Automatic car insurance business handling method based on AI customer service robot

Abstract

The invention relates to the technical field of insurance business processing, and particularly discloses an automatic car insurance business handling method based on an AI customer service robot. The method comprises the following steps: receiving and standardizing business related information through an information acquisition unit, screening effective information through an information quality evaluation model, extracting key fields by means of OCR text recognition, and generating structured insurance data in combination with multi-modal consistency comparison; a vehicle information system is called to verify data integrity and accuracy in real time and automatically complement and correct errors, and compliance verification is completed through an underwriting rule engine according to a preset rule of an insurance company to generate standardized insurance data; a rate engine is called based on the data to generate a quotation scheme, the quotation scheme is submitted to an underwriting system to generate an underwriting conclusion after being confirmed by the user, a payment two-dimensional code is automatically generated after underwriting is passed, an electronic insurance policy is pushed when the user pays successfully, and meanwhile, whole-process data is stored persistently, and a service ledger is generated.

Description

Technical Field

[0001] This invention relates to the field of insurance business processing technology, and in particular to a method for automating the processing of auto insurance business based on an AI customer service robot.

[0002] Next, we will describe the specific embodiments of the claims. For each step of a claim I input, its specific embodiment will be generated. Please describe it from the perspective of an insurance company, providing a brief description of each step in one or two paragraphs. Background Technology Currently, insurance companies still largely rely on manual processes for handling auto insurance business, including data collection, information entry, insurance plan development, quotation calculation, underwriting approval, payment and policy issuance, and policy delivery. In the traditional process, users need to provide various documents such as ID card, vehicle registration certificate, driver's license, vehicle photos, new vehicle invoice, and vehicle certificate of conformity. Staff must manually verify the authenticity of these documents, including checking the license plate number, vehicle identification number (VIN), and engine number on the vehicle registration certificate, and verifying the policyholder's identity and the vehicle's past accident records through an online system. After confirming the documents are correct, the data entry personnel must enter the relevant information into the business system item by item and generate an insurance plan and quotation based on the vehicle's condition and insurance requirements. Because different insurance companies have different restrictions on insurance combinations, coverage categories, coverage ranges, and vehicle compliance, the quotation and underwriting processes often require repeated manual verification, resulting in long processing times and a high risk of omissions and errors.

[0003] In the underwriting process, routine business relies on automated system underwriting. However, many special cases (such as modified vehicles, older vehicles, and new vehicles without registration) still require manual underwriting personnel to review supplementary information and conduct risk assessments. After underwriting approval, customers need to complete the payment themselves. The system then generates an electronic policy and delivers it via SMS, email, etc., and finally, staff complete the policy archiving. The entire process involves a large number of repetitive manual operations, with a heavy workload for data entry, accuracy relying on human experience, and a disconnect between the front-end and back-end systems. Seamless integration of data verification, quotation calculation, and underwriting review is difficult, resulting in significant time consumption in overall business processing and issues such as mis-entry, omissions, and inconsistent rules. This makes it difficult to meet the current needs of online and intelligent auto insurance business.

[0004] Furthermore, the information submitted by users through channels such as WeChat Work, WeChat, and websites is complex and inconsistent in format, potentially including blurry images, irrelevant pictures, missing fields, or conflicting content. Existing systems lack the ability to automatically identify, verify, and compare image content, text content, and user input across modalities, leading to a high reliance on manual processing for data organization, rule checks, and underwriting decisions. With the increasing volume of auto insurance business and the accelerating demand for digitalization, traditional processing methods can no longer meet the requirements for high efficiency, high accuracy, and low labor costs. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide an automated method for handling auto insurance business based on AI customer service robots, in order to solve the problems of disorganized data sources, unstable image quality, low data recognition accuracy, cumbersome manual verification, difficulty in automatically comparing cross-modal data, complex underwriting rules of various insurance companies that are difficult to automate, and the fragmented business links of quotation, underwriting, payment and policy issuance that cannot be automatically completed in the existing auto insurance application process.

[0006] This invention provides a method for automating auto insurance transactions based on an AI customer service robot, comprising: Receive business-related information submitted through the information acquisition unit and perform format standardization processing; identify valid business-related information and filter out irrelevant business-related information through the information quality evaluation model. Perform OCR text recognition on the valid business-related information to extract key fields from the business-related information, and perform multimodal consistency comparison between the recognition results and user input to generate structured insurance data; The vehicle information system is invoked to perform real-time verification of the structured insurance data, identify missing, conflicting, or incorrect information, and automatically generate completion suggestions or error correction prompts to form a complete set of insurance data. The insurance application information is input into the underwriting rule engine, which automatically verifies the vehicle's compliance, document validity, and insurance restrictions according to the preset rules of different insurance companies. Data that does not comply with the rules is prompted or corrected, and standardized insurance application data is generated. Based on the standardized insurance application data, the insurance company's internal rate engine is used to generate a quote for premiums and coverage. After the user confirms the target insurance plan, an internal underwriting request is constructed and submitted to the company's underwriting system to generate an underwriting conclusion. Upon successful underwriting, a payment QR code is automatically generated. After receiving the user's payment feedback, the electronic policy is downloaded and pushed to the user. The insurance application data, quote information, underwriting results, and policy files are persistently stored, and corresponding business ledger records are generated simultaneously.

[0007] Specifically, the step of performing OCR text recognition on the valid business-related information to extract key fields from the business-related information, and performing multimodal consistency comparison between the recognition results and user input to generate structured insurance data includes: The image in the input business-related information is classified by document type, and the corresponding key field area is automatically located based on the document layout structure model. The key field area is then cropped to form the target image block. A deep learning OCR engine is used to perform text recognition on the target image block, and a language model is used to perform semantic consistency correction on the recognized content. The OCR text, image region features, and user-input text data are input into a cross-modal consistency comparison model, and the consistency score between fields is obtained through attention calculation. The fusion confidence of the fields is calculated based on OCR confidence, image clarity score, multimodal consistency score and similarity of historical insurance data, and data completion is triggered to obtain the completed fields in the case of insufficient confidence or missing fields. The completed fields, after consistency verification and completion, are organized hierarchically according to a preset field tree model to generate structured insurance data.

[0008] Specifically, the cross-modal consistency comparison model includes an image modality encoder, a text modality encoder, a cross-modal alignment module, a consistency score calculation module, and a confidence fusion module; The image modal encoder is used to convert the target image into an image feature vector; The text modal encoder is used to convert OCR text and user input text into text feature vectors respectively; The cross-modal alignment module performs semantic alignment between image feature vectors and text feature vectors based on a cross-attention mechanism; The consistency score calculation module is used to generate field-level consistency scores based on semantic alignment results; The confidence fusion module generates the final field confidence score based on the consistency score, OCR confidence score, and historical data matching degree.

[0009] Specifically, the step of calling the vehicle information system to perform real-time verification of the structured insurance data, identifying missing, conflicting, or incorrect information, and automatically generating completion suggestions or error correction prompts to form a complete set of insurance data includes: Perform integrity checks on preset key fields in structured insurance data and identify missing and unverified fields to generate a corresponding list of verification tasks; According to the verification task list, a real-time query request is initiated to the vehicle information platform to obtain vehicle registration information, technical parameters, historical records or vehicle status data. The data returned by the vehicle information platform is compared with the structured insurance data field by field semantically, and the consistency and credibility of each field are determined by similarity calculation, rule verification or logical relationship judgment. Based on the comparison results, identify field conflicts, logical anomalies, or format errors, and classify and label them according to the type or severity of the anomalies; Based on the anomaly classification results, the system automatically generates field completion suggestions or error correction prompts, performs automatic completion on data with high credibility, and integrates the completed fields with the corrected content to form a complete set of insurance application materials.

[0010] Specifically, the process of inputting the insurance application information into the underwriting rules engine, automatically verifying vehicle compliance, document validity, and insurance restrictions according to the preset rules of different insurance companies, prompting or correcting data that does not comply with the rules, and generating standardized insurance application data includes: Based on the insurance company identifier, insurance region information, and vehicle type parameters contained in the insurance application materials, the corresponding insurance company's underwriting rule set is loaded from the preset rule base, and a rule matching process applicable to the insurance application materials is constructed based on the rule set. Each rule in the underwriting rule set is logically parsed, and the parsed rule conditions are compared with the vehicle attributes, document information, insurance parameters and insured amount information in the insurance application materials at the field level to determine whether the vehicle compliance, document validity and various insurance restrictions are met. Based on the field-level comparison results, the system automatically identifies projects that violate the rules and classifies non-compliant items into uninsurable projects, projects requiring adjustment, or projects with suggestive anomalies. At the same time, it generates corresponding anomaly tags for each type of anomaly. For items in the classification results that can be automatically repaired, automatic correction is performed according to the constraints in the rule set, including adjusting the excess insured amount, replacing illegal insurance combinations, correcting document format information, or supplementing missing fields to form a version of the insurance application materials; The verified and automatically corrected insurance application information is formatted according to a preset field template to generate standardized insurance application data that meets the requirements of the insurance company's quotation interface, and the standardized data is then output to the quotation module.

[0011] Specifically, the set of underwriting rules includes: Rules for vehicle age restrictions, vehicle type insurance restrictions, vehicle technical parameters, and vehicle usage for vehicle compliance verification; Rules for verifying the validity of documents, including rules for the validity period of ID cards, rules for vehicle registration information, and rules for the integrity of document formats; Rules for combining insurance products, upper and lower limits of coverage, and deductible range used for verifying insurance product combinations and coverage limits; Risk control includes rules for incident frequency, geographical risk, and vehicle usage risk.

[0012] Specifically, the process of generating a premium and coverage quote based on the standardized insurance data using the insurance company's internal rate engine, constructing an internal underwriting request after the user confirms the target insurance plan, submitting it to the company's underwriting system, and generating an underwriting conclusion includes: Based on the vehicle information, policyholder information and insurance type configuration in the standardized insurance data, a data structure that conforms to the internal rate calculation system of this insurance company is generated, and it is constructed into an internal quotation calculation request. Input the internal quote calculation request into the insurance company's preset rate engine and coverage calculation module, and automatically generate the corresponding premium amount, coverage and insurance combination results based on the company's current rate rules, risk coefficient, coverage terms and regional applicability; The rates and coverage results are organized into one or more optional insurance plans and presented to the user; after the user selects and confirms the target plan, a corresponding internal underwriting request data packet is generated. The underwriting request data packet is used to determine the underwriting conditions: When the plan meets the company's automatic underwriting rules, the automatic underwriting process is triggered directly; When the plan involves high-risk information, special vehicles, or factors that require manual judgment, the relevant information should be submitted to the manual underwriting review module to assist underwriters in completing the review. Based on the review results of automatic or manual underwriting, generate underwriting conclusions including underwriting approval, underwriting rejection, or prompts for supplementary information, and output the approved plan to the subsequent payment and billing process.

[0013] Specifically, after underwriting approval, a payment QR code is automatically generated; after receiving user payment feedback, the electronic policy is downloaded and pushed to the user; and the insurance application data, quotation information, underwriting results, and policy files are persistently stored. Simultaneously, corresponding business ledger records are generated, including: Based on the approved insurance plan, a payment request is constructed and sent to the insurance company's payment interface to generate a payment QR code, which is then pushed to the user. Listen for payment status receipts from the payment platform and verify the payment amount, order number, and payment signature to confirm the authenticity of the payment result; After confirming successful payment, a policy download request is sent to the insurance company's underwriting system to obtain the electronic policy document, and the integrity of the electronic policy document is verified. Verified electronic policies will be pushed to users via WeChat, email, or mobile applications, while providing users with the policy number and activation information. The insurance application materials, quotation information, underwriting results, payment vouchers and electronic policy files are persistently stored, and a business ledger record containing information on the entire business process is generated.

[0014] Specifically, the information quality evaluation model includes a content type identification module, an image quality assessment module, a layout structure integrity detection module, a text validity verification module, and a multimodal consistency analysis module; Among them, the content type recognition module is used to determine whether business-related information belongs to a predefined document type; the image quality assessment module is used to generate an image quality score based on the indicators of clarity, exposure, and occlusion; and the layout structure integrity detection module is used to identify whether the key areas of the document are complete. The text validation module is used to validate the format of user-input text. The multimodal consistency analysis module is used to determine whether the image content is consistent with the user input.

[0015] Specifically, the information acquisition unit includes at least one information acquisition channel; The information acquisition channels include any one or more of WeChat, WeChat Work, mobile application clients, web pages, QR code upload portals, or third-party platform interfaces.

[0016] The automated processing method for auto insurance business based on AI customer service robots provided by this invention has the following beneficial effects: This invention achieves intelligent, standardized, and automated processing of the entire auto insurance application process by constructing a multi-channel information acquisition mechanism, an information quality evaluation model, OCR text recognition, a multimodal consistency comparison model, a structured processing module for insurance application data, an underwriting rule engine, an automated quotation and underwriting process, and an automated payment and electronic policy delivery mechanism. This solution automatically filters valid data, improves the accuracy of image and text recognition, reduces data entry errors, and automatically verifies data integrity and field conflicts, significantly reducing the workload of manual data entry and underwriting. Simultaneously, it matches underwriting conditions according to different rules, automatically generates available quotations, and triggers the underwriting and policy issuance process, improving business processing efficiency and pass rate. Furthermore, by automatically generating payment QR codes, automatically downloading and pushing electronic policies, automatically archiving business data, and generating business ledgers, this solution significantly improves the flow efficiency and data management capabilities of auto insurance business, effectively reduces labor costs, decreases error rates, and improves overall service quality and user experience. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments of the present invention will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, and these are all within the protection scope of the present invention.

[0018] Figure 1 This is a flowchart illustrating a method for automating car insurance transactions based on an AI-powered customer service robot. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. In the description of the present invention, it should be understood that the terms "center," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicate orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements, but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, the element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. Where there is no conflict, embodiments of the present invention and the various features thereof can be combined with each other, all of which are within the scope of protection of the present invention.

[0020] Example 1 Please see Figure 1 This invention provides a method for automating auto insurance transactions based on an AI customer service robot, including: Receive business-related information submitted through the information acquisition unit and perform format standardization processing; identify valid business-related information and filter out irrelevant business-related information through the information quality evaluation model. Specifically, the insurance company's internal business processing system receives business-related information submitted by users through WeChat, WeChat Work, mobile application clients, web pages, or QR code upload channels via the information acquisition unit.

[0021] The business-related information includes various forms of materials such as ID card photos, vehicle registration certificate photos, vehicle exterior photos, license plate photos, vehicle purchase invoices, vehicle qualification certificates, vehicle information manually entered by the user, and insurance requirements.

[0022] After receiving the above information, the system prioritizes the standardization of the collected images, texts and electronic documents, including orientation correction, resolution compression and color space unification for image files, and character encoding conversion, space removal and format compatibility processing for text content, in order to form a standardized data format suitable for subsequent automatic recognition and structured extraction.

[0023] After data standardization, the system invokes its internally integrated information quality evaluation model to verify the validity of submitted content. This model performs document type identification, image quality analysis, layout structure integrity detection, and text compliance verification on each piece of business information to determine whether the document is valid for processing auto insurance business. It also identifies whether images are blurry, reflective, obstructed, distorted, or have missing key areas, which could affect subsequent identification. Based on the quality score and validity label output by the model, the system determines clear, complete documents that conform to the expected format as valid business-related information, while filtering out irrelevant images, substandard images, or incomplete content.

[0024] Perform OCR text recognition on the valid business-related information to extract key fields from the business-related information, and perform multimodal consistency comparison between the recognition results and user input to generate structured insurance data; Specifically, the system first calls the corresponding layout structure model based on the document classification results to automatically locate key field areas in images such as vehicle registration certificates, ID cards, vehicle qualification certificates, and license plate photos. These areas include the license plate number, vehicle identification number (VIN), engine number, vehicle owner's name, address, usage type, registration date, curb weight, and approved number of passengers. The system then crops these located areas into independent target image blocks and inputs them into a deep learning OCR engine for character recognition. Finally, a language model is used to perform format correction and semantic consistency adjustment on the recognized text to improve the accuracy and readability of the recognition results.

[0025] After completing OCR recognition, the system further combines the text information input by the user to perform multimodal consistency comparison on data from different sources for the same field.

[0026] Specifically, the system inputs OCR text, visual features of the corresponding image region, and user-inputted text into a cross-modal consistency comparison model. It calculates field-level consistency scores through image modality coding, text modality coding, and semantic alignment mechanisms to determine the degree of consistency between data from different sources. For fields with high consistency, the system directly uses their content. For fields with conflicts or low confidence, the system performs confidence fusion based on OCR confidence, image quality scores, and the matching degree of historical insurance records to generate more reliable field values. Finally, the system hierarchically organizes the field content obtained through consistency verification and confidence fusion according to a preset field tree model, constructing standardized structured insurance data from vehicle basic information, policyholder information, document information, and insurance configuration information.

[0027] The vehicle information system is invoked to perform real-time verification of the structured insurance data, identify missing, conflicting, or incorrect information, and automatically generate completion suggestions or error correction prompts to form a complete set of insurance data. Specifically, the system first automatically identifies the fields that need to be verified based on the structured insurance data, including the vehicle identification number (VIN), license plate number, vehicle brand and model, vehicle type, owner information, registration date, usage nature and technical parameters, and generates a field verification list to determine the query type and interface to be initiated subsequently.

[0028] The system connects to the vehicle information system based on the field validation checklist and sends real-time verification requests to the corresponding interfaces according to different query requirements. For example, the system can obtain vehicle manufacturing and registration parameters through the VIN query interface, and obtain information such as the vehicle's current status, transfer records, and deregistration status through the license plate query interface. The field values ​​and status information returned by the vehicle information system are uniformly converted into a comparable data structure by the system and parsed and mapped according to internal field standards.

[0029] After receiving the data returned by the vehicle information system, the system compares the returned results with the structured insurance data field by field. Through field similarity calculation, regular expression rule verification, format consistency verification, and business logic relationship judgment, the system identifies whether any fields in the structured insurance data are missing, incorrect, or conflicting. For example, if the system finds a difference in letter positions between the VIN provided by the user and the VIN returned by the registration platform, or if the license plate color does not match the vehicle type, the system identifies it as an abnormal field; if a key field is missing in the vehicle registration information, it is determined to be a missing field.

[0030] For identified anomalies, the system automatically generates corresponding completion suggestions or correction prompts based on the anomaly type. For example, when a field is detected as missing, the system automatically completes the field value based on authoritative data returned by the vehicle information system; when fields conflict, the system recommends the optimal field value and provides correction suggestions based on the credibility of the field source, the platform's authority, and the priority of field rules; for anomalies that cannot be automatically corrected, the system generates a prompt message that requires confirmation from the user or a human underwriter.

[0031] Finally, the system reconstructs the insurance data set based on the completed and corrected field content to meet the requirements of consistency of vehicle registration information, field completeness, and accuracy of business logic, thereby generating a complete insurance data set that can be directly used for underwriting rule verification, rate calculation, and underwriting processing.

[0032] The insurance application information is input into the underwriting rule engine, which automatically verifies the vehicle's compliance, document validity, and insurance restrictions according to the preset rules of different insurance companies. Data that does not comply with the rules is prompted or corrected, and standardized insurance application data is generated. Specifically, the system first automatically identifies the fields that need to be verified based on the structured insurance data, including the vehicle identification number (VIN), license plate number, vehicle brand and model, vehicle type, owner information, registration date, usage nature and technical parameters, and generates a field verification list to determine the query type and interface to be initiated subsequently.

[0033] The system connects to the vehicle information system based on the field validation checklist and sends real-time verification requests to the corresponding interfaces according to different query requirements. For example, the system can obtain vehicle manufacturing and registration parameters through the VIN query interface, and obtain information such as the vehicle's current status, transfer records, and deregistration status through the license plate query interface. The field values ​​and status information returned by the vehicle information system are uniformly converted into a comparable data structure by the system and parsed and mapped according to internal field standards.

[0034] After obtaining the data returned by the vehicle information system, the system compares the returned results with the structured insurance data field by field. Through field similarity calculation, regular expression rule verification, format consistency verification, and business logic relationship judgment, the system identifies whether there are missing, incorrect, or conflicting fields in the structured insurance data.

[0035] For example, when the system finds that there are differences in the letter positions between the VIN provided by the user and the VIN returned by the registration platform, or that the license plate color does not match the vehicle type, the system identifies them as abnormal fields; when there are missing key fields in the vehicle registration information, they are determined to be missing fields.

[0036] For identified anomalies, the system automatically generates corresponding completion suggestions or correction prompts based on the anomaly type. For example, when a field is detected as missing, the system automatically completes the field value based on authoritative data returned by the vehicle information system; when fields conflict, the system recommends the optimal field value and provides correction suggestions based on the credibility of the field source, the platform's authority, and the priority of field rules; for anomalies that cannot be automatically corrected, the system generates a prompt message that requires confirmation from the user or a human underwriter.

[0037] Based on the standardized insurance data, the insurance company's internal rate engine is called to generate a quote for premium and coverage. After the user confirms the target insurance plan, an internal underwriting request is constructed and submitted to the company's underwriting system to generate an underwriting conclusion. Specifically, the system first constructs a quotation data package that meets the requirements of the company's internal rate engine based on the vehicle attributes, usage, insured region, insurance type combination, and coverage amount settings contained in the standardized insurance data, and then inputs it into the rate engine for calculation. The rate engine automatically generates the premium amount, coverage, applicability of additional clauses, and related rate parameters for this insurance policy based on the company's current rate coefficients, risk algorithms, coverage rules, and regional factors.

[0038] After obtaining internal quotes, the system organizes the premiums and coverage into a presentable insurance plan and displays it to the user. When the user confirms the target plan, the system constructs an internal underwriting request data packet based on the plan and submits it to the company's internal underwriting system for automatic or manual underwriting assisted review. The automatic underwriting module conducts a multi-dimensional review of the vehicle condition, policyholder information, vehicle purpose, past accident records, and the risk matching of the insurance plan based on the company's preset underwriting rules, risk assessment models, and historical underwriting strategies. If all conditions meet the automatic underwriting rules, the system directly generates an underwriting approval result.

[0039] For insurance applications that do not meet the conditions for automatic underwriting, such as cases where the vehicle has undergone special modifications, the vehicle information has historical anomalies, the risk level is high, or some fields require manual judgment, the system will submit the relevant insurance application materials and risk warnings to the manual underwriting auxiliary review module. Manual underwriters will view basic information, verification information, consistency comparison results, and risk assessment prompts through the system interface, complete the necessary manual review operations, and submit the underwriting conclusion through the system.

[0040] After underwriting approval, a payment QR code is automatically generated. After receiving payment feedback from the user, the electronic policy is downloaded and pushed to the user. The insurance data, quotation information, underwriting results and policy files are persistently stored, and corresponding business ledger records are generated at the same time. Specifically, the system generates a payment request data packet based on the approved insurance plan and calls the insurance company's internal payment interface or cashier service to generate a corresponding payment order number. The system automatically generates a payment QR code based on the order number and the premium amount. This QR code can be displayed to the user through mobile applications, web pages, or WeChat Work, etc., to complete the online payment of the insurance premium.

[0041] After the QR code is generated, the system activates the payment status monitoring module to obtain the payment progress in real time through the payment platform's callback interface or polling mechanism. Upon receiving a payment success receipt, the system verifies key information such as the payment order number, transaction amount, and payment signature to ensure the authenticity and consistency of the payment result. If the payment result verification is successful, the system will automatically call the electronic policy generation and download interface of the insurance company's underwriting system to request the generation of an electronic policy based on the application number, underwriting number, and order number, and obtain the electronic policy file containing the policy text, liability terms, and payment information.

[0042] After successfully downloading the electronic policy document, the system performs integrity verification on the file, including digital signature verification, file format parsing, and file hash verification, to ensure the policy's origin is trustworthy and its content has not been tampered with. Upon completion of verification, the system sends the electronic policy to the user via WeChat, mobile application push notifications, or email, and marks the delivery status and user confirmation information in the push notification record. Simultaneously, the system synchronizes the policy number, effective date, coverage scope, and payment status to the user's interface, enabling the user to immediately access effective coverage information.

[0043] After data delivery, the system persistently stores all data involved in this business process, including standardized insurance application data, quotation information, underwriting records, payment vouchers, electronic policy documents, and user notification receipts, and writes this information into the company's internal business ledger management module. The system automatically generates business ledger records containing business identifiers, timestamps, underwriting types, payment status, policy numbers, and processing logs to facilitate subsequent operational analysis, risk auditing, regulatory reporting, and financial reconciliation, achieving automated closed-loop management of the entire underwriting process.

[0044] Furthermore, the step of performing OCR text recognition on the valid business-related information to extract key fields from the business-related information, and performing multimodal consistency comparison between the recognition results and user input to generate structured insurance data includes: The image in the input business-related information is classified by document type, and the corresponding key field area is automatically located based on the document layout structure model. The key field area is then cropped to form the target image block. Specifically, the insurance company's internal intelligent business processing system calls a deep learning OCR engine to perform text recognition on the target image blocks cropped in the aforementioned steps. The OCR engine, based on deep learning models such as Convolutional Neural Networks (CNN), Long Short-Term Memory Networks (LSTM), and Attention mechanisms, performs end-to-end character recognition in the image blocks, automatically outputting text sequences including letters, numbers, and Chinese characters, and generating corresponding character-level confidence scores. This process is robust to complex fonts, different shooting angles, lighting differences, and slightly blurred images. After completing the initial OCR recognition, the system further combines a language model to perform semantic consistency correction on the recognized content. The language model can be pre-trained based on a large-scale auto insurance industry text corpus, capable of capturing contextual relationships and common rules between fields. For example: the semantic and logical relationships between the length and character set constraints of VIN codes, the regional coding rules of license plate numbers, the check digit rules of ID card numbers, date formats, unit symbols, and commonly used fields.

[0045] The system performs semantic analysis and contextual verification on the OCR output text input language model, and automatically corrects possible recognition errors, missing characters, confusion, or abnormal arrangement.

[0046] For example, when "0" and "o" are confused in the VIN recognition result, the language model corrects it according to the VIN format rules and context; when the license plate number characters are not fully recognized, it is corrected by combining the regional code and common license plate patterns.

[0047] Through this step, the system can convert the OCR engine recognition results into highly reliable text output that conforms to business logic and field rules, providing accurate and reliable text information for subsequent multimodal consistency comparison, confidence fusion, and structured insurance data generation.

[0048] A deep learning OCR engine is used to perform text recognition on the target image block, and a language model is used to perform semantic consistency correction on the recognized content. Specifically, the insurance company's internal intelligent business processing system calls a deep learning OCR engine to perform text recognition on the target image blocks cropped in the aforementioned steps. The OCR engine, based on deep learning models such as Convolutional Neural Networks (CNN), Long Short-Term Memory Networks (LSTM), and Attention mechanisms, performs end-to-end character recognition in the image blocks, automatically outputting text sequences including letters, numbers, and Chinese characters, and generating corresponding character-level confidence scores. This process can robustly recognize complex fonts, images from different shooting angles, with varying lighting conditions, and slightly blurred images.

[0049] After initial OCR recognition, the system further integrates a language model to perform semantic consistency correction on the recognized content. The language model, pre-trained on a large-scale auto insurance industry text corpus, can capture contextual relationships and common rules between fields, such as: VIN length and character set constraints, regional coding rules for license plate numbers, check digit rules for ID card numbers, date formats, unit symbols, and semantic logical relationships between commonly used fields. The system inputs the OCR output text into the language model for semantic analysis and contextual verification, automatically correcting potential recognition errors, missing characters, confusion, or arrangement anomalies. For example, when "0" and "o" are confused in the VIN recognition result, the language model corrects it based on VIN format rules and context; when license plate number characters are incompletely recognized, it corrects them by combining regional coding and common license plate patterns.

[0050] Through this step, the system can convert the OCR engine recognition results into highly reliable text output that conforms to business logic and field rules, providing accurate and reliable text information for subsequent multimodal consistency comparison, confidence fusion, and structured insurance data generation.

[0051] The OCR text, image region features, and user-input text data are input into a cross-modal consistency comparison model, and the consistency score between fields is obtained through attention calculation. Specifically, within the insurance company's internal system, the text identified by the previous OCR step, the key field features of the corresponding image, and the text data manually entered by the user in the form are all uniformly input into a cross-modal consistency comparison model. This model automatically performs semantic alignment between image features and text features through an attention mechanism, and calculates the consistency score between each field.

[0052] For example, the system compares the ID number recognized by OCR with the ID number entered by the user, and also considers image clarity and regional features to determine the reliability of the field information. This process effectively identifies potential input errors or OCR recognition biases, providing a high-confidence data foundation for subsequent field completion and structured insurance data generation.

[0053] The fusion confidence of the fields is calculated based on OCR confidence, image clarity score, multimodal consistency score and similarity of historical insurance data, and data completion is triggered to obtain the completed fields in the case of insufficient confidence or missing fields. Specifically, for each key field, the system calculates the final fusion confidence score by combining the confidence score of the OCR recognition result, the clarity score of the corresponding image region, the multimodal consistency score, and the similarity with historical insurance data. This fusion confidence score comprehensively reflects the reliability of the field information. For example, if the OCR text is blurry or has recognition errors, but the image region is clear and consistent with the user input, the fusion confidence score may still be high; conversely, if multiple indicators are low, the field's credibility will significantly decrease. In this way, the system can quantify the accuracy of each field, providing a basis for judgment in subsequent automated processing. For cases with insufficient fusion confidence or missing fields, the system automatically triggers a data completion mechanism. The completion process includes retrieving similar cases from historical insurance records to recommend possible values, deriving missing information based on rules, or prompting the user for necessary supplementary information. After completion, the field is updated in the structured insurance data, ensuring the completeness and accuracy of the insurance information. This supports the smooth execution of underwriting rule verification, quotation calculation, and underwriting processes, while significantly reducing manual intervention and improving business processing efficiency.

[0054] The completed fields, after consistency verification and completion, are organized hierarchically according to a preset field tree model to generate structured insurance data. Specifically, the system organizes the fields, after multimodal consistency verification and automatic completion, hierarchically according to a pre-defined field tree model. This model categorizes key fields such as identity information, vehicle information, insurance type configuration, and coverage parameters into parent-child hierarchical relationships based on the logical structure of the insurance application data. For example, vehicle information serves as the top-level node, with child nodes including license plate number, vehicle identification number (VIN), engine number, and registration date, ensuring clear data structure and logical consistency for each field. This hierarchical organization allows the generated structured insurance data to be directly accessed by the underwriting rule engine, quotation module, and underwriting system, avoiding manual processing or repetitive calculations. In subsequent quotation, underwriting, and policy issuance processes, the system can directly access each field and its relationships, quickly performing rule verification, calculating premiums, and generating underwriting conclusions, while ensuring data integrity and consistency, improving business processing efficiency, and reducing error risk.

[0055] Furthermore, the cross-modal consistency comparison model includes an image modality encoder, a text modality encoder, a cross-modal alignment module, a consistency score calculation module, and a confidence fusion module; The image modal encoder is used to convert the target image into an image feature vector; The text modal encoder is used to convert OCR text and user input text into text feature vectors respectively; The cross-modal alignment module performs semantic alignment between image feature vectors and text feature vectors based on a cross-attention mechanism; The consistency score calculation module is used to generate field-level consistency scores based on semantic alignment results; The confidence fusion module generates the final field confidence score based on the consistency score, OCR confidence score, and historical data matching degree.

[0056] Furthermore, the step of calling the vehicle information system to perform real-time verification of the structured insurance data, identifying missing, conflicting, or incorrect information, and automatically generating completion suggestions or error correction prompts to form a complete set of insurance data includes: Perform integrity checks on preset key fields in structured insurance data and identify missing and unverified fields to generate a corresponding list of verification tasks; Specifically, the insurance company's system performs integrity checks on key fields pre-defined in the generated structured insurance application data, including information such as ID number, vehicle registration number, vehicle type, chassis number, engine number, insurance type configuration, and coverage amount. The system automatically identifies missing fields or fields requiring further verification, such as incomplete vehicle registration information or abnormal insurance type configurations, and compiles these fields into a verification task list, providing clear task guidance for subsequent comparisons with the vehicle information system. By generating the verification task list, the system can manage the fields to be verified in an orderly manner, ensuring that every key data point is checked and verified, avoiding omissions or errors. This list serves as the basis for calling the vehicle information system for real-time queries and comparisons, making the automatic verification process more efficient and traceable, while also providing a basis for subsequently automatically generating completion suggestions or error correction prompts, thereby improving the completeness and reliability of the insurance application data.

[0057] According to the verification task list, a real-time query request is initiated to the vehicle information platform to obtain vehicle registration information, technical parameters, historical records or vehicle status data. Specifically, the system initiates a real-time query request to the vehicle information platform for the verification task list generated in the previous step, in order to obtain the relevant vehicle registration information, technical parameters, historical accident records, and current vehicle status data. The query process can automatically match the VIN, license plate number, or engine number in the insurance application materials to ensure the accuracy of the request and return detailed vehicle information in a short time. The obtained data is then compared field-by-field with the structured insurance application data to determine the completeness and accuracy of the insurance application materials. For example, the system can verify whether the vehicle registration date and vehicle model match the insurance information, and whether there are any unrecorded accidents or violations. This real-time query mechanism significantly improves data verification efficiency, reduces manual verification work, and provides a reliable basis for subsequent anomaly identification and automatic completion.

[0058] The data returned by the vehicle information platform is compared with the structured insurance data field by field semantically, and the consistency and credibility of each field are determined by similarity calculation, rule verification or logical relationship judgment. Specifically, the system performs a field-by-field semantic comparison between the data returned by the vehicle information platform and the structured insurance application data. The system performs similarity calculations for each field, such as comparing the consistency of key fields like the vehicle identification number (VIN), engine number, and license plate number. Simultaneously, it performs logical checks based on preset rules, such as ensuring the vehicle registration date is not later than the insurance application date and that the vehicle model falls within the company's coverage scope. Through these comparisons and rule checks, the system can determine the consistency and reliability of each field. Fields that match perfectly are marked as high reliability, while fields with discrepancies or anomalies are recorded as low reliability or anomalies. This process automatically detects information conflicts, format errors, or logical anomalies, providing accurate evidence for subsequent field completion, error correction prompts, and the generation of complete insurance application data, thereby improving the accuracy and reliability of the insurance application data.

[0059] Based on the comparison results, identify field conflicts, logical anomalies, or format errors, and classify and label them according to the type or severity of the anomalies; Specifically, the system automatically identifies anomalies in each field based on the comparison results between the data returned by the vehicle information platform and the structured insurance application data. These anomalies include field conflicts, logical inconsistencies, and formatting errors. For example, if the vehicle registration information does not match the vehicle identification number (VIN) entered by the user, or if the coverage amount exceeds the insurance type limit, the system will mark that field as an anomaly. Identified anomaly fields are categorized according to their type and severity: conflicting information, logical inconsistencies, and formatting issues are grouped separately. High-severity anomalies (such as mismatched key document information) are prioritized, while medium- and low-severity anomalies are addressed with prompts or subsequent corrections based on established rules. This categorization and marking provides a clear basis for subsequent automatic completion, error correction prompts, and the generation of complete insurance application data, improving the accuracy and efficiency of insurance data processing.

[0060] Based on the anomaly classification results, automatic field completion suggestions or error correction prompts are generated. Automatic completion is performed on data with high credibility, and the completed fields are integrated with the correction content to form a complete set of insurance information. Specifically, based on the results of the previous anomaly classification step, the system automatically generates completion suggestions or error correction prompts for each anomaly field. For example, for a missing license plate number field, the system can infer possible values ​​based on historical insurance data or vehicle registration information; for incorrectly formatted document numbers, the system will provide automatic correction suggestions according to the standard format. The system will also determine which data is highly reliable based on the field's fusion confidence level, and automatically perform completion operations for these fields without manual intervention. After completion and error correction, the system integrates the updated fields back into the structured insurance data, forming a complete, accurate, and business-compliant insurance data set. This data set can be directly used for underwriting rule verification, quotation calculation, and underwriting processes, ensuring data integrity and consistency while significantly reducing manual data entry and verification work, improving the efficiency and accuracy of automated processing in auto insurance business.

[0061] Furthermore, the step of inputting the insurance application information into the underwriting rule engine, automatically verifying the vehicle's compliance, document validity, and insurance restrictions according to the preset rules of different insurance companies, and prompting or correcting data that does not comply with the rules and generating standardized insurance application data includes: Based on the insurance company identifier, insurance region information, and vehicle type parameters contained in the insurance application materials, the corresponding insurance company's underwriting rule set is loaded from the preset rule base, and a rule matching process applicable to the insurance application materials is constructed based on the rule set. Specifically, the system first retrieves key information from the insurance application materials, such as the insurance company's identification, the region of coverage, and vehicle type parameters. Based on this information, the system loads the corresponding insurance company's underwriting rule set from a pre-set rule base. These rule sets include aspects such as vehicle compliance, document validity, insurance combination restrictions, coverage limits, and risk control requirements. Subsequently, the system constructs a dedicated rule matching process for the current insurance application materials based on the loaded rule set. This process compares each field in the insurance application materials with the rule conditions one by one to automatically determine whether the vehicle, documents, and insurance types meet the insurance company's underwriting requirements. In this way, the system can quickly identify data that does not comply with the rules in an automated process, providing a standardized and callable data foundation for subsequent automatic correction, quotation calculation, and underwriting processes.

[0062] Each rule in the underwriting rule set is logically parsed, and the parsed rule conditions are compared with the vehicle attributes, document information, insurance parameters and insured amount information in the insurance application materials at the field level to determine whether the vehicle compliance, document validity and various insurance restrictions are met. Specifically, the system's underwriting rule engine logically parses each rule in the loaded rule set, transforming it into a computable conditional expression. For example, the vehicle age restriction rule is parsed as "the difference between the vehicle registration date and the current date must not exceed X years," the insurance combination rule is parsed as "insurance A and insurance B cannot be selected simultaneously," and the document validity period rule is parsed as "the validity period of the ID card must cover the policy start date." The system then compares these parsed rule conditions with each field in the application materials, including vehicle attributes, document information, insurance parameters, and coverage information. Through field-level comparison, the system can determine whether each piece of application information meets the requirements for vehicle compliance, document validity, and other underwriting restrictions. This provides a basis for automatically identifying non-compliant items, generating anomaly markers, or performing subsequent corrections, ensuring that the application data complies with the insurance company's underwriting standards.

[0063] Based on the field-level comparison results, the system automatically identifies projects that violate the rules and classifies non-compliant items into uninsurable projects, projects requiring adjustment, or projects with suggestive anomalies. At the same time, it generates corresponding anomaly tags for each type of anomaly. Specifically, based on the results of the previous field-level comparison, the system will automatically identify fields and items that violate underwriting rules. For example, if the vehicle's age exceeds the company's maximum insurable years, this field will be marked as abnormal; if the insurance combination does not comply with the rules, the system will mark the relevant insurance fields as non-compliant items. Identified non-compliant items will be categorized into three types according to processing priority and severity: Uninsurable items (such as invalid key documents or vehicles that are prohibited from being insured) will directly block the insurance application process; Adjustments to items (such as exceeding the coverage limit or inappropriate insurance combination) are permitted to be made by the system or the user. For items with suggestive exceptions (such as incorrect formatting of non-critical fields), only a notification is provided.

[0064] Each type of anomaly will generate a corresponding anomaly marker for subsequent automatic correction, manual review, or system prompts, thereby ensuring that the insurance application materials meet the underwriting requirements and are easy to process.

[0065] For items in the classification results that can be automatically repaired, automatic correction is performed according to the constraints in the rule set, including adjusting the excess insured amount, replacing illegal insurance combinations, correcting document format information, or supplementing missing fields to form a version of the insurance application materials; Specifically, for any abnormal items identified as automatically repairable in the previous classification results, the system will perform automatic corrections based on the constraints in the underwriting rule set. For example, when the insured amount exceeds the prescribed limit, the system will automatically adjust the amount to the allowable range; for insurance combinations that do not comply with the rules, the system can replace or delete illegal insurance products to meet underwriting requirements; if there are errors in the document format, the system will automatically correct it to the standard format; for missing key fields, the system can derive appropriate values ​​and complete them based on historical records or rules. Through these automatic correction operations, the version of the insurance application materials generated by the system has been corrected according to the rules and meets the underwriting standards and business process requirements of the insurance company. The corrected materials can be directly used by the quotation, underwriting, and policy issuance modules, ensuring that the insurance information is complete, standardized, and can be used for subsequent automated processing, while reducing the risk of manual intervention and errors.

[0066] The verified and automatically corrected insurance application data is formatted according to a preset field template to generate standardized insurance application data that meets the requirements of the insurance company's quotation interface, and the standardized data is output to the quotation module. Specifically, the insurance application materials, after being verified and automatically corrected according to underwriting rules, are formatted according to a preset field template. This template arranges key data such as identity information, vehicle information, insurance combinations, and coverage parameters in the application materials according to standardized field names and data types, ensuring complete matching with the data format of the insurance company's internal quotation interface. During the formatting process, the system also unifies data types, units, and date formats to ensure the standardized and complete structure of the insurance application data. After formatting, the generated standardized insurance application data is output to the quotation module for subsequent premium calculation and coverage generation. Since the data strictly conforms to the quotation interface requirements, the quotation module can directly call it without additional manual adjustments or repeated verification, thereby automating and streamlining the quotation process and ensuring the accuracy and consistency of the insurance application data.

[0067] Furthermore, the set of underwriting rules includes: Rules for vehicle age restrictions, vehicle type insurance restrictions, vehicle technical parameters, and vehicle usage for vehicle compliance verification; Rules for verifying the validity of documents, including rules for the validity period of ID cards, rules for vehicle registration information, and rules for the integrity of document formats; Rules for combining insurance products, upper and lower limits of coverage, and deductible range used for verifying insurance product combinations and coverage limits; Risk control includes rules for incident frequency, geographical risk, and vehicle usage risk.

[0068] Furthermore, the step of sending a quote request to one or more insurance companies based on the standardized insurance data, obtaining the corresponding rates and coverage, and performing automatic or manual underwriting assisted review on the user-confirmed plan to generate an underwriting conclusion includes: Based on the vehicle information, policyholder information and insurance type configuration in the standardized insurance data, a data structure that conforms to the internal rate calculation system of this insurance company is generated, and it is constructed into an internal quotation calculation request. Specifically, the system organizes vehicle information, policyholder information, and insurance type configurations from standardized insurance application data into a data structure that conforms to the company's internal rate calculation system. For example, the system maps information such as vehicle brand, model, vehicle age, purpose, insurance type, and coverage amount to internal rate calculation fields and generates calculation parameters according to the company's rate rules. Subsequently, the system constructs these parameters into an internal quotation calculation request, which serves as input to the internal rate engine. Based on this request, the internal quotation engine automatically calculates the corresponding premium amount, coverage scope, and applicable insurance type combinations, and generates one or more insurance plans for the user to choose from, providing a standardized data foundation for user confirmation and subsequent underwriting processes.

[0069] Input the internal quote calculation request into the insurance company's preset rate engine and coverage calculation module, and automatically generate the corresponding premium amount, coverage and insurance combination results based on the company's current rate rules, risk coefficient, coverage terms and regional applicability; Specifically, the generated internal quote calculation request is input into the company's pre-set rate engine and coverage calculation module. The rate engine automatically calculates the premium for each type of insurance based on the company's current rate rules, vehicle type, policyholder information, historical risk data, and regional risk coefficients, while also considering coverage terms, deductibles, and regional applicability. The coverage calculation module automatically generates a complete coverage combination and insurance liability list based on the selected insurance types and coverage amounts in the insurance plan. Ultimately, the system output includes the premium, coverage, and combination options for each type of insurance for user selection, while providing standardized, directly accessible data for subsequent underwriting processes, achieving automation and accuracy in quote calculation.

[0070] The rates and coverage results are organized into one or more optional insurance plans and presented to the user; after the user selects and confirms the target plan, a corresponding internal underwriting request data packet is generated. Specifically, the system organizes vehicle information, policyholder information, and insurance type configurations from standardized insurance application data into a data structure that conforms to the company's internal rate calculation system. For example, the system maps information such as vehicle brand, model, vehicle age, purpose, insurance type, and coverage amount to internal rate calculation fields and generates calculation parameters according to the company's rate rules. Subsequently, the system constructs these parameters into an internal quotation calculation request, which serves as input to the internal rate engine. Based on this request, the internal quotation engine automatically calculates the corresponding premium amount, coverage scope, and applicable insurance type combinations, and generates one or more insurance plans for the user to choose from, providing a standardized data foundation for user confirmation and subsequent underwriting processes.

[0071] The underwriting request data packet is used to determine the underwriting conditions: When the plan meets the company's automatic underwriting rules, the automatic underwriting process is triggered directly; When the plan involves high-risk information, special vehicles, or factors that require manual judgment, the relevant information should be submitted to the manual underwriting review module to assist underwriters in completing the review. Specifically, the system will assess the underwriting conditions of the generated underwriting request data packet. If the insurance plan meets the company's automatic underwriting rules, such as complete vehicle information, legal insurance combination, and acceptable historical claim frequency, the system will directly trigger the automatic underwriting process, and the underwriting engine will generate an underwriting conclusion without manual intervention. If the insurance plan involves high-risk information, special vehicles (such as modified vehicles or old vehicles), or factors requiring manual judgment, the system will submit the relevant information to the manual underwriting review module. Underwriters can complete the review based on the auxiliary information and comparison results provided by the system, including verifying the authenticity of the information, risk assessment, and the applicability of special clauses, thereby ensuring that high-risk insurance plans are fully reviewed and a reasonable underwriting decision is made.

[0072] Based on the review results of automatic or manual underwriting, generate underwriting conclusions including underwriting approval, underwriting rejection, or prompts for supplementary information, and output the approved plan to the subsequent payment and billing process; Specifically, the system generates an underwriting conclusion based on the review results of automated or manual underwriting. If the insurance plan complies with the underwriting rules and the risk is acceptable, the system will generate an "underwriting approved" conclusion; if there are serious violations or high-risk situations, the system will generate an "underwriting rejected" conclusion; for cases with incomplete information or questions, the system will generate a "suggestion for supplementary information" to guide the user to supplement or correct relevant information. For approved plans, the system will output the insurance application materials, underwriting results, and related identifiers to the payment and policy issuance module. The payment module then generates a payment request and QR code. After the user completes the payment, the system will automatically trigger the electronic policy issuance and push process, achieving a seamless connection from underwriting to policy generation, while ensuring the integrity and traceability of business data.

[0073] Furthermore, after underwriting approval, a payment QR code is automatically generated. Upon receiving user payment feedback, the electronic policy is downloaded and pushed to the user. The insurance application data, quotation information, underwriting results, and policy files are persistently stored, and corresponding business ledger records are generated, including: Based on the approved insurance plan, a payment request is constructed and sent to the insurance company's payment interface to generate a payment QR code, which is then pushed to the user. Specifically, once underwriting is approved, the system will construct a payment request based on the premium amount, policyholder information, and payment channel in the insurance plan and send it to the company's payment interface. The payment interface will return a unique payment QR code, which the system will then push to the user, displaying it through WeChat, mobile applications, or web pages for convenient scanning to complete the payment. During the push process, the system will ensure that the QR code corresponds one-to-one with the insurance plan, policy number, and order information to avoid confusion or duplicate generation of payment information. After the user scans the code to pay, the system can monitor the payment status in real time and obtain payment feedback, providing accurate payment confirmation information for subsequent electronic policy download and issuance processes.

[0074] Listen for payment status receipts from the payment platform and verify the payment amount, order number, and payment signature to confirm the authenticity of the payment result; Specifically, the system continuously monitors payment status receipts returned by the payment platform, including information such as payment amount, order number, payment time, and digital signature. The system compares the received payment information with the premium amount and order number in the approved insurance plan, and verifies the legality of the payment signature to ensure the receipt information has not been tampered with. Through this verification process, the system can confirm the authenticity and integrity of the user's payment, preventing policy issuance failure due to payment anomalies or incorrect information. Only after the payment information passes verification will the system proceed with the electronic policy generation and push process, thus ensuring the security and reliability of the payment and policy issuance stages.

[0075] After confirming successful payment, a policy download request is sent to the insurance company's underwriting system to obtain the electronic policy document, and the integrity of the electronic policy document is verified. Specifically, once the payment status is successfully confirmed, the system sends a policy download request to the underwriting system to obtain the corresponding electronic policy document. The system matches the request with the policy number, policyholder information, and order number of the insurance plan to ensure that the obtained policy is consistent with the payment record and underwriting results. After obtaining the electronic policy document, the system performs integrity verification, including checking the file format, digital signature, and policy number consistency to ensure that the electronic policy has not been tampered with or lost. Only policy documents that pass verification will be included in the subsequent push and archiving process, ensuring that the policy information obtained by the user is accurate, complete, and traceable.

[0076] Verified electronic policies will be pushed to users via WeChat, email, or mobile applications, while providing users with the policy number and activation information. Specifically, verified electronic policies are pushed to users via WeChat, email, or mobile applications. During the push process, the system sends the policy document along with the policy number, policyholder information, and policy effective date, ensuring users can clearly access policy details and status. Simultaneously, the system provides confirmation information, such as the policy number, insurance start and end dates, and coverage details, facilitating user verification of insurance information. This method allows users to promptly obtain valid electronic policies, improving information transparency and service experience, while also providing reliable documentation for subsequent claims or business inquiries.

[0077] The insurance application materials, quotation information, underwriting results, payment vouchers and electronic policy files are persistently stored, and a business ledger record containing information on the entire business process is generated.

[0078] Specifically, the system will persistently store application materials, quotation information, underwriting results, payment vouchers, and electronic policy documents to ensure the integrity, security, and traceability of all business data. Data storage can utilize a structured database or file storage system, and supports version management and access control to meet compliance and auditing requirements.

[0079] Simultaneously, the system generates a business ledger containing information on the entire business process, detailing the operations and status of each step from insurance application, document verification, quotation calculation, underwriting results, payment completion to electronic policy delivery. This business ledger not only facilitates internal management and risk control but also provides complete and reliable data for user inquiries, claims verification, and regulatory review.

[0080] Furthermore, the information quality evaluation model includes a content type identification module, an image quality assessment module, a layout structure integrity detection module, a text validity verification module, and a multimodal consistency analysis module; Among them, the content type recognition module is used to determine whether business-related information belongs to a predefined document type; the image quality assessment module is used to generate an image quality score based on the indicators of clarity, exposure, and occlusion; and the layout structure integrity detection module is used to identify whether the key areas of the document are complete. The text validation module is used to validate the format of user-input text. The multimodal consistency analysis module is used to determine whether the image content is consistent with the user input.

[0081] Furthermore, the information acquisition unit includes at least one information acquisition channel; The information acquisition channels include any one or more of WeChat, WeChat Work, mobile application clients, web pages, QR code upload portals, or third-party platform interfaces.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for automating auto insurance business processing based on AI customer service robots, characterized in that, include: Receive business-related information submitted through the information acquisition unit and perform format standardization processing; identify valid business-related information and filter out irrelevant business-related information through the information quality evaluation model. Perform OCR text recognition on the valid business-related information to extract key fields from the business-related information, and perform multimodal consistency comparison between the recognition results and user input to generate structured insurance data; The vehicle information system is invoked to perform real-time verification of the structured insurance data, identify missing, conflicting, or incorrect information, and automatically generate completion suggestions or error correction prompts to form a complete set of insurance data. The insurance application information is input into the underwriting rule engine, which automatically verifies the vehicle's compliance, document validity, and insurance restrictions according to the preset rules of different insurance companies. Data that does not comply with the rules is prompted or corrected, and standardized insurance application data is generated. Based on the standardized insurance data, the insurance company's internal rate engine is called to generate a quote for premium and coverage. After the user confirms the target insurance plan, an internal underwriting request is constructed and submitted to the company's underwriting system to generate an underwriting conclusion. After underwriting approval, a payment QR code is automatically generated. After receiving payment feedback from the user, the electronic policy is downloaded and pushed to the user. The insurance application data, quotation information, underwriting results and policy files are persistently stored, and corresponding business ledger records are generated at the same time.

2. The method for automating auto insurance business based on AI customer service robots according to claim 1, characterized in that, The step of performing OCR text recognition on the valid business-related information to extract key fields from the business-related information, and performing multimodal consistency comparison between the recognition results and user input to generate structured insurance data includes: The image in the input business-related information is classified by document type, and the corresponding key field area is automatically located based on the document layout structure model. The key field area is then cropped to form the target image block. A deep learning OCR engine is used to perform text recognition on the target image block, and a language model is used to perform semantic consistency correction on the recognized content. The OCR text, image region features, and user-input text data are input into a cross-modal consistency comparison model, and the consistency score between fields is obtained through attention calculation. The fusion confidence of the fields is calculated based on OCR confidence, image clarity score, multimodal consistency score and similarity of historical insurance data, and data completion is triggered to obtain the completed fields in the case of insufficient confidence or missing fields. The completed fields, after consistency verification and completion, are organized hierarchically according to a preset field tree model to generate structured insurance data.

3. The method for automating auto insurance business based on AI customer service robots according to claim 2, characterized in that, The cross-modal consistency comparison model includes an image modal encoder, a text modal encoder, a cross-modal alignment module, a consistency score calculation module, and a confidence fusion module; The image modal encoder is used to convert the target image into an image feature vector; The text modal encoder is used to convert OCR text and user input text into text feature vectors respectively; The cross-modal alignment module performs semantic alignment between image feature vectors and text feature vectors based on a cross-attention mechanism; The consistency score calculation module is used to generate field-level consistency scores based on semantic alignment results; The confidence fusion module generates the final field confidence score based on the consistency score, OCR confidence score, and historical data matching degree.

4. The method for automating auto insurance business based on AI customer service robots according to claim 1, characterized in that, The process of calling the vehicle information system to perform real-time verification of the structured insurance data, identifying missing, conflicting, or incorrect information, and automatically generating completion suggestions or error correction prompts to form a complete set of insurance data includes: Perform integrity checks on preset key fields in structured insurance data and identify missing and unverified fields to generate a corresponding list of verification tasks; According to the verification task list, a real-time query request is initiated to the vehicle information platform to obtain vehicle registration information, technical parameters, historical records or vehicle status data. The data returned by the vehicle information platform is compared with the structured insurance data field by field semantically, and the consistency and credibility of each field are determined by similarity calculation, rule verification or logical relationship judgment. Based on the comparison results, identify field conflicts, logical anomalies, or format errors, and classify and label them according to the type or severity of the anomalies; Based on the anomaly classification results, the system automatically generates field completion suggestions or error correction prompts, performs automatic completion on data with high credibility, and integrates the completed fields with the corrected content to form a complete set of insurance application materials.

5. The method for automating auto insurance business processing based on AI customer service robots according to claim 1, characterized in that, The process of inputting the insurance application information into the underwriting rule engine, automatically verifying vehicle compliance, document validity, and insurance restrictions according to the preset rules of different insurance companies, prompting or correcting data that does not comply with the rules, and generating standardized insurance application data includes: Based on the insurance company identifier, insurance region information, and vehicle type parameters contained in the insurance application materials, the corresponding insurance company's underwriting rule set is loaded from the preset rule base, and a rule matching process applicable to the insurance application materials is constructed based on the rule set. Each rule in the underwriting rule set is logically parsed, and the parsed rule conditions are compared with the vehicle attributes, document information, insurance parameters and insured amount information in the insurance application materials at the field level to determine whether the vehicle compliance, document validity and various insurance restrictions are met. Based on the field-level comparison results, the system automatically identifies projects that violate the rules and classifies non-compliant items into uninsurable projects, projects requiring adjustment, or projects with suggestive anomalies. At the same time, it generates corresponding anomaly tags for each type of anomaly. For items in the classification results that can be automatically repaired, automatic correction is performed according to the constraints in the rule set, including adjusting the excess insured amount, replacing illegal insurance combinations, correcting document format information, or supplementing missing fields to form a version of the insurance application materials; The verified and automatically corrected insurance application information is formatted according to a preset field template to generate standardized insurance application data that meets the requirements of the insurance company's quotation interface, and the standardized data is then output to the quotation module.

6. The method for automating auto insurance business based on AI customer service robots according to claim 5, characterized in that, The underwriting rule set includes: Rules for vehicle age restrictions, vehicle type insurance restrictions, vehicle technical parameters, and vehicle usage for vehicle compliance verification; Rules for verifying the validity of documents, including rules for the validity period of ID cards, rules for vehicle registration information, and rules for the integrity of document formats; Rules for combining insurance products, upper and lower limits of coverage, and deductible range used for verifying insurance product combinations and coverage limits; Risk control includes rules for incident frequency, geographical risk, and vehicle usage risk.

7. The method for automating auto insurance business processing based on AI customer service robots according to claim 1, characterized in that, The process of generating a premium and coverage quote based on the standardized insurance data using the insurance company's internal rate engine, and then, after the user confirms the target insurance plan, constructing an internal underwriting request and submitting it to the company's underwriting system to generate an underwriting conclusion, includes: Based on the vehicle information, policyholder information and insurance type configuration in the standardized insurance data, a data structure that conforms to the internal rate calculation system of this insurance company is generated, and it is constructed into an internal quotation calculation request. Input the internal quote calculation request into the insurance company's preset rate engine and coverage calculation module, and automatically generate the corresponding premium amount, coverage and insurance combination results based on the company's current rate rules, risk coefficient, coverage terms and regional applicability; The rates and coverage results are organized into one or more optional insurance plans and presented to the user; after the user selects and confirms the target plan, a corresponding internal underwriting request data packet is generated. The underwriting request data packet is used to determine the underwriting conditions: When the plan meets the company's automatic underwriting rules, the automatic underwriting process is triggered directly; When the plan involves high-risk information, special vehicles, or factors that require manual judgment, the relevant information should be submitted to the manual underwriting review module to assist underwriters in completing the review. Based on the review results of automatic or manual underwriting, generate underwriting conclusions including underwriting approval, underwriting rejection, or prompts for supplementary information, and output the approved plan to the subsequent payment and billing process.

8. The method for automating auto insurance business based on AI customer service robots according to claim 1, characterized in that, After underwriting approval, a payment QR code is automatically generated. Upon receiving user payment feedback, the electronic policy is downloaded and pushed to the user. The insurance application data, quotation information, underwriting results, and policy files are persistently stored, and corresponding business ledger records are generated, including: Based on the approved insurance plan, a payment request is constructed and sent to the insurance company's payment interface to generate a payment QR code, which is then pushed to the user. Listen for payment status receipts from the payment platform and verify the payment amount, order number, and payment signature to confirm the authenticity of the payment result; After confirming successful payment, a policy download request is sent to the insurance company's underwriting system to obtain the electronic policy document, and the integrity of the electronic policy document is verified. Verified electronic policies will be pushed to users via WeChat, email, or mobile applications, while providing users with the policy number and activation information. The insurance application materials, quotation information, underwriting results, payment vouchers and electronic policy files are persistently stored, and a business ledger record containing information on the entire business process is generated.

9. The method for automating auto insurance business based on AI customer service robots according to claim 1, characterized in that, The information quality evaluation model includes a content type identification module, an image quality assessment module, a layout structure integrity detection module, a text validity verification module, and a multimodal consistency analysis module. Among them, the content type recognition module is used to determine whether business-related information belongs to a predefined document type; the image quality assessment module is used to generate an image quality score based on the indicators of clarity, exposure, and occlusion; and the layout structure integrity detection module is used to identify whether the key areas of the document are complete. The text validation module is used to validate the format of user-input text. The multimodal consistency analysis module is used to determine whether the image content is consistent with the user input.

10. The method for automating auto insurance business based on AI customer service robots according to claim 1, characterized in that, The information acquisition unit includes at least one information acquisition channel; The information acquisition channels include any one or more of WeChat, WeChat Work, mobile application clients, web pages, QR code upload portals, or third-party platform interfaces.

Images