The State of Third-Party Insurance Verification – Research Report
November 11, 2021
Analog vs Digital? You Need Both.
The insurance industry isn’t well known for its quick adoption of cutting-edge technology, but the draconian process of verifying certificates of insurance to prove that a third party meets corporate requirements is not going away anytime soon.
Enterprises should have a combination of processes that meet today’s needs while technology companies work to build a more digital future for insurance verification, which may take some time and require some hand-holding. Most risk managers are only familiar with the concept of using human eyes to scan paper COI documentation, so when a digital verification solution offers to automate the task, they’re understandably skeptical.
Having an insurance verification solution that incorporates both physical and digital documentation is effective for two reasons: it gives enterprise risk managers the flexibility to verify and manage COIs as they choose, and it slowly introduces the concept of verifying COIs digitally to help traditionalists ease into digital insurance verification.
The question then becomes, how can enterprises digitize the static, standard COI documents so that artificial intelligence technologies can efficiently and accurately verify them, with the goal of eventually automating this highly manual process?
Digitizing Standard Documents
Document digitization has been around for a long time, but the fundamental reason why it’s so difficult is because it has to be so accurate that it doesn’t generate more human labor than is necessary to help the technology function properly. If a document is digitized slightly incorrectly, it may generate not just a little bit of human work to correct errors, but rather, as much – if not more – work to get everything right than it would have if a human being just did it themselves from the get-go.
Document digitization requires sophisticated technology, and Optical Character Recognition (OCR) – which is more of a utility technology in this application – is not going to be the solution. OCR can be applied to a page of free-form text, like a crossword puzzle, and data can be extracted that way, but that information isn’t necessarily usable.
For an automation pipeline to work properly, the data has to be right and the machine has to know when it’s failed and it shouldn’t create more work for humans. The technology not only has to be accurate and precise, but it also has to be done at scale, and that is what makes document digitization a fairly difficult and sophisticated problem that most organizations don’t want to deal with.
Digitizing a standard document like an Acord 25 is not about character recognition, it’s about partitioning. The text must be accurately partitioned in order to extract the useful components so they can be pulled into a database with other information that’s relevant to a particular field so an enterprise can process and sort the data types by category.
How Evident Works
Evident’s technology is leveraging machine learning techniques to extract meaningful information from Acord 25 documents almost 100% of the time, with near-perfect accuracy for many of the fields. Our technology has created a fine dividing line between what data we have and what we know has been parsed accurately so we know, specifically, where we need human assistance.
Evident has found that the difficulty in digitizing these standard documents isn’t “bad scans,” it’s the nature of PDF images that are compressed so they can be sent via email and then are reconstructed into a format that has small artifacts and pixelation properties throughout the image, making it difficult for our techniques to accurately find and partition the data.
This isn’t necessarily a weakness, rather, it’s just an added layer of complexity. Sometimes, image quality can degrade to a level where it’s difficult for anyone to decipher, much less a sophisticated software solution that has been accurately extracting personal data from standardized identification documents for many years.