Accelerating Document Processing for Government Agencies Demo

Mike Ryan: Hi, my name is Mike Ryan, Senior Solutions Engineer at Hyperscience. This demo will showcase how Hyperscience can help agencies automate tedious manual work associated with document processing.

For over five years, both the VA and the SSA have been leveraging Hyperscience’s Hypercell to process over 1 billion pages per year across hundreds of document types. They’re currently using digital workers to streamline mail rooms, data entry, and repetitive decision-making tasks to ensure our veterans and citizens are supported in a timely manner.

Let’s take a look at Hyperscience in action. First, we need documents. Hyperscience uses API-based connectors to automatically ingest documents from your systems, folders, or inboxes. However, documents can be manually uploaded as well, regardless of file type.

Hyperscience automatically identifies the document type and begins the data extraction process using machine learning. The digital worker identifies where key fields are on the page and only asks for help from a human when it’s uncertain. An intuitive user interface allows reviewers to quickly validate what the digital worker has identified, and then tag any low-confidence fields.

Next, the digital worker reads all of the fields across the documents, but for the tough-to-read fields, defers to a human. That’s how our customers get the best of both digital workers and humans to achieve 99% accuracy and 90-plus percent automation.

As you can see, we’ve processed almost a hundred fields in just a few minutes. If we click into the W-9, you can review what human peers have transcribed as well as the digital worker. Our proprietary machine learning models have been trained to read cursive, hand print, checkboxes, and signatures, even when a “one” and an “L” look the same. We use context to understand human intent and even look outside the box to transcribe the field correctly.

Now that we are done processing these documents, all of this data is available in machine-readable format and can be sent downstream to automatically populate your systems of record or data analytics engines. Additionally, Hyperscience’s automated redaction can identify PII and create redacted copies of your documents for compliance purposes.

For unique business processes, you can design your own workflow to streamline your work. Here’s a claims process that involves case routing and approvals, data validations across documents, and integrations to Large Language Models. It even allows for customized user interfaces for reviewers to interact with the documents and make informed decisions.

Here’s a medical claim and supporting medical evidence that has been submitted into Hyperscience. The digital worker packages these documents into a case that can be tracked and managed end-to-end. Hyperscience extracts relevant case details from the documents, and then provides recommendations.

The first date of symptoms clearly occurred before the policy start date, so Hyperscience recommends claim rejection, but allows the human reviewer to make the final decision. Hyperscience has also pulled the offending clause from the database and used an LLM to summarize the doctor’s notes for easy reference later on.

To learn more about how we can help you achieve hyperautomation, send us an email at [email protected]. Thank you.