US-20260127676-A1 - AUTOMATED DOCUMENT ANALYSIS AND CERTIFICATE OF INSURANCE GENERATION FROM UNSTRUCTURED INSURANCE COVERAGE DOCUMENTS

US20260127676A1US 20260127676 A1US20260127676 A1US 20260127676A1US-20260127676-A1

Abstract

In an illustrative embodiment, systems and methods for automatically generating a certificate of insurance (COI) include extracting, from a request document, certificate of insurance parameters detailing a request for COI, extracting, by at least one artificial intelligence classifier from unstructured insurance policy document(s), policy attribute values, converting each policy attribute value to a vector form, applying artificial intelligence to analyzing the vector forms in light of a knowledge ontology to identify a set of COI field values for populating a certificate of insurance template, and populating the certificate of insurance template with the set of COI field values to generate the COI.

Inventors

Sudipto Shankar Dasgupta
Raju Debroy
Abdul Razack
Shazia Pappa
Liam Montgomery
Erin Alder
Anthony Amaro
Amanda Van Heck
Russell Lee Antrim

Assignees

AON RISK SERVICES, INC. OF MARYLAND

Dates

Publication Date: 20260507
Application Date: 20241101

Claims (20)

1 . A system for automating creation of ad hoc certificates of insurance, the system comprising: one or more first non-transitory storage devices configured to store a knowledge ontology comprising semantically linked relationships between a plurality of insurance policy attributes and a plurality of fields within at least one certificate of insurance template; one or more second non-transitory storage devices configured to store semantically linked high-dimensional vector representations of portions of each insurance policy document of a set of insurance policy documents, wherein the semantically linked high-dimensional vector representations comprise a plurality of insurance policy attribute values corresponding to at least a subset of the plurality of insurance policy attributes; one or more third non-transitory storage devices configured to store at least one large language model (LLM) tuned to assess completeness and accuracy of a subset of a plurality of insurance policy attribute values to be used for certificate of insurance generation; and processing circuitry configured to identify, via a widget, bot, or extension to an email program, availability of an email, collect, from the email, an email data set comprising a metadata portion including a subject line and a sender identifier, and a body text portion, identify, within the email data set, a plurality of certificate of insurance parameters detailing a request for a certificate of insurance, wherein the plurality of certificate of insurance parameters comprises information identifying a requesting party and one or more of a policy number, a policy holder name, a policy holder contact information, or a type of insurance policy, apply a business relationship-directed semantic ontology comprising corporate relationships to enrich details regarding the policy holder contact information and the requesting party, upon identifying a discrepancy between one or both of the policy holder contact information or the requesting party and the business relationship-directed semantic business ontology, present, at a first interactive graphic user interface, any inconsistent information to obtain correction. using the email data set including the enriched policy holder contact information, identify at least one unstructured insurance policy document, apply at least one natural language processing technique to extract, from each unstructured insurance policy document of the at least one unstructured insurance policy document, a set of insurance policy attribute values corresponding to at least a portion of the plurality of insurance policy attributes, classify, by at least one trained classification model, each respective insurance policy attribute value according to a classification schema organizing the plurality of insurance policy attributes of the knowledge ontology, convert each respective insurance policy attribute value of the set of insurance policy attribute values to a numeric format arranged in a respective high-dimensional vector form of a set of high-dimensional vector forms, store the high-dimensional vector forms of the set of insurance policy attribute values to the one or more second non-transitory storage devices, apply a respective attribute tag of a plurality of attribute tags to each high-dimensional vector form of the set of high-dimensional vector forms according to the classifying, wherein at least a portion of the plurality of attribute tags reference a respective data entry field of a plurality of data entry fields of a topic certificate of insurance template of the at least one certificate of insurance template, prepare, for presentation on a display of a remote computing device, a second interactive graphical user interface for reviewing each respective insurance policy attribute value of the set of insurance policy attribute values and a descriptor corresponding to the attribute tag applied to the respective insurance policy attribute value, wherein the descriptor corresponds to a label of an information field of the topic certificate of insurance template, receive, responsive to presenting the second interactive graphical user interface, one or more modified insurance policy attribute values, using the one or more modified insurance policy attribute values, update training of the at least one trained classification model, and update the set of high-dimensional vector forms according to the one or more modified insurance policy attribute values, cluster the set of high-dimensional vector forms into semantic relationships linked within the one or more second non-transitory storage devices, responsive to an indication of acceptance, received via the second interactive graphical user interface, of the set of insurance policy attribute values, analyze, by the at least one LLM, a plurality of vector representations stored to the one or more second non-transitory storage devices to identify a set of insurance policy attribute values of the plurality of insurance policy attribute values relevant to the request for the certificate of insurance, wherein the plurality of insurance policy attribute values comprises one or more values corresponding to one or more of an effective date, an expiration date, a limit, or a deductible, and the plurality of vector representations comprise at least a portion of the set of vector representations of the at least one unstructured insurance policy document, match each data entry field of a plurality of data entry fields of the topic certificate of insurance template with a respective one or more insurance policy attribute values of the set of insurance policy attribute values or one or more certificate of insurance parameters of the plurality of certificate of insurance parameters according to a set of attribute tags of the plurality of attribute tags corresponding to a set of vector representations of the plurality of vector representations used in identifying the set of insurance policy attribute values, wherein the matching is performed in accordance with the knowledge ontology, present, on the display of the remote computing device, a third interactive graphical user interface for reviewing and revising a digital rendition of the certificate of insurance, the digital rendition of the certificate of insurance comprising a) based on the matching of each data entry field with the respective one or more insurance policy attribute values, a plurality of filled data fields comprising values of at least a portion of the plurality of insurance policy attributes and at least a portion of the plurality of certificate of insurance parameters, and b) at least one edit control configured to enable editing of at least a portion of the plurality of filled data fields, receive, responsive to presenting the third interactive graphical user interface, one or more modified values, each modified value corresponding to a different insurance policy attribute of the plurality of insurance policy attributes, and using the one or more modified values, convert the values of the plurality of filled data fields into at least one formal COI certificate document.
2 . (canceled)
3 . The system of claim 1 , wherein identifying the plurality of certificate of insurance parameters comprises applying at least one natural language processing technique to extract at least a portion of the certificate of insurance parameters from the email data set.
4 . (canceled)
5 . The system of claim 1 , wherein the semantically linked high-dimensional vector representations are stored within a semantic graph structure.
6 . (canceled)
7 . The system of claim 1 , wherein the at least one LLM matches each data entry field of the topic certificate of insurance template with the respective one or more attribute values.
8 . The system of claim 1 , wherein the digital rendition of the certificate of insurance comprises a thumbnail image of each page of the certificate of insurance.
9 . (canceled)
10 . The system of claim 1 , further comprising one or more fourth non-transitory storage devices configured to store at least one machine learning model configured to recognize the plurality of insurance policy attributes, wherein the at least one natural language processing technique formats content of the at least one unstructured insurance policy document for processing by the at least one machine learning model; and the at least one machine learning model extracts the set of insurance policy attribute values.
11 . The system of claim 1 , wherein clustering the set of high-dimensional vector forms of the set of insurance policy attribute values comprises clustering the high-dimensional vector forms into semantic relationships linked within a knowledge graph, wherein the knowledge graph is stored to the one or more second non-transitory storage devices.
12 . The system of claim 1 , further comprising a second knowledge ontology comprising semantically linked relationships between a plurality of business entities, wherein: the plurality of business entities comprises a plurality of insurance policy holders; and the processing circuitry is further configured to analyze at least a portion of the plurality of certificate of insurance parameters in view of the second knowledge ontology to enhance entity details related to at least one of the policy holder name, the policy holder contact information, a certificate holder, or an insurance carrier.
13 . The system of claim 1 , wherein at least a portion of the processing circuitry comprises hardware logic hard-coded or programmed into the portion of the processing circuitry.
14 .- 20 . (canceled)
21 . The system of claim 1 , wherein the at least one LLM is fined-tuned using a respective corpus of documents comprising insurance policies and certificates of insurance.
22 . The system of claim 21 , wherein the at least one LLM comprises a plurality of LLMs, each document of the respective corpus of documents used to fine-tune a respective LLM of the plurality of LLMs being truth labeled according to a different business ontology of a set of business ontologies.
23 . The system of claim 22 , wherein the set of business ontologies comprises at least one of a customer services ontology, an insurance policy ontology, or a certificate of insurance formatting ontology.
24 . The system of claim 1 , wherein analyzing, by the at least one LLM, the plurality of vector representations comprises prompting the at least one LLM to apply a portion of the plurality of high-dimensional vector forms and the knowledge ontology to collect and provide a set of field values for populating a plurality of attribute fields of the topic certificate of insurance template.
25 . The system of claim 1 , wherein identifying the at least one unstructured insurance policy document comprises identifying at least one attachment to the email as the at least one unstructured insurance policy document.
26 . The system of claim 1 , wherein the processing circuitry is further configured to analyze the email data set to identify the topic certificate of insurance template from the at least one certificate of insurance template, wherein the at least one certificate of insurance template comprises a plurality of certificate insurance templates.

Description

BACKGROUND A certificate of insurance (COI) serves as proof that a business or individual holds an active insurance policy. A COI is often requested by businesses or individuals before getting into a contract or agreement for any commercial business transactions. It provides key details of the insurance policy (e.g., certificate holder's name, producers, insurance carrier, policy number, effective dates, types of coverage, limits, additional insured, endorsements, etc.). In the commercial insurance business space, it is common practice for entities to share property and casualty insurance coverages for a variety of reasons, including loan closings, building access, and regulatory. Often one or both entities will require very specific details to be displayed on the COI. The COI producers (e.g., brokers) must gather these specific details from a collection of unstructured documents and painstakingly validate the resultant certificate against the policies to ensure the COI is in fact properly representing the coverage. A COI produced in between renewal periods are referred to as “ad hoc.” Brokers receive hundreds of thousands of requests for ad hoc certificates each year. Ad hoc certificates are often difficult to forecast and needed urgently for a specific business reason, creating friction between the broker, the client, and (where applicable) the client's business partner. Oftentimes, the information necessary to fill in the key details for the ad hoc certification is scattered across various documents retained by various systems. Further, there is limited standardization among the generally manual workflows used across clients and industries for generating ad hoc certificates. Hence, this process is highly manual, time-consuming, and error prone, resulting in a poor client and broker experience. The churn in this process creates material business expenditures for the client and the broker. The inventors recognized that a highly automated mechanism for generating certificates of insurance, especially ad hoc certifications, would create many benefits, including reducing the time for certificate generation, reducing mistakes within the document typically introduced through human error, and aiding producers in consistently producing a more accurate COI. These benefits would result in both a better end user experience for both clients and brokers, as well as significantly reduced operational expenditure. SUMMARY OF ILLUSTRATIVE EMBODIMENTS BRIEF DESCRIPTION OF THE DRAWINGS In one aspect, the present disclosure relates to automatically extracting detailed information from insurance certificate request documents, comparing the information with existing policy documents, and flagging any deviation within the details of the collective documents. A fine-tuned Large Language Model (LLM)-driven extraction of the detailed information from the policies and contract documents, for example, may be augmented with an insurance-specific ontology to drive processing of the output in a standardized format despite variation in the source document formatting (e.g., due to variations in carrier document templates, variations in terminology applied by the drafters, etc.). Flagged deviation(s), for example, may be presented for end user review and resolution. In one aspect, the present disclosure relates to methods and systems for automatically generating certificates of insurance documents from policy data extracted from existing policy documents in accordance with policy parameters. The policy parameters may be entered by a requesting user and/or extracted from one or more files related to a certificate of insurance request. The files, in some examples, may include one or more emails, one or more email attachments, one or more electronically submitted files, and/or one or more voicemail recordings. In one aspect, the present disclosure relates to methods and systems for populating attribute fields of a certificate of insurance template with details extracted from a collection of unstructured documents. The attributes, for example, may be organized in accordance with an insurance servicing knowledge ontology, pairing each data entry field of the certificate of insurance template with one or more corresponding attributes related to a certificate of insurance request or derived from an insurance policy. Business entity attributes, in another example, may be organized in a data store in a manner capturing relationships between entities. The methods and systems, further to the example, may identify entities related to certificate of insurance requests, such as policy holders, COI requesters, and insurance carriers, based in part on the relationship connections between entities within the data store. The data store may be organized, in some examples, as a knowledge ontology or taxonomy. In one aspect, the present disclosure relates to a system for automating creation of ad hoc certificates of insurance, the system including one or more