AI-powered processing of business documents
Did you know that the average cost of manually processing a paper invoice lies between $15 and $40? According to The CFO’s Guide To Digitizing B2B Payments, the average overall annual cost incurred by businesses from processing paper invoices is $171,000. Think about that for a second. What drives that cost and how can that cost be reduced?
Besides storage and shipping/mailing, the majority of the cost stems from human labor. This includes transferring document content to digital data, for instance the entry of an invoice into accounting software. Added to that is the cost of lost discounts and penalties incurred due to errors or late payments, simply because the human element is the bottleneck.
Several industries have made great progress in standardizing the exchange of information. The healthcare world created the Health Language (HL7) for the exchange of patient related information, which is now transitioning to the more modern FHIR standard. The commerce world settled on the Electronic Data Interchange (EDI) standard. But adoption has been an issue. EDI implementations and integrations to back-office systems are expensive, and with a variety of standards like ANSI, EDIFACT and TRADAFACT, maintenance of document format changes is complex.
Paper documents are here to stay
The simple fact is that paper documents are not going away any time soon. But it begs for a more efficient and accurate form of automation.
Many larger businesses are using 3rd party services to digitize paper documents via OCR. While costly, it has the benefit that storage, searching and retrieval of documents is far more convenient and much faster. However, OCR has fairly strict requirements for legibility. Faded ink and variations in character sets can lead to incomplete or inaccurate digitization. And even when OCR correctly recognizes and digitizes characters – it still does nothing to interpret the content of a document. It lacks the ability to automate processes based on understanding the data captured from paper.
AI for document recognition
While we all know Google for its search engine and web analytics tools, Google Cloud has been making great strides in providing the building blocks for cloud computing, data analytics & machine learning technology to power data-driven business applications. One example of these building blocks is the Google Cloud-based Document AI solution, or DocAI for short. In Google’s own words: “Document AI is a document understanding solution that takes unstructured data (documents, forms, etc.) and makes the data easier to understand, analyze, and consume by providing structure through content classification, entity extraction, advanced searching, and more.”
Given the vast amount of data Google processes every day, DocAI has become very proficient in the accurate recognition of data in paper documents, across languages and formats. But DocAI is an engine, a platform, not an off-the-shelf solution. Google Cloud has left the development of practical business applications to its partners, and provides a complete set of APIs that allow software developers to embed or access the engine in other solutions.
UnDoc - Intelligent processing of business documents
With its roots in enterprise resource planning and business process automation, Vanenburg – as an experienced Google Cloud partner – used its expertise to create a unique solution around DocAI to streamline and reduce the cost of processing paper documents. Vanenburg’s solution, called UnDoc, utilizes the full power of DocAI for accurate, automatic recognition and processing of paper documents, using a set of pre-trained parsers and templates. UnDoc adds a web-based user interface for monitoring the flow of documents, including the ability to ‘teach’ the DocAI engine whenever it does not recognize a data element on a page. This enables DocAI to accurately recognize that data element on a similar document going forward. In addition, UnDoc enables the automated insertion of processed data into backend systems via APIs.
Paper documents are here to stay, for the near foreseeable future. But it is no longer necessary to spend inordinate amounts of resources and money on the accurate processing of these documents. Instead, more time can be spent on value added tasks.