How Automation Hero uses accurate AI to process documents

Before Dr. Alan Turing designed the first computer, people merely dreamed of intelligent machines that could read paperwork and do most of their grunge work for them. Science-fiction movies depict advanced software processing large amounts of documents to find hidden insights that save the day. Today this is available in real life from progressive-thinking software providers. 

One of them, San Francisco-based Automation Hero, today launched v6.0 of its Hero Platform, a SaaS service the company claims takes a quantum leap in OCR (optical character recognition) document-processing accuracy. It also can read handwriting with 70-80% accuracy. 

Yes, this includes doctors’ notoriously bad handwriting, Automation Hero founder and CEO Stefan Groschupf told VentureBeat. 

The Hero Platform has shown, both in beta trials and in actual production, that it can unlock millions of dollars in return on investment (ROI) for companies such as MarkerStudy, turning documents into actionable insights in claims processing, supply chain optimization, fraud detection and customer-interaction automation.

Faster new AI engine

The latest version of the AH platform provides a faster new AI engine for document processing and automation, Groschupf told VentureBeat. The company’s deep learning-based approach together with its fast-reading OCR process can turn structured or unstructured documents – such as contracts, invoices, receipts, prescriptions, doctor notes and purchase orders – immediately into highly accurate and actionable data, Groschupf said. 

Modeled on cognitive science, Automation Hero’s OCR engine turns scans of documents into data similar to how humans read even the most difficult handwriting with contextual cues. The patent-pending technology is combined with an intuitive-to-use natural language understanding engine, all within a single end-to-end platform, Groschupf said. 

While this seems far-fetched, the goal is to deliver the highest ROIs in the industry within weeks without any data science or large training data required, Groschupf said. 

“Our job at Automation Hero is to accelerate health care, get insurance claims paid out faster, and find bad actors that over-invoice or bring to the surface contracts that can be renegotiated,” Groschupf told VentureBeat. “Our mission is to provide overwhelmed employees with the AI superpowers to work faster through the mountain of documents that keeps stacking up on their desk or in their inbox.

“We have AI that goes to a document management tool and processes the data to make it actionable to automate certain business processes.”

Faster by 40% than Google’s OCR, Automation Hero claims

Based on new benchmarking data, the technology outperformed handwriting OCR alternatives ABBYY and Google Vision by approximately 280% and 40% respectively, Groschupf said. It also outperformed ABBYY and Rossum in AI-based invoice processing in a separate benchmark by 60% and 100% respectively, he said.

Insurance claims processing is Automation Hero’s best business right now, Groschupf told VentureBeat. 

“Eighty percent of all data in companies is in documents. This is a treasure chest that historically was not tapped,” Groschupf said. “And if you think about all of this business intelligence and ETL, it’s a multi-billion dollar market in which we are all operating on 20% structured data. Historically, humans have stored so much unstructured data – emails, contracts, images, invoices, infrastructure data, etc. – that it took days or weeks to get through previously.

“We can do things that you couldn’t do before, such as constantly mining hundreds of thousands of contracts, automating 60,000 claims a day, and so on. We really want to be a bicycle for people’s minds, mining way more documents than we could do before.”

Automation Hero’s unique approach to OCR has decreased tedious claims administration work by hundreds of hours, transforming the way we handle customer claims and manage employee satisfaction, Eoin Grace, deputy head of IT at Markerstudy, said in a media advisory. 

“Previously, we were manually deciphering doctors’ diagnoses, stamped addresses, and handwritten notations, across a wide variety of claims forms,” Grace said. “Automation Hero’s intelligent OCR technology delivers above-human accuracy.”

Originally appeared on: TheSpuzz