Documents that contain unstructured data present problems for most insurance companies’ insurance policy automation solutions because they are not easily digestible by computers.

Claims processing, for example, often involves a review of notes from an adjuster based on conversations with the claimant. The adjuster’s notes are primarily free-form, following no standard format, with plenty of variation from one adjuster to the next. An adjuster’s file may also include photos, along with reports from doctors and lawyers. In short, it’s a significant amount of data, nearly all of it unstructured – and hence a prime candidate for claims process automation.

To process a claim, someone has to pore through hundreds of pages of documents, extract pertinent bits of information and input them into a downstream claims processing system. Relevant data may include the claim number, policy number, date and time of loss, location, coverage limits, and more.

That data entry job is labor-intensive and time-consuming, not to mention error-prone, making it ripe for insurance document automation. Companies have tried using insurance process automation tools based on keywords and rules, with less-than-stellar results.

Keyword and rule-based approaches to insurance automation use templates that define precisely where the data you want to extract is located in a given document, along with a slew of rules defining what data to extract and what to do with it.

In practice, an insurer often hires a consulting firm to write countless rules and templates to try to account for every variation in the documents the company needs to process. That may work, briefly – until a document comes along that doesn’t neatly fit into any of the rules or templates the consultants created. Once this happens, the entire system breaks down – and that’s why you need to deploy intelligent automation in insurance.

If you think of an insurance adjuster’s notes, it’s easy to see how it won’t take long at all until a new type of document comes along, making the rule-based approach all but futile. On top of that, it’s horrifically costly, whether you use outside consultants or in-house resources.

Automate Complex Insurance Process Such as Underwriting, Claims, Customer Onboarding & More.

SCHEDULE YOUR DEMO TODAY

OCR is machine learning technology that can convert documents such as PDFs into a machine-readable format. That’s useful, but it still leaves you dealing with templates to extract useful information and all the issues that still exist.

Robotic process automation for insurance suffers from much the same problem when it comes to these types of use cases and insurance workflow automation in general. It’s why we often hear that robotic process automation for insurance claims, and other use cases, falls short.

There are a number of RPA use cases in insurance. RPA is great at automating processes that involve the exact same steps each time. Say, for instance, an insurance data entry clerk entered the exact same keystrokes in the same order time after time into a claims processing system. That would be a process that’s ripe for RPA.

But, as explained above, that’s not at all how the process works. Rather, it requires a human being to make judgment calls about which data to extract and enter. Any insurance process that involves unstructured documents – which is most of them – will suffer the same problem. (RPA can, however, complement IPA in an insurance automation solution. More on that here.)

As implemented by Indico, intelligent document processing technology is fundamentally different from RPA and templated approaches because IPA can understand document context much like a human does. That’s because it’s based on a model that incorporates some 500 million labeled data points, enough to enable it to understand human language and context.

Having a large set of data to “train” brings intelligence to any artificial intelligence solution. But to utilize AI in insurance, even the largest insurance companies would be hard-pressed to create their own model based on that much data.

Indico’s Unstructured Data Platform then applies technology known as transfer learning to create custom models that can tackle virtually any downstream task – including claims automation and other common insurance use cases. The result is insurance companies can automate processes using a relatively small number of documents to train the model. What’s more, you don’t need data scientists to make it all work. Rather, the professionals on the front lines train the model – those who know the processes best. (For a deeper dive, check out our Intelligent Process Automation page.)

Applying artificial intelligence to insurance claims

Advanced intelligent document processing platforms employ artificial intelligence technologies including natural language processing, machine learning and transfer learning. Together, they enable employees who process insurance claims day-to-day to build claims process automation models – without involving data scientists or even IT. That helps deliver a higher level of model accuracy and significant scalability in terms of an insurer’s ability to roll out automation models across the company.

Insurance claims automation: email processing

Chief among the areas that are ripe for automated claims processing is the corporate email in-box. It’s common for insurance companies to have a single email address to which clients send claims information. Some use the same address for general information requests, including quotes on new business. Or, customers mistakenly use a generic address to submit a claim.

Many of these emails contain not only potentially complex requests, but attachments, such as ACORD forms, photos showing home or vehicle damage and the like. An intelligent document processing platform can essentially read and triage each email, determining where each should be routed.

The platform can also extract and read attachments, often dealing with them on its own. The Indico Unstructured Data Platform, for example, includes a large library of ACORD forms, often resulting in straight-through processing of the forms. In general, the Indico Data platform can ensure each email is entered to the correct workflow for efficient, automated processing.

Automating first notice of loss (FNOL)

No matter how an insurance company receives word of a client claim, it’s sure to come with plenty of documentation. Much of it will be unstructured, making it beyond the scope of a robotic process automation solution.

An intelligent document processing tool, however, can process any type of document, structured or unstructured. In the insurance first notice of loss scenario, that may mean accepting the initial claim (the FNOL), then validating the claim is covered by the client’s policy. Much of that process can be automated with a tool that “reads” the claim form, extracts pertinent data and inputs it to a claims management tool. Here again, STP may apply to simple claims while others can be prepared with all pertinent information already assembled for an adjuster, greatly reducing time spent on the assessment process.

Meet compliance requirements with explainable AI

A key concern in a highly regulated industry like insurance is the ability to explain why an automated, AI-based solution makes the decisions it does. If a claim is rejected, the insurance company must be able to explain in simple terms the reason behind that decision. A sound intelligent automation tool will make that easy, with an audit trail that makes clear – in plain English – the rationale behind each claim decision. Additionally, given insurance claims process owners create the automated models, the models naturally reflect the way they think during the claims assessment process.

Outsmart insurance fraud perpetrators

Automation is crucial for insurance companies if they are to keep up with insuretech startups, or simply compete effectively vs. traditional competitors. It’s also important to helping detect fraud.

More than 7,000 insurance companies collect over $1 trillion in premiums each year, the FBI estimates, making them a prime target for illegal activity. The total cost of insurance fraud exceeds $40 billion per year, the FBI says – and that doesn’t include health insurance. That costs the average U.S. family between $400 and $700 per year in increased premiums.

Automation can help insurance companies do better. To learn more, see our page, Intelligent Document Processing for Property & Casualty Insurance or click below for an interactive demo, a free trial or to get in touch with any questions.

Some insurance automation use cases in life and health insurance may involve both robotic process automation and intelligent process automation, as complementary technologies.

RPA is great at automating repetitive tasks to make a process less labor-intensive for humans and works well with deterministic business processes that involve structured data. IPA, on the other hand, can automate processes that involve unstructured data.

A common IPA and RPA use case, then, is to use IPA to “read” unstructured data and translate it into a structured format that an RPA tool can then process. For example, the RPA ingests documents and send them to IPA solution for classification and data extraction. Once extracted, the IPA tool puts the data into a structured format that the RPA tool can work with, such as a spreadsheet. The RPA tool can then take in the now-structured data and populate a downstream system, such as a customer relationship management (CRM) tool.

Keeping up with the pace of business is difficult under any circumstances. But if insurance companies are to achieve digital transformation, it’s imperative. Indico’s intelligent document processing solution can help in that effort while delivering significant benefits, including: