Making smart data decisions for your RPA tool
In retrospect, company file sharing paved the way for document management applications. Then, these document management tools led to cloud-based information management systems. From there, remotely accessible content and cloud computing led to Big Data analytics. And, the demand for process automation has led Robotic Process Automation (RPA) to seek document capture tools that can interpret content to power their digital workflows. As technological innovations guide the way for data-based insights and automated task processing, companies are realizing the importance of existing textual data within their documents and file stores.
With the aim of providing document intelligence to robotic workflows, most RPA systems offer Optical Character Recognition (OCR) tools out of the box. A recently published Gartner report on the RPA market makes the assertion that OCR “is tangential to the core of RPA.” However, if the goal is to replace a manual process with a robot, traditional OCR barely scratches the surface of document intelligence. Smart content capture – the AI-powered evolution of OCR technology – is required for true process automation.
When working with companies on digital transformation or document process automation projects, I regularly answer the question, “Do I really need a separate capture product? Why not just use the OCR activity/tool/VBO available in my RPA application?” An example comes to mind when I asked myself a similar question mid-way through a recent home project. Standing in front of a paneled wall with a picture frame on the floor at my feet, a level in one hand and a nail in the other, I realized I left the hammer in the garage. “Why not use the handle of this screwdriver to hammer the nail into the paneling?” If the wood is soft, the nail is sharp and the wielder of the screwdriver has precise aim, it is possible. But change any of these circumstances, and you’ll quickly realize this makeshift tool replacement is out of its league. The same is true when a developer or business analyst tries to force-fit OCR into document-centric RPA workflow when a smart, machine learning-powered content capture tool is needed.
Simply put, there is a world of difference between an OCR engine and an intelligent capture tool. An OCR or ICR (Intelligent Character Recognition often used for detecting handwriting) engine recognizes characters on an image and outputs text in a machine-readable format. Some OCR products include tools for textual pattern or fixed-form extraction, but these methods of identifying key text values from a document are rarely scalable or effective across a variety of document types.
The same Gartner report summarizes, “despite the claims of some of the RPA vendors, there are only limited opportunities to use machine learning in the core of RPA itself.” When it comes to document-based processes, the standard image-to-text tools like Google, Microsoft, Amazon and ABBYY OCR, are insufficient to meet the demands of a workflow that includes variable documents and textual repetition.
There are key document workflow attributes that should trigger an analyst or RPA developer to employ a smart capture tool rather than a traditional OCR engine:
- If there is variability in the text or layout of the documents that are being processed will not be enough. For example, if you consider invoices or purchase orders, OCR alone or OCR use in conjunction with some form of template-based extraction will not be usable if the template varies.
- If files need to be separated into individual documents prior to inclusion in an automated process, such as stacks of documents that were scanned in batches, OCR alone will fall short of expectations.
- If there is repetition in the key values or index fields (multiple dates or addresses in a contract or mortgage loan) needed for document workflow, again, OCR alone will not accurately extract the necessary data.
To learn more about how Ephesoft Transact integrates with best in class RPA systems to provide document intelligence to robotic action, click here. For a demonstration or to connect with an Ephesoft sales representative, email sales@ephesoft.com.
