The scope of this automation entails taking an incoming stream of different types of documents, from various input streams including mobile, and quickly verifying the document type and performing data location and extraction—this is much more complex than sorting and searching data from the digital archive and places this technology into a “front office” process that is focused as much on improving the user experience through quicker turnarounds as it is on cost reduction.
The second broad trend is an aggregation of multiple automation capabilities within a knowledge worker-oriented workflow. Not long ago, if an organization were to automate document processing, it would typically be through multiple standalone solutions that were placed at the very beginning of a capture-oriented workflow. Keeping with the progression of automation, today, the different capabilities, whether it is automated document classification, document separation, or the data location extraction function, they are embedded into the by business process itself rather than exist as a stand-alone capture solution.

Can you share a case study of how Parascript helped a company overcome their document processing challenges?

One of our clients, a large bank, is quite active in the secondary loan market. They were facing challenges in managing the home mortgages. Most mortgages are originated by companies that aren’t necessarily banks and a major percentage of those loans are sold in the secondary loan market. In some cases, a loan is originated by a particular company and a few months later, it gets sold to another company that services the mortgage. This system required the bank to follow the protocol and redo all the loan documentation review and verification processes, even though the loan is already created and sanctioned. Since the bank relied on manual operations for the verification process, it became a tedious and never-ending cycle of events for the bank, urging them to seek an automation solution that could seamlessly address the problem. To solve this, Parascript created an automated document identification process capable of taking all the documents and automating the identification of documents. While doing so, it would also separate the files and send it to the appropriate staff, who can easily sift through the necessary documents and cross-verify.

How would you describe the unique value proposition of Parascript?

The most striking factor is our in-house machine learning capabilities that provide a much deeper-level of remediation for a client’s problem. The second unique factor is our take on the precision of the automation solution. While most other vendors concentrate on only functional configuration— aiding the software to identify different documents and rules of a company— Parascript spends ample time on tuning and optimizing those systems to enhance the throughput of automation. It is a time-consuming proposition, where we take samples and compare them with the system output document-wise and department-wise, but the result is an advanced breed of document automation that, in certain applications can achieve 99.5 percent output accuracy.

What are Parascript’s future endeavors?

We recently announced the release of our Cascade Classifier. At a very image-based level, classifiers are trained to recognize different documents such as proofs of income and appraisals. However, there would be cases where the classifiers will not be able to differentiate a particular application or document type from another. The only way it can be optimally tuned is by analyzing all of the output, identifying where the system has made mistakes and then making adjustments which can take literally hundreds of hours. Our cascade classifier, offering 100 percent automation of the analysis and tuning, effectively fulfills these obligations. As a result, the final output is 25 percent better than any solution in the market.

With that, we are also geared toward bringing scalability of the document automation solution, making it more accessible to small and medium-sized enterprises. Most of these small companies do not have the technology expertise or time to adequately configure and optimize a document automation system. Thus, they resort to processing methods that are sub-optimal or continue using manual workflows and processes. Our upcoming release is targeted toward addressing this problem. With our new release—an advanced machine learning capability—any mid-sized business can take our software and optimize their processes using various samples to train the software. We believe this capability is going to bring a sea change in the adoption of document automation, especially robotic process automation that increasingly needs access to document-oriented data.