Leading AI cloud provider H2O.ai announced the general availability of H2O Document AI, a machine learning service that understands, processes, and manages the large volume and types of documents and unstructured text data that businesses and organisations handle every day.
H2O Document AI streamlines operations, reduces costs, and discovers new information and insights contained in documents. H2O Document AI “learns as it goes,” continuously improving processing accuracy using H2O.ai’s latest innovations in machine learning and deep learning to achieve automation across business verticals and use cases not previously possible.
In today’s digital world, core business processes rely on documents and unstructured or semi-structured data that contain valuable information critical to business operations. The definition of a ‘document’ continues to expand and includes PDFs, emails, scans, images, paper and web forms, faxes and e-faxes, chats from chatbots, free-form text, and more. 80 per cent of enterprise data is unstructured or in a format that is not machine-readable or readily available.
However, traditional AI document processing solutions use Optical Character Recognition (OCR) or Robotic Process Automation (RPA), limited by rules-based and template-driven constraints. Additionally, the OCR and RPA solutions have limited capabilities to self-learn. Not surprisingly, the results from these existing document processing solutions are often lacklustre. Without other options, some organisations do with sub-optimal products/solutions that don’t scale and increase inefficiencies.
Other organisations have not adopted any automation and rely on people to review, process, and act on documents. The manual labour required to process documents is tedious, can be error-prone, and keeps workers from doing more meaningful, impactful, and enjoyable work. In either scenario, using existing automation or no automation, handling documents is expensive and time-consuming — companies spend an average of $20 to file and store a single document, employees spend up to 50 per cent of their time searching for information and can take, on average, 18 minutes to locate a single document.
H2O Document AI helps organisations quickly and accurately process documents and unstructured text data to increase productivity and find hidden insights. H2O Document AI provides automation not previously possible by combining state-of-the-art Intelligent Character Recognition (ICR), Natural Language Processing (NLP), computer vision and layout intelligence.
H2O Document AI comes pre-built with recipes created by H2O.ai’s two dozen Grandmasters (best in the world data scientists); a sophisticated labelling and training workflow; self-service capabilities to build, deploy, and manage high-accuracy AI models to classify documents and pages, in addition to extracting value, and meaning. The service has flexible out-of-the-box document pre-and post-processing and seamlessly integrates with customers’ existing business processes and workflows. H2O Document AI identifies entities and classifies them, and understands the document and constituent pages, sections, and layout to provide additional context that can help drive decision-making and lead to an improved end-customer experience.
Document data can easily be used to build intelligent searches on extensive archives of documents, or it can be loaded into a datastore to be queried, analysed, audited, or used by other applications. H2O Document AI provides a workflow from labelling to training to scoring to low touch integrations and consumption options. It is customisable and extensible and works in conjunction with H2O AI Cloud, so customers can quickly scale to additional use cases. And, the types of documents that benefit from AI-enabled ICR and NLP are numerous, including mortgage paperwork, receipts, applications of all kinds (life insurance, jobs, loans), tax documents, employment records, and more.
“Our banking, insurance, health, audit, and public sector customers each process billions of documents every year. Documents are the fastest-growing source of data in the enterprise, ranging from contracts, bank statements, invoices, payroll reports, regulatory reports, and medical referrals to customer conversations in text, chat, and email,” said Sri Ambati, CEO and founder, H2O.ai. “H2O Document AI enables customers to sieve intelligence across a wide variety of document types not possible before, with unprecedented accuracy and speed. With H2O Document AI, businesses can now seamlessly integrate insights from documents to their feature stores and transactional systems to delight their customers.”
Health systems, as an example, receive millions of faxed documents annually, including patient referrals, prescription refill requests, durable medical equipment requests, lab results, and school forms. Without AI technology, people must review the documents multiple times to determine the document type, which patient it pertains to, and what precisely must be done to further process or respond to the document’s contents.
The Center for Digital Health Innovation (CDHI) at the University of California, San Francisco (UCSF), is collaborating with H2O.ai to develop and train AI algorithms to recognise these various document types, analyse the contents, extract relevant data, and appropriately route information and requests to systems or individuals as necessary for follow-through. Once trained, these algorithms will enable CDHI’s referral automation software to significantly speed the processing of 1.4 million faxes UCSF Health receives each year.