The Digital Gold Rush: Insurance Document Prospecting, Refining, and Taking Action

Most leaders at insurance carriers today would agree that data is one of their company’s most valuable assets, but most would also admit that one area they could improve is documents (i.e., unstructured data).

Common questions and challenges:

• The number of documents is increasing while the number of hours in a workday is not, and we are all under pressure to do more with fewer people.
• How much money and productivity are we losing looking for information and what is the cost of not finding it?
• Are we playing “Concentration” with our documents (i.e., child’s game where you turn over cards to try to find a match). Searching through documents can be like that with many windows open trying to piece together a bigger picture or story.
• Are we getting that information to the person who needs it to act when it can make the most impact?
• How do we help knowledge workers be more efficient, effective, and happier (not just create more distractions for them).

Documents contain a wealth of information that frequently does not exist in any insurance business system, because it is unstructured, highly variable, and hard to extract/manage. This may be your opportunity to start your own “digital gold rush” to realize untapped value.

Mining is a useful analogy when examining the handing, processing, and using documents:

• Prospecting (Identifying Documents): Locating the “gold’ in claims forms, customer communications, policy documents, etc.
• Exploration (Document Classification): Documents are classified similarly to the way mining sites are surveyed.
• Drilling (Extracting Metadata): Data is extracted like ore from the earth.
• Planning and Designing (Workflows): Mining engineers design the most efficient way to extract gold, AI engineers do the same with data.
• Crushing and Grinding (Data Cleaning): Data becomes information as additional services are applied.   Extracted data may need cleaning or transforming, like crushing the ore.
• Separation Process (Data Integration): Gold from other particles.  Extracted data is enriched, integrated, and stored for later use.
• Refining (Analysis and Actions):  The process yields richer insight that can be used to initiate action.
• Reclamation (Data Maintenance): It’s important to manage data like any natural resource.

Let’s examine these in the context of a typical insurance carrier and their challenges:

• Prospecting (Identifying Documents) & Planning/Designing (Workflows to Ingest Documents)

Every insurance company has “at least” one and probably multiple repositories where they accumulate documents as they arrive from the post office, email, vendors, consumer uploads, and not to mention all the documents insurance companies generate as part of doing business (e.g., policies, bills, reports, audits, etc.).

• Exploration (Classification), Drilling (Metadata Extraction), Crushing and Grinding (Cleaning)

Common barriers to effective document handling start with getting them into a repository with metadata.

• Document volume and variety
• Documents with no or inconsistent structure, ambiguity and “noisy”
• Limited scanning, OCR and data extraction capabilities
• Classification, tagging and indexing can be manually intensive

• Separation (Data Integration) Many companies stop short here by not enriching their documents with other data sources or integrating documents and byproducts into their business processes/systems.  Until recently, the technology to extract intent and other information was not as capable or available to companies as it is today.
• Refining (Analysis and Actions): The final stage of refining in gold mining can be compared to the process of analyzing the processed data to derive actionable insights like trend identification, anomaly spotting and predictions.   Getting actionable information to an adjuster, underwriter, actuary, etc. when it’s needed is crucial.
• Reclamation (Data Maintenance): Like returning the land to its original or better state, maintaining the quality of stored data over time through regular audits, updates, and error corrections is essential.

How Innova’s Upcoming IRIS Framework will Address these Challenges Through the Full Life Cycle

Innova’s approach to Intelligent Document Processing enabled by artificial intelligence will be encapsulated in our IRIS Framework which stands for Insurance Research Intelligent Synopsis.

• Prospecting (Identifying Documents): We work with you to identify your documents, how they come in, where they are stored and how you “would like to use them” in your business.  We provide/build connectors as needed to access your documents.
• Exploration (Document Classification): Machine learning models are used (framework) or custom classifiers created to classify your documents for routing through pipelines and to notify people based on intent/interest/role.
• Drilling (Extracting Metadata): Relevant metadata is extracted from the documents by file type to support additional steps in the pipeline, enabling enrichment, and indexing for searches.
• Planning and Designing (Pipelines/Workflows): Our AI “mining” engineers design the most efficient way to extract your gold. We work with you to design workflows to efficiently review, process, and administer them for effectiveness and safety.
• Crushing and Grinding (Data Cleaning): AI technologies like OCR, entities, key phrases, captions, summarization, etc. All documents are vectorized to index them semantically. This and all other relevant extractions and derivations are indexed for searchability.
  Extracted data frequently needs to be cleaned or transformed, like crushing the ore. This might involve identifying irrelevant data, resolving discrepancies, standardizing data format and business record match/confirmation. “Human in the loop” (HITL) processes are provided to resolve exceptions that require or merit their attention.
• Separation Process (Data Integration): Just like separating gold from other particles, the extracted data is integrated and stored in appropriate data stores for future use.  This includes various forms or enrichment to make routing/”refining” more targeted, accurate and efficient.
• Refining (Analysis and Actions): The final stage of refining in gold mining can be compared to the process of analyzing the data to derive actionable insights.
  • Identifying trends, spotting anomalies, predicting future scenarios, etc. using machine learning models for clustering, classification, and regression.
  • Notifying interested parties who have subscribed to document event topic(s) by intent/entity, providing them with a document summary, explaining why the event was triggered and facilitating access to the full document for their review.
  • Conducting conversational, semantic search to obtain answers to natural language questions, not just a list of matches for a literal search
  • Generating custom and standard reports using generative AI to find and organize the information to make the task easier.
• Reclamation (Data & Process Maintenance): Managing the integrity of data through regular audits, updates, and error correction is essential. In addition to basic data conservancy, we must hold ourselves accountable to how we use AI to ensure it is ethical, fair, reliable, safe, secure, inclusive, and transparent. This requires autonomous and human-in-the-loop monitoring to be built into both our systems and our processes.

In addition to IRIS, we are creating an assessment process called CORNEA, which stands for Current Operations Review and Novel Enhancements Assessment where we will examine your document handing processes/technology to identify gaps and make recommendations which can include IRIS and other automations.

Innova Solutions is currently working on IRIS and will be launching the first iteration soon. Watch our website or reach out to your client partner for more information.