In today’s mortgage market you can’t pick up an industry publication or attend a trade show and not hear someone talk about Artificial Intelligence (AI). Many of the claims talk about how AI will disrupt the mortgage industry or radically change how mortgages are originated or processed.
Some of it is pure hype and some of it includes technology that can clearly streamline processes and reduce cost. So how do you filter fact from fiction? Let’s take a look at how AI-based textual analysis is actually at work in the mortgage industry and the different approaches some vendors are using.
The most fundamental difference in approach between an AI textual analysis and the majority of other “advanced” document recognition technologies is that AI textual analysis treats variable layout documents as unstructured documents whereas most other prominent solutions treat them more as semi-structured documents.
To illustrate the difference in methodology let us consider a customer wishing to process pay stub documents. An AI textual analysis implementation would typically be deployed with one set of completely generic rules designed to encompass all variations of the “Pay Stub” document type from any company. Because alltext is evaluated by the AI engine, the rules can flex with the layout and verbiage changes just as a human does when reading the page. The solution is also capable of being configured to perform conditional processing for specific exceptions to generic rules (per-layout exceptions) but it is not typically necessary to do this.
Most other modern advanced document recognition technologies treat document variations as semi-structured documents. These solutions typically either:
a) Remember as many of the variations as is practical and process each variation with layout-specific templates for processing Or b) Apply a mix of layout-specific processing and some generic processing (usually the higher incidence layouts are processed with layout-specific or templated processing)
The obvious advantage of the AI textual analysis generic approach is that, as new layouts appear or existing layouts change, the software is better equipped to deal with these new variations. This advantage was recently validated at an account with over 35,000 layout variations where rules had been in place for five years with not a single modification. An audit of this client’s processes after five years revealed an identical automation rate to that when the system was first deployed. This outcome was observed despite the fact that a significant portion of the originally dominant layouts had been transferred to EDI processingand were therefore bypassing the system now.
As you can see the varying approaches often produce different levels of success. But how can you determine which is best for your organization? Get a demo and hope that the solution is more than smoke and mirrors? Up until now that is usually the case. To overcome much of the confusion and disappointment, a better evaluation process in many cases may go a long way towards greatly minimizing the risks involved in choosing a vendor who can actually deliver AI-based textual analysis.
In order to quickly understand AI-based textual analysis and its capabilities, a blind test with several sample files should be considered the gold standard for an evaluation. This is especially true when it comes to the challenges presented with the many and varying document types and quality levels of document images found in the mortgage industry. Asking vendors if they are willing to perform a test on a never before seen sample set of typical loan files on site and in sight of your evaluation team, is a great first step in separating fact from fiction.
Ideally, an evaluation should be setup as a one day event to hedge against any vendor refining their results. This kind of test is intended to demonstrate the validity of the vendors’ out-of-the-box capabilities so that prospects can be assured that they are considering a proven, robust and scalable solution ready to deliver productivity improvements in weeks rather than months or years.
For qualified opportunities, Paradatec will perform this process, which enables prospective clients to quickly understand the overall levels of automation, and speed improvements they will be able to achieve with their technology. Download our whitepaper on AI based textual analysis https://www.paradatec.com/wp-content/uploads/2018/12/CompetitiveMethodologies_2018_Final.pdf.
About The Author
Mark Tinkham is Director of Business Alliances at Paradatec, Inc. Over the past twenty-five plus years, Mark has worked for technology companies that deliver innovative solutions to the financial services industry. For the past ten years, his primary focus has been bringing efficiencies to the mortgage market through industry leading Optical Character Recognition (OCR).