Tech, ExplainedFintech
Product

AI for insurance submissions and claims

The submission, read and filled by AI.

BEFORE ITAFTER ITAI for insurancesubmissions and claimsFurtherAICOVERSDocument readingSubmissionsQuote comparisonLoss noticesInsurance agencymanagement systemsends theclient'sdocumentssends thefilledsubmissionUnderwriter workbenchopensthe claimInsurance core systemCONCEPTS IT USESHuman in the loopBEFORE ITAFTER ITAI for insurancesubmissions and claimsFurtherAICOVERSDocumentreadingSubmissionsQuotecomparisonLoss noticesInsurance agencymanagement systemsends theclient'sdocumentssends thefilledsubmissionUnderwriterworkbenchopensthe claimInsurancecore systemCONCEPTS IT USESHuman in the loopBEFORE ITInsurance agency management system↓ sends the client's documentsAI for insurancesubmissions and claimsFurtherAIDocument readingSubmissionsQuote comparisonLoss noticesAFTER IT↓ sends the filled submissionUnderwriter workbench↓ opens the claimInsurance core systemCONCEPTS IT USESHuman in the loop
AI for insurance submissions receives a client's documents from the agency management system, sends the filled submission to the underwriter workbench, and records each first notice of loss in the insurance core system.
opens a page inside this one a page in the same scene a page in another scene a concept page a label only, with no page

AI for insurance submissions and claims is software that uses an AI model to read insurance documents and fill in the forms that people used to type by hand. It does this at two points: when a business applies for cover, and when a customer first reports a loss.

The first point is the submission. A submission is the application for cover that a broker sends to several insurers on behalf of a business client. To prepare it, the broker reads the client's documents: old policies, lists of buildings and vehicles, and records of past claims. These often arrive as PDFs, spreadsheets and scanned pages. A large language model (LLM) can now read them and fill in the submission. When the quotes come back, the same model can put them side by side, with the price, the cover and the exclusions of each. Exclusions are the losses that a policy does not cover.

Take, for example, a broker preparing a submission for a food-processing company with three plants. The client sends 64 pages of documents. The AI model reads them in about five minutes. It fills in 180 fields of the submission, including each plant's address, building value and claims for the last five years. It marks 6 fields where the documents disagree with each other. The broker checks those 6 fields, corrects two of them, and sends the submission from the insurance agency management system. At each insurer, an underwriter reviews it on an underwriter workbench.

The second point is the first notice of loss (FNOL), the first report that a customer makes after something goes wrong. It may be an email, a call recording or a form with photographs. The model reads it and picks out the policy number, the date, what happened and what was damaged. Then it opens the claim in the insurer's claims management system.

Older tools for this job are called intelligent document processing (IDP). They read fixed fields from forms whose layout they already know. An LLM can also read documents it has never seen before, in any layout. But it can make mistakes, so a person checks its work before anything is sent or paid. This arrangement is called human in the loop.

Tech, Explained is a part of ingrid.fyi. Every page explains one piece of technology in plain words.