A large cabinet level federal civilian agency that provides near-comprehensive healthcare services and non-healthcare benefits.
Today, millions of claims are filed with the agency. These claims come in a variety of formats including mail, fax, and electronic submission. Almost all of the claims received contain a degree of handwritten text. Faced with a growing backlog of hundreds of thousands of claims and about 3 months processing time, the agency embarked on a 5-year contract.
The contract was for a variety of upgrades, including the automatic processing of incoming mail claims. It’s the first step in a larger digitization effort for the agency, with the vision of providing greater visibility and transparency about where claims are in the process.
The agency wanted to speed up processing times without sacrificing accuracy. It knew the variability and diversity of documents received required an intelligent, flexible solution. To complicate matters further, these are large file submissions – often downloaded from the agency’s website, filled out by hand and mailed in – nearly all contain handwritten text and other difficult conditions that Optical Character Recognition [OCR] and other legacy products cannot reliably handle.
With a tight timeframe for implementation and the help of IBM and their federal systems integration partner, Aptive, the agency was up & running with Hyperscience roughly two months after the contract was signed.
Using Hyperscience today, the agency can quickly and efficiently classify documents while reliably extracting handwritten information. In the first three months, more than 1 million pages, or 115,000 claims, were processed automatically.
The agency measures success based on the rate of forms that can be accurately processed end-to-end. Whereas with their old solution they were achieving roughly 77% accuracy rates, today with Hyperscience, they’re targeting – and exceeding – over 96% accuracy and 99% automation.