaining the challenges faced and what results were achieved.
CHALLENGE
Our client conducted a search for a document scanning solution that could perform with greater effectiveness. Traditional document extraction solutions typically relied upon templates or keyword searching for information capture. The challenge is that in most cases, the complexity of document formats in the Title industry limited scanning accuracy. Manual data capture must then be performed, which results in three main issues:
- Slow processing time
- Manual data entry errors, and
- High cost to utilize on-shore resources or subject matter experts
In any given month, this Title company was tasked with pulling from one to 50 data points from public records that are extracted from up to 3,000 county websites. This Title company data extraction case study utilized source documents that could be classified as both semi-structured and unstructured, as described below.
SEMI-STRUCTURED DOCUMENTS
While much of the formatting might be the same, the key data elements varied. For example, a table, list, or fields could randomly come and go on a page. In this case, a template could not ease data extraction challenges. This was the case with loan applications, credit reports, and tax documents.
UNSTRUCTURED DOCUMENTS
Most of the index information in these materials is buried within the depth of a paper, often as words or sentences in a paragraph. The data positioning continually changes, so the only consistency is the language around the index. Deed, Release, Correspondence, Contracts, and notes are examples of unstructured content with this unique challenge.
SOLUTION
This client selected Axis Technical Group for their document classification and data extraction solution that leverages artificial intelligence to steadily improve performance. By choosing Axis Technical Group, it was possible to take advantage of artificial intelligence that could learn how to operate with greater efficiency. As an advanced technology software solution, Axis Technical Group could be used to classify and extract data from both structured and unstructured content.
Using proprietary algorithms, including those used to perform Natural Language Processing (NLP), this solution was able to read and extract data from sentences, paragraphs, and entire pages written in natural English.
Axis Technical Group solutions have been proven to accurately deliver 80% or more of our client’s documents within the first 30 days. Since this Axis solution is a true machine learning system, it gets smarter as it processes more document types. This intelligence has led to an increase in the company’s scanning success rate to be close to 90%. These results were achieved without requiring any manual intervention.
BENEFITS
Results achieved by Axis Technical Group’s client deploying this new solution include:
- Reduced title clearance time from 2-4 hours to just 20 minutes
- Achieved a 10x reduction in manual data entries, a drop from 4-5% (industry average) down to 0.5% to 1%
- Better utilization of Subject Matter Experts to reengineer business processes and workflows instead of performing manual data entry
- Unlocked new business growth and scalability potential
- Delivered higher customer satisfaction to clients through faster turnaround times
NEXT STEPS
Our client insisted on anonymity in this case study to avoid exposing its new competitive advantage. The company is now actively planning on applying an Axis AI solution to other functions and business units heavily reliant upon data extraction from both structured and unstructured sources.