'Intelligent Ingestion': Time to Truly Go Digital

The industry kids itself about having gone paperless. In fact, we still use the same processes we used in the 17th century. It's time for a change.

Skyscrapers against a dark blue sky with a cyber circle in the background


--A typical process involves routing the submission to a lower-cost resource, often offshore, who might extract and convert 50 of the 500 important pieces of information in it into digital data.

--But tools that combine natural language processing with computer vision can now extract all the key data from both structured and unstructured documents with great accuracy and speed.


For over a decade, we have pretended to be paperless. We tout the fact that the file cabinets, the seven-part specialized file folders and massive mail rooms have all been eliminated and reduced, and we claim we are paperless. But we know we aren't. Today’s paper is PDFs, Excels and Adobe.

The filing cabinets have been replaced with digital folders and the mail rooms with emails and digital workflows. But insurance isn’t paperless; it just pretends to be.

It doesn’t have to be that way. The technology to become truly digital exists. We just need to take the leap.

One of the biggest hurdles to being truly digital as opposed to being a digital paper industry involves the initial ingestion or digitization of the data. We have had advanced OCR and computer vision solutions for a while. These have been great at extracting the information from digital forms and standardized templates but have not met the needs of more complex undertakings such as commercial insurance submissions.

The typical commercial insurance new business submission or quote request can contain an application, loss runs, statement of values, insurance certificates, financial statements and many other documents, depending on the type of insurance. A typical commercial insurance submission will contain 300 to 500 pieces of information. Information that is invaluable in understanding, evaluating and quoting a piece of business.

The untapped potential of dark data

Our processes today to extract the data from these submissions are archaic. A typical process involves routing the submission to a lower-cost resource, often offshore, who will extract a minimum set of fields to set up the submission and some basic rating information. On a good day, they might extract and convert 50 of the 500 pieces of information into digital data by entering it into the system. The rest is left in the documents as dark data. Data that the carriers have, but that is never digitally exposed or available.

Then the electronic file, carrying the electronic documents, is sent along to the underwriter, where these digital documents are opened again and again because the data isn’t available. Other than the file folder and the documents being made out of bits and bytes rather than paper and ink, it is the same process as 300 years ago.

It doesn’t have to be that way. We have seen what insurance can become with speed, efficiency and precision when the process is digital with simpler and homogeneous risks in personal lines, but the promise is there for more complex insurance such as group benefits, commercial lines and specialty insurance. And it starts by being able to digitally ingest the data.

This is where we have seen a technological leap. With the integration of more advanced machine learning tools that can combine natural language processing with computer vision, data can now be extracted from both structured and unstructured documents with high degrees of accuracy and speed. In fact, this is one of the hottest emerging technology areas in insurance today with a wide array of players and investors. Take the case of a life insurer in China that has deployed an intelligent risk control system that enables end-to-end automation of insurance applications. (See, Fuel the future of insurance)

See also: Seeing Through Digital Glasses

Learn from the emerging leaders

One of the other emerging leaders in this space is a company called MEA out of the U.K. What makes MEA unique is that it was founded by insurance executives who understand the unique challenges involved in complex insurance documents as well as having a key understanding of the terminology, variability and complexity involved. Their solution has focused on building deep expertise and a broad insurance-specific extraction catalog around core insurance concepts, starting with submission documents that allow their solution to be quickly adapted to new insurance areas. The best part is that, because their team deeply understands insurance, working with them does not require you to train their team on what insurance means.

We have worked with MEA on several engagements and tests throughout Europe and the U.S. The breadth of their solution has allowed us to evaluate a wide range of lines of business, business processes and insurance entities including carriers, MGAs and brokers. They can consistently compete in terms of speed, accuracy and quality in their testing and execution. It is really possible to be going from evaluation to use of this type of solution within a few short months.

So, what does this mean for our digital paper world today? Well, it means insurers now have a real choice to begin a digital journey. This has been the hope and dream for a while, but technology has really caught up to that vision of being digital – starting with intelligent ingestion.

Creating truly touchless processes

There are several different ways in which to now employ it. It starts by identifying a pseudo-paperless process that exists in your organization today and targeting the documents that it ingests. Submissions are an obvious choice, but claims, bordereaux, invoice receipts, audits, etc. are all also possible. Then design how you want the digital process to work. You can choose to ingest and directly process the data or take a more cautious approach that still includes some level of human review or human insight.

The choice should depend on the complexity and significance of the data and your comfort with implementing it, but long term you should expect that at least some portion of your ingestion will be able to be touchless. The other decision to make is whether you are only going to extract the data that you use today or do you want to extract everything in the document. This is the 50 versus 500 question for submissions. But you may require some other changes and other technology to aid a true digital transformation. We will discuss those elements in a future blog.

In the meantime, however, isn’t it time that your insurance process was no longer from the 17th century? Isn’t it time that we moved from passing along the digital paper in email and workflow systems to building truly digital processes? Isn’t it time to start to build your company’s intelligent ingestion solution? Let’s start to build real digital insurance.

You can find this article originally published here.

Michael Reilly

Profile picture for user MichaelReilly

Michael Reilly

Michael Reilly is a managing director in Accenture's insurance strategy practice.

He has 20 years experience helping insurance companies to transform underwriting operations and organizations around the world; he has led large-scale commercial insurance transformation programs in underwriting, policy, business intelligence and mergers and acquisitions.

Reilly has also co-written and presented multiple articles on underwriting, analytics and knowledge management and worked at General Accident Insurance prior to joining Accenture.


Read More