Method and system for financial intelligence auditing

Abstract

The application discloses a financial intelligent auditing method and system, and belongs to the technical field of auditing. The method and system realize fast auditing of regular compliant bills through AI automatic auditing, do not need manual intervention, greatly shorten the auditing period, and help employees to complete pre-application, trip construction, bill submission and other operations through an AI reimbursement assistant, so that the preparation time of the employees and the basic auditing workload of financial personnel are reduced. Through combination of preset auditing rules and an AI auditing model, various defect problems (such as repeated reimbursement, trip conflict, inconsistency between expenses and check-in places and the like) in the business travel reimbursement are accurately identified, subjective errors and omissions in manual auditing are avoided, and the financial risk of an enterprise is reduced. Only abnormal bills are shunted to manual auditing, reasonable distribution of auditing resources is realized, financial personnel focus on processing of complex abnormal problems, the pertinence and efficiency of manual auditing are improved, and the AI generated preliminary auditing opinions provide a reference for the financial personnel, and the cost of manual judgment is reduced.

Description

Technical Field

[0001] This invention belongs to the field of auditing technology, specifically relating to a financial intelligent auditing method and system. Background Technology

[0002] In corporate financial management, travel expense reimbursement is a high-frequency and complex core process, and its review quality and efficiency directly impact the company's cash flow, cost control, and employee experience. Currently, traditional travel expense reimbursement review models largely rely on manual processing, which has many shortcomings that urgently need to be addressed: Firstly, the review process is inefficient. As businesses expand, the number of travel expense reports surges. Manually verifying each report requires a significant amount of time and effort from finance staff, resulting in long review cycles, affecting the timeliness of reimbursement payments, and reducing employee satisfaction.

[0003] Secondly, the accuracy of the audit is difficult to guarantee. Manual auditing is easily affected by subjective factors and fatigue, making it difficult to accurately identify hidden problems in expense reports, such as duplicate reimbursements within the same time period, conflicts between hotel check-in time and travel itinerary in other locations, discrepancies between the location of expenses and the employee's clock-in location, illegal listing of entertainment expenses during travel, and non-compliant entertainment expenses in other locations. These problems can easily lead to the loss of corporate funds and increase financial risks.

[0004] Third, inconsistent audit standards. Differences in the understanding and implementation of reimbursement policies among different finance personnel can lead to varying audit results for similar reimbursement documents, affecting the fairness and standardization of the audit process. Furthermore, employees' insufficient awareness of reimbursement policies often results in the submission of non-compliant documents due to a lack of understanding of policy requirements, further increasing the audit workload.

[0005] Fourth, the reimbursement process lacks a closed loop. The existing reimbursement process often has breakpoints in the "submission-review-payment" process. Employees cannot check the reimbursement progress and reasons for abnormalities in real time and need to communicate repeatedly with finance personnel. After problems are found during the review, the feedback and rectification process is not smooth, making it difficult to form a complete closed loop of "application-review-rectification-review-payment".

[0006] Fifth, business-related traceability is difficult. In the traditional reimbursement model, the association information between travel reimbursement documents and corresponding projects and customers is unclear, making it difficult to accurately collect project costs and analyze customer-related travel expenses, which is detrimental to corporate cost control and business decision-making.

[0007] Sixth, there is a waste of human review resources. A large number of routine and compliant expense reports occupy limited human review resources. In scenarios where human intervention is only required when anomalies occur, the traditional model fails to achieve a reasonable allocation of review resources, resulting in high labor costs.

[0008] In recent years, the application of artificial intelligence technology in various fields has gradually deepened, providing possibilities for upgrading financial auditing models. Although some existing intelligent financial auditing solutions attempt to introduce AI technology, they are mostly limited to simple document information recognition, failing to achieve deep interaction with employees and unable to meet the preliminary needs such as travel application creation, itinerary construction, and policy consultation. At the same time, they lack the accuracy to identify core defects in travel reimbursement, lack multi-dimensional data analysis support, and have not formed a complete reimbursement audit closed loop, making it difficult to fundamentally solve the drawbacks of the traditional model.

[0009] Therefore, there is an urgent need for a financial intelligent auditing method and system that integrates AI interaction, multi-dimensional analysis, and closed-loop management. This system should be designed for travel expense reimbursement scenarios, enabling intelligent processing of the entire process from pre-reimbursement application and reimbursement review to post-reimbursement inquiry. It should accurately identify various reimbursement anomalies, rationally allocate manual review resources, improve auditing efficiency and accuracy, reduce corporate financial risks, and optimize the employee reimbursement experience. Summary of the Invention This invention aims to solve the problems of low efficiency, poor accuracy, lack of process closure, and difficulty in tracing business relationships in existing travel expense reimbursement review processes. It provides an AI-based intelligent financial review method and system, which automates the entire travel expense reimbursement process, accurately identifies anomalies, and enables efficient human-machine collaboration through intelligent means. At the same time, it builds a complete reimbursement closure loop, improves review efficiency and standardization, reduces corporate financial costs and risks, and optimizes the employee reimbursement experience.

[0010] The present invention employs the following technical solution.

[0011] A financial intelligent auditing method includes: S1: Execute the pre-reimbursement interaction stage, interact with employees through the AI ​​reimbursement assistant to complete the pre-operation related to business travel; S2: During the expense reimbursement submission phase, employees submit travel expense reimbursement documents through the AI ​​expense reimbursement assistant; S3: AI-automated review stage. The financial intelligent review system performs multi-dimensional automated review of reimbursement data based on preset review rules and trained AI review models. S4: Perform audit result judgment and sorting. Based on the AI ​​automatic audit results, the expense reimbursement documents are divided into compliant documents and abnormal documents; S5: During the manual review stage, finance personnel review the abnormal documents that have been diverted, refer to the preliminary review opinions generated by AI, give the final review results, and enter the final review opinions into the system. S6: In the post-reimbursement inquiry phase, employees can use the AI ​​reimbursement assistant to check the reimbursement progress and review results; S7: In the closed-loop management phase, the financial intelligent audit system summarizes and analyzes the data from the entire process, updates the reimbursement policy knowledge base and the training data of the AI ​​audit model, and optimizes the audit rules.

[0012] Preferably, S1 specifically includes: S11: Quickly create travel requests. Employees can send travel request requests to the AI ​​expense reimbursement assistant using natural language or preset templates. The AI ​​expense reimbursement assistant extracts key information from the request and generates a travel request form. S12: Constructing travel itineraries. The AI ​​expense reimbursement assistant automatically captures or assists employees in entering transportation and accommodation information based on their travel needs, constructs a complete travel itinerary, and associates it with the corresponding project or customer information. S13: Policy consultation interaction. Employees consult the AI ​​expense reimbursement assistant about travel expense reimbursement policies through natural language. The AI ​​expense reimbursement assistant provides real-time feedback on the corresponding policy content based on a preset expense reimbursement policy knowledge base.

[0013] Preferably, S2 specifically includes: AI-powered expense reimbursement assistant helps employees verify and supplement expense documents, ensuring the completeness of expense reimbursement information, and uploads the complete expense data to the financial intelligent auditing system.

[0014] Preferably, S3 specifically includes: S31: Basic information review, verify the completeness and format of expense reimbursement documents, and confirm the authenticity and validity of the associated project / customer information; S32: Defect and problem identification, through multi-dimensional data comparison, to identify typical abnormal problems in travel expense reimbursement; S33: Expense standard review, compare the reimbursement expense details with the preset travel expense standard, and identify expenses exceeding the standard; S34: Verification of voucher authenticity. Using OCR and image recognition technologies, verify the authenticity and completeness of reimbursement vouchers and identify false or tampered vouchers.

[0015] Preferably, S32 specifically includes: a) Duplicate reimbursements within the same time period: Compare the time of expense occurrence in the reimbursement documents with the approved / pending reimbursement documents to identify duplicate reimbursements for the same expense item within the same time period; b) Hotel and off-site transportation conflicts: Compare the hotel check-in / check-out times in the travel itinerary with the departure / arrival times of off-site transportation to identify time logic conflicts; c) Inconsistency between the place of expense occurrence and the place of attendance: Obtain employee attendance data during business trips, compare the place of expense occurrence with the place of attendance, and identify irregular expense claims outside of business trips; d) Hospitality expenses incurred during travel: Compare the travel time periods of the transportation itinerary with the time when hospitality expenses were incurred to identify any unauthorized hospitality expense claims during the travel period; e) Compliance review of out-of-town hospitality expenses: Verify the approval process for out-of-town hospitality expenses, the relevance between the hospitality recipients and the reason for the trip, and identify non-compliant out-of-town hospitality expenses.

[0016] Preferably, S4 specifically includes: S41: If the expense report is normal, it is considered a compliant report and will be automatically approved and proceed to the next payment process. S42: If there are any abnormalities in the expense reimbursement documents, they will be judged as abnormal documents. The AI ​​audit model will automatically generate preliminary audit opinions and transfer the abnormal documents and preliminary audit opinions to the manual audit stage.

[0017] Preferably, the preset audit rules in S3 include basic information verification rules, expense standard rules, business logic rules, and relationship rules, which can be configured and updated by the company's financial personnel according to actual business needs; the AI ​​audit model is trained by a large amount of historical reimbursement data (including compliant documents, abnormal documents and corresponding audit opinions), has self-learning ability, and can continuously optimize the recognition accuracy based on the summary and analysis results of the whole process data.

[0018] Preferably, in S12, the AI ​​expense reimbursement assistant can automatically retrieve flight tickets, high-speed rail tickets, and hotel order information by connecting to third-party transportation and accommodation platform interfaces.

[0019] Preferably, the multi-dimensional audit in S3 also includes the matching degree analysis of travel expenses with related projects / clients. Combined with project budget data, it audits whether travel expenses are within the project budget, providing support for project cost control.

[0020] A financial intelligent auditing system for implementing the above-mentioned financial intelligent auditing method includes: AI Interaction Module: Also known as AI Expense Reimbursement Assistant, it is used to interact with employees throughout the entire process, including travel application creation, travel itinerary construction, policy consultation, guidance on submitting expense reimbursement documents, and expense reimbursement progress inquiry. It supports natural language interaction and has functions such as information extraction, data verification, and real-time feedback. Data collection and integration module: Used to collect travel-related data, including employee expense reports, scanned copies of vouchers, related project / customer information, connect to the attendance system to obtain clock-in data, connect to third-party transportation and accommodation platforms to obtain itinerary order data, connect to the project management system to obtain project budget data, and standardize and integrate the collected data; AI Automatic Review Module: Based on preset review rules and trained AI review models, it performs multi-dimensional automated review of integrated expense reimbursement data, identifies various abnormal issues, and automatically generates preliminary review opinions. Review and triage module: Based on the results of AI-automated review, compliant documents are automatically transferred to the payment process, while abnormal documents and preliminary review opinions are triaged to the manual review module; Manual review module: Used by finance personnel to review abnormal documents, enter final review comments, and support the editing, modification and feedback of review comments; The query and feedback module is used by employees to query the reimbursement progress, review results and reasons for abnormalities, receive review comments from finance personnel, and support the resubmission of rectified documents. Closed-loop management and analysis module: used to summarize the entire process of review data, conduct multi-dimensional analysis, update the reimbursement policy knowledge base and the training data of the AI ​​review model, and optimize the review rules; Data storage module: Used to store data for the entire reimbursement process, including travel application data, itinerary data, reimbursement document data, audit result data, project / customer related data, policy knowledge base data, etc., to ensure data security and traceability. The beneficial effects of the present invention are as follows: Compared with the prior art, the technical effects of the present invention include: 1. Improve review efficiency: AI-powered automatic review enables rapid review of routine and compliant documents without human intervention, significantly shortening the review cycle; at the same time, the AI ​​expense reimbursement assistant helps employees complete pre-application, itinerary construction, and document submission, reducing employee preparation time and the basic review workload of finance personnel.

[0021] 2. Improve audit accuracy: By combining preset audit rules with AI audit models, we can accurately identify various defects in travel expense reimbursements (such as duplicate reimbursements, itinerary conflicts, and discrepancies between expenses and check-in locations), avoiding subjective errors and omissions in manual audits and reducing corporate financial risks.

[0022] 3. Optimize human-machine collaboration efficiency: Only abnormal documents are diverted to manual review, achieving a reasonable allocation of review resources. This allows finance personnel to focus on handling complex and abnormal issues, improving the pertinence and efficiency of manual review. At the same time, the preliminary review opinions generated by AI provide reference for finance personnel, reducing the cost of manual judgment.

[0023] 4. Enable business-related traceability: By linking travel expense reimbursement documents with project and customer information, it is easier to collect project costs and analyze customer-related expenses, providing data support for enterprise business decisions.

[0024] 5. Build a complete reimbursement loop: Form a complete closed loop of "application-submission-review-feedback-rectification-review-payment", so that employees can check the progress and reasons for anomalies in real time. The rectification process is smooth and improves the employee reimbursement experience. At the same time, through the summary and analysis of closed-loop data, the review rules and AI models are continuously optimized to improve the system's adaptability and stability.

[0025] 6. Standardize reimbursement management: Ensure consistency in review standards through unified review rules and policy knowledge base; AI reimbursement assistant answers policy inquiries in real time, reduces the submission of non-compliant documents by employees, and improves the standardization of reimbursement management. Attached Figure Description Figure 1 This is a flowchart of the financial intelligent auditing method described in this invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, any other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0027] like Figure 1 As shown, the financial intelligent auditing method of the present invention includes: S1: Execute the pre-reimbursement interaction stage, interact with employees through the AI ​​reimbursement assistant to complete the pre-operation related to business travel; In a preferred but non-limiting embodiment of the present invention, S1 specifically includes: S11: Quickly create travel requests. Employees can initiate travel request requests to the AI ​​expense reimbursement assistant through natural language or preset templates. The AI ​​expense reimbursement assistant extracts key information from the request (including traveler, travel time, destination, and reason for travel) and generates a travel request form. S12: Constructing travel itineraries. The AI ​​expense reimbursement assistant automatically captures or assists employees in entering transportation information (including train / flight number, departure time, arrival time, and ticket price for air tickets and high-speed rail tickets) and accommodation information (including hotel name, check-in time, check-out time, and room rate) based on the travel needs provided by employees, constructs a complete travel itinerary, and associates it with the corresponding project or customer information. S13: Policy consultation interaction. Employees consult the AI ​​reimbursement assistant about travel reimbursement policies through natural language. The AI ​​reimbursement assistant provides real-time feedback on the corresponding policy content (including expense standards, reimbursement scope, and required vouchers) based on a preset reimbursement policy knowledge base.

[0028] S2: During the expense reimbursement submission phase, employees submit travel expense reimbursement documents through the AI ​​expense reimbursement assistant; In a preferred but non-limiting embodiment of the present invention, S2 specifically includes: The AI ​​expense reimbursement assistant helps employees verify and supplement document information, ensuring that the expense reimbursement documents are complete (including travel information, expense details, scanned copies of vouchers, and related project / customer information), and uploads the complete reimbursement data to the financial intelligent auditing system.

[0029] S3: AI-automated review stage. The financial intelligent review system performs multi-dimensional automated review of reimbursement data based on preset review rules and trained AI review models. In a preferred but non-limiting embodiment of the present invention, S3 specifically includes: S31: Basic information review, verify the completeness and format of expense reimbursement documents, and confirm the authenticity and validity of the associated project / customer information; S32: Defect and problem identification, through multi-dimensional data comparison, to identify typical abnormal problems in travel expense reimbursement; In a preferred but non-limiting embodiment of the present invention, S32 specifically includes: a) Duplicate reimbursements within the same time period: Compare the time of expense occurrence in the reimbursement documents with the approved / pending reimbursement documents to identify duplicate reimbursements for the same expense item within the same time period; b) Hotel and off-site transportation conflicts: Compare the hotel check-in / check-out times in the travel itinerary with the departure / arrival times of off-site transportation (airfare / high-speed rail) to identify time logic conflicts; c) Inconsistency between the place of expense occurrence and the place of attendance: Obtain employee attendance data during business trips, compare the place of expense occurrence with the place of attendance, and identify irregular expense claims outside of business trips; d) Hospitality expenses incurred during travel: Compare the travel time periods of the transportation itinerary with the time when hospitality expenses were incurred to identify any unauthorized hospitality expense claims during the travel period; e) Compliance review of out-of-town hospitality expenses: Verify the approval process for out-of-town hospitality expenses, the relevance between the hospitality recipients and the reason for the trip, and identify non-compliant out-of-town hospitality expenses.

[0030] S33: Expense standard review, compare the details of reimbursement expenses with the preset travel expense standards (including transportation class standards, accommodation expense limits, and meal allowance standards), and identify expenses exceeding the standards; S34: Verification of voucher authenticity. Using OCR and image recognition technologies, verify the authenticity and completeness of reimbursement vouchers (invoices, itinerary slips) and identify false or tampered vouchers.

[0031] S4: Perform audit result judgment and sorting. Based on the AI ​​automatic audit results, the expense reimbursement documents are divided into compliant documents and abnormal documents; In a preferred but non-limiting embodiment of the present invention, S4 specifically includes: S41: If the expense report is normal, it is considered a compliant report and will be automatically approved and proceed to the next payment process. S42: If there is an anomaly in the reimbursement document, it will be judged as an abnormal document. The AI ​​audit model will automatically generate a preliminary audit opinion (clearly specifying the type of anomaly, the cause of the anomaly, and the corresponding policy basis), and the abnormal document and the preliminary audit opinion will be diverted to the manual audit stage. S5: During the manual review stage, finance personnel review the abnormal documents that have been diverted, refer to the preliminary review opinions generated by AI, give the final review result (pass, return for rectification, or reject), and enter the final review opinion into the system. S6: In the post-reimbursement inquiry stage, employees can check the reimbursement progress and review results through the AI ​​reimbursement assistant. If the review fails, they can view the reasons for the abnormality and the review comments in real time, and resubmit the document after completing the rectification according to the review comments. S7: In the closed-loop management phase, the financial intelligent audit system summarizes and analyzes the data from the entire process, updates the reimbursement policy knowledge base and the training data of the AI ​​audit model, optimizes the audit rules, and forms a complete reimbursement closed loop of "application-submission-audit-feedback-rectification-review-payment".

[0032] In a preferred but non-limiting embodiment of the present invention, the preset audit rules in S3 include basic information verification rules, expense standard rules, business logic rules, and relationship rules, which can be configured and updated by the company's financial personnel according to actual business needs; the AI ​​audit model is trained by a large amount of historical reimbursement data (including compliant documents, abnormal documents and corresponding audit opinions), has self-learning ability, and can continuously optimize the recognition accuracy based on the summary and analysis results of the whole process data.

[0033] In a preferred but non-limiting embodiment of the present invention, in S12, the AI ​​expense reimbursement assistant can automatically retrieve air tickets, high-speed rail tickets, and hotel order information by connecting to third-party transportation and accommodation platform interfaces, eliminating the need for employees to manually enter the information and improving the efficiency and accuracy of itinerary construction.

[0034] In a preferred but non-limiting embodiment of the present invention, the multi-dimensional review in S3 also includes a matching analysis of travel expenses with related projects / clients, and, in conjunction with project budget data, reviews whether travel expenses are within the project budget, thus providing support for project cost control.

[0035] The present invention also provides a financial intelligent auditing system for implementing the above-mentioned financial intelligent auditing method, comprising: AI Interaction Module: Also known as AI Expense Reimbursement Assistant, it is used to interact with employees throughout the entire process, including travel application creation, travel itinerary construction, policy consultation, guidance on submitting expense reimbursement documents, and expense reimbursement progress inquiry. It supports natural language interaction and has functions such as information extraction, data verification, and real-time feedback. Data collection and integration module: Used to collect travel-related data, including employee expense reports, scanned copies of vouchers, related project / customer information, connect to the attendance system to obtain clock-in data, connect to third-party transportation and accommodation platforms to obtain itinerary order data, connect to the project management system to obtain project budget data, and standardize and integrate the collected data; AI Automatic Review Module: Based on preset review rules and trained AI review models, it performs multi-dimensional automated review of integrated expense reimbursement data, identifies various abnormal issues, and automatically generates preliminary review opinions. Review and triage module: Based on the results of AI-automated review, compliant documents are automatically transferred to the payment process, while abnormal documents and preliminary review opinions are triaged to the manual review module; Manual review module: Used by finance personnel to review abnormal documents, enter final review comments, and support the editing, modification and feedback of review comments; The query and feedback module is used by employees to query the reimbursement progress, review results and reasons for abnormalities, receive review comments from finance personnel, and support the resubmission of rectified documents. Closed-loop management and analysis module: used to summarize the entire process of audit data, conduct multi-dimensional analysis (including reimbursement efficiency analysis, anomaly type statistics, project cost collection analysis, policy implementation analysis), update the reimbursement policy knowledge base and the training data of the AI ​​audit model, and optimize audit rules; Data storage module: Used to store data for the entire reimbursement process, including travel application data, itinerary data, reimbursement document data, audit result data, project / customer related data, policy knowledge base data, etc., to ensure data security and traceability. The specific embodiments of the present invention are as follows: (I) System Architecture Deployment The financial intelligent auditing system described in this invention adopts a distributed architecture deployment, including three core layers: the client layer, the application service layer, and the data layer. 1. Client-side layer: This includes the employee side and the finance side. The employee side supports access to the AI ​​expense reimbursement assistant through various channels such as WeChat Work, DingTalk, Web, and mobile APP, enabling operations such as travel application, itinerary construction, policy consultation, expense submission, and progress inquiry. The finance side accesses the system through the Web, enabling operations such as abnormal document review, review rule configuration, and data statistical analysis.

[0036] 2. Application Service Layer: Deploys AI interaction module, data collection and integration module, AI automatic review module, review and triage module, manual review module, query and feedback module, and closed-loop management and analysis module. Each module realizes data interaction and collaborative work through API interfaces. Among them, the AI ​​interaction module adopts natural language processing (NLP) technology, supports both voice and text interaction methods, and has the functions of intent recognition, information extraction, and semantic understanding. The AI ​​automatic review module is deployed on a GPU server to ensure the efficiency of model operation.

[0037] 3. Data Layer: A distributed database (such as the Hadoop Distributed File System) is used to store all process data, including structured data (such as expense report details, audit results, project / customer information) and unstructured data (such as scanned copies of vouchers and itinerary images). At the same time, a data encryption module is deployed to encrypt sensitive financial data to ensure data security. It connects with the company's existing attendance system, project management system, and third-party transportation (Ctrip, Fliggy) and accommodation (Huazhu, Jinjiang) platforms to achieve real-time data synchronization through API interfaces.

[0038] (II) System Initialization Configuration Before the system goes live, finance personnel complete the initial configuration through the finance terminal: 1. Configure preset audit rules: including basic information verification rules (such as expense reports must include travel information, expense details, and scanned copies of vouchers), expense standard rules (such as different levels of employees' travel class standards and accommodation expense limits, for example, directors can travel in business class with an accommodation expense limit of 800 yuan / night; ordinary employees can travel in economy class with an accommodation expense limit of 400 yuan / night), business logic rules (such as prohibiting the application for entertainment expenses during travel, and requiring approval forms for entertainment in other locations), and association rules (such as the project code must be consistent with the project code in the enterprise project management system).

[0039] 2. Build a reimbursement policy knowledge base: Input relevant policy documents on corporate travel reimbursement, including the scope of reimbursement, voucher requirements, approval process, and rules for handling special circumstances. Use NLP technology to segment and semantically annotate the policy text to build a structured policy knowledge base, supporting the AI ​​reimbursement assistant to quickly retrieve and provide feedback.

[0040] 3. Train the AI ​​audit model: Collect historical expense reimbursement data from the past 3-5 years (including compliant documents, abnormal documents, and corresponding audit comments), clean and label the data (labeling the type and cause of the abnormality), and train the AI ​​audit model using deep learning algorithms (such as the CNN-LSTM hybrid model). Optimize the model parameters through multiple iterations to ensure that the model's recognition accuracy for various abnormal issues is no less than 95%. At the same time, set up a model self-learning mechanism and regularly fine-tune the model using new data summarized by the closed-loop management and analysis module.

[0041] (III) Specific Operation Procedure Examples Taking the example of an employee A of a certain company needing to travel to Shanghai for business expansion due to the business needs of client C of project B, the specific implementation process of this invention will be explained in detail: 1. Pre-reimbursement interaction stage: Employee A sent a voice message to the AI ​​expense reimbursement assistant via WeChat Work: "I want to apply for a business trip to Shanghai from March 10th to March 15th, 2026, to coordinate with client C's project B business." The AI ​​expense reimbursement assistant used voice recognition and intent recognition to extract key information (traveler: A; travel dates: March 10th-15th, 2026; destination: Shanghai; reason: business expansion with client C for project B), generated a travel application form, and sent it to employee A for confirmation. After employee A confirmed that everything was correct, the AI ​​expense reimbursement assistant further inquired: "Do you need assistance in creating the travel itinerary? You can automatically retrieve flight and hotel information." Employee A replied: "Yes, please help me book the high-speed train from Beijing to Shanghai on March 10th and the high-speed train from Shanghai back to Beijing on March 15th. The accommodation should be near client C's company." The AI ​​expense reimbursement assistant connects to the 12306 platform and the Huazhu Hotels platform. Based on employee A's needs, it selects suitable high-speed train routes (G101 Beijing South Station - Shanghai Hongqiao Station on March 10, 08:00-12:30; G102 Shanghai Hongqiao Station - Beijing South Station on March 15, 14:00-18:30) and hotels (Shanghai XX Hotel, located near client C's company, stay from March 10 to March 15, room rate 380 yuan / night), generates a complete business trip itinerary, links it to project B and client C's information, and sends it to employee A for confirmation. After employee A confirms the itinerary, the AI ​​expense reimbursement assistant prompts: "Business trip itinerary confirmed. You can consult the reimbursement policy at any time, such as 'Shanghai business trip accommodation expense standards'."

[0042] 2. Reimbursement submission stage: After the business trip, Employee A submitted expense reports to the AI ​​expense reimbursement assistant via WeChat Work, uploading photos of high-speed rail tickets, hotel invoices, and expense details (high-speed rail tickets: 480 yuan / ticket, 2 tickets total; hotel expenses: 1900 yuan; meal allowance: 300 yuan / day, 5 days total, 1500 yuan). The AI ​​expense reimbursement assistant used OCR recognition technology to extract the ticket information and verified the consistency between the expense details and the itinerary. It found that the hotel expenses of 1900 yuan matched the 380 yuan / night × 5 nights = 1900 yuan in the itinerary, and the high-speed rail ticket information matched the itinerary. At the same time, it verified the completeness of the information for related items B and customer C, and prompted Employee A to supplement the travel summary description before uploading the complete expense data to the financial intelligent audit system.

[0043] 3. AI-automated review stage: After receiving expense reimbursement data, the financial intelligent auditing system initiates AI-powered automatic auditing: (1) Basic information review: Verify that the information on the reimbursement documents is complete (including itinerary information, expense details, scanned copies of receipts, project / customer association information), the format is standardized, the project code of project B is consistent with the code in the project management system, and the information of customer C is true and valid; (2) Defect identification: a) Duplicate reimbursements within the same time period: Comparing the historical reimbursement data of employee A in the system, no other travel reimbursement documents were found during the period from March 10 to March 15, 2026, and there were no duplicate reimbursements; b) Hotel and transportation conflict: Comparing the itinerary, the hotel check-in time (March 10) is logically consistent with the arrival time of the high-speed train from Beijing to Shanghai (March 10, 12:30), and the hotel check-out time (March 15) is logically consistent with the departure time of the high-speed train from Shanghai back to Beijing (March 15, 14:00). There is no time conflict. c) The location where the expense occurred is inconsistent with the location where the employee clocked in: By connecting to the attendance system, the clock-in data of employee A from March 10th to March 15th was obtained. All of the clock-in data were from Shanghai, which is consistent with the location where the expense occurred. There were no abnormalities. d) Hospitality expenses incurred during the trip: Review the expense details; no hospitality expenses were declared, and there were no abnormalities. e) Compliance review of out-of-town hospitality expenses: No out-of-town hospitality expenses were declared, and there were no abnormalities; (3) Expense standard review: Employee A is a regular employee, the class of transportation is economy class (in compliance with the standard), the accommodation fee is RMB 380 / night (lower than the upper limit of RMB 400 / night for regular employees, in compliance with the standard), the meal allowance is RMB 300 / day (in compliance with the company’s travel meal allowance standard in Shanghai), and there are no expenses exceeding the standard; (4) Verification of the authenticity of vouchers: The anti-counterfeiting marks of high-speed rail tickets and hotel invoices are verified by image recognition technology, and the authenticity of the invoices is verified by connecting with the tax system to confirm that the vouchers are genuine and valid.

[0044] 4. Review Result Determination and Distribution: The AI-powered automatic review found no irregularities in the expense report, determined it to be a compliant document, automatically approved it, and transferred it to the company's financial payment system to arrange for the payment of the reimbursement.

[0045] 5. Post-reimbursement inquiry stage: Employee A asked the AI ​​expense reimbursement assistant via WeChat: "What is the progress of my expense reimbursement for my business trip to Shanghai in March?" The AI ​​expense reimbursement assistant checked the system and replied: "Your March 2026 Shanghai business trip expense reimbursement form (item number: BX20260316001) has been automatically reviewed by AI and has been transferred to the payment system. It is expected to arrive within 3 working days."

[0046] 6. Example of handling abnormal documents: If employee A's expense report includes an entertainment expense of 800 yuan for 13:00 on March 10 (during the travel period, i.e., the G101 high-speed train arrives at Shanghai Hongqiao Station at 12:30, and the entertainment expense occurred at 13:00, which is a subsequent and adjacent time period during the travel period, and no approval form for entertainment in another location was provided): During the AI-automated review phase, the system identified two anomalies: "Entertainment expenses incurred during travel" and "No approval form for out-of-town entertainment." It automatically generated preliminary review comments: "1. The entertainment expenses occurred at 13:00 on March 10, which is close to the arrival time of the high-speed train from Beijing to Shanghai (12:30). This falls within the relevant time period of travel. According to the company's travel reimbursement policy, entertainment expenses are prohibited from being claimed during travel. 2. This entertainment expense is for out-of-town entertainment, and no corresponding approval form was provided. This does not meet the requirements for out-of-town entertainment reimbursement. Please provide the approval form or an explanation."

[0047] The audit and routing module routed the abnormal document and preliminary audit comments to the manual audit stage. The finance staff logged into the finance terminal, viewed the abnormal document and preliminary audit comments, and after further verification, gave the final audit opinion: "Return for rectification. The entertainment expense declaration during the travel period needs to be deleted. If it is necessary out-of-town entertainment, a complete out-of-town entertainment approval form needs to be added and resubmitted."

[0048] When employee A checked the reimbursement progress through the AI ​​reimbursement assistant, he received feedback that the review was not approved, along with the specific reasons for the abnormality and the review comments. After deleting the entertainment expense application based on the comments and resubmitting it, the system restarted the AI ​​automatic review. After the review was approved, it proceeded to the payment process.

[0049] 7. Closed-loop management and analysis: The closed-loop management and analysis module summarizes all data from the entire reimbursement process, including application time, submission time, review time, expense details, and related project / customer information. It generates monthly travel reimbursement analysis reports, including anomaly type statistics (such as the percentage of duplicate reimbursements and itinerary conflicts), project expense aggregation analysis (such as the summary of travel expenses for Project B), and review efficiency analysis (such as the AI-automated approval rate and average review time). Based on the analysis results, it updates the reimbursement policy knowledge base (such as supplementing the submission requirements for out-of-town hospitality approval forms) and adds new anomaly cases to the training data of the AI ​​review model to optimize the model's recognition accuracy.

[0050] (iv) Key Technology Description 1. Natural Language Processing Technology: The AI ​​interaction module adopts the BERT model based on the Transformer architecture to achieve accurate recognition and understanding of natural language. It supports voice and text interaction, can accurately extract key information from travel applications, quickly retrieve reimbursement policy knowledge base and provide accurate answers.

[0051] 2. AI Review Model: A CNN-LSTM hybrid model is adopted. The CNN layer is used to extract structured features (such as expense amount, time, and location) from the expense report, and the LSTM layer is used to capture logical relationships in the time series (such as the logic of the trip time and the correlation between the expense occurrence time and the check-in time). Through training with a large amount of historical data, it can accurately identify various abnormal issues.

[0052] 3. Multi-dimensional data comparison technology: By connecting to multiple systems (attendance, project management, third-party transportation / accommodation platforms, tax systems) through the data collection and integration module, real-time synchronization and multi-dimensional comparison of cross-system data are achieved to ensure the comprehensiveness and accuracy of the audit.

[0053] 4. Data Encryption and Security Technology: Sensitive financial data is encrypted and stored using the AES encryption algorithm, and access control is set (employees can only view their own expense reimbursement data, and finance personnel can only view and approve related data) to ensure data security and privacy.

[0054] The beneficial effects of the present invention are as follows: Compared with the prior art, the technical effects of the present invention include: 1. Improve review efficiency: AI-powered automatic review enables rapid review of routine and compliant documents without human intervention, significantly shortening the review cycle; at the same time, the AI ​​expense reimbursement assistant helps employees complete pre-application, itinerary construction, and document submission, reducing employee preparation time and the basic review workload of finance personnel.

[0055] 2. Improve audit accuracy: By combining preset audit rules with AI audit models, we can accurately identify various defects in travel expense reimbursements (such as duplicate reimbursements, itinerary conflicts, and discrepancies between expenses and check-in locations), avoiding subjective errors and omissions in manual audits and reducing corporate financial risks.

[0056] 3. Optimize human-machine collaboration efficiency: Only abnormal documents are diverted to manual review, achieving a reasonable allocation of review resources. This allows finance personnel to focus on handling complex and abnormal issues, improving the pertinence and efficiency of manual review. At the same time, the preliminary review opinions generated by AI provide reference for finance personnel, reducing the cost of manual judgment.

[0057] 4. Enable business-related traceability: By linking travel expense reimbursement documents with project and customer information, it is easier to collect project costs and analyze customer-related expenses, providing data support for enterprise business decisions.

[0058] 5. Build a complete reimbursement loop: Form a complete closed loop of "application-submission-review-feedback-rectification-review-payment", so that employees can check the progress and reasons for anomalies in real time. The rectification process is smooth and improves the employee reimbursement experience. At the same time, through the summary and analysis of closed-loop data, the review rules and AI models are continuously optimized to improve the system's adaptability and stability.

[0059] 6. Standardize reimbursement management: Ensure consistency in review standards through unified review rules and policy knowledge base; AI reimbursement assistant answers policy inquiries in real time, reduces the submission of non-compliant documents by employees, and improves the standardization of reimbursement management. 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 it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention without departing from the spirit and scope of the present invention. Any modifications or equivalent substitutions should be covered within the protection scope of the claims of the present invention.

Claims

1. A financial intelligent auditing method, characterized in that, include: S1: Execute the pre-reimbursement interaction stage, interact with employees through the AI ​​reimbursement assistant to complete the pre-operation related to business travel; S2: During the expense reimbursement submission phase, employees submit travel expense reimbursement documents through the AI ​​expense reimbursement assistant; S3: AI-automated review stage. The financial intelligent review system performs multi-dimensional automated review of reimbursement data based on preset review rules and trained AI review models. S4: Perform audit result judgment and sorting. Based on the AI ​​automatic audit results, the expense reimbursement documents are divided into compliant documents and abnormal documents; S5: During the manual review stage, finance personnel review the abnormal documents that have been diverted, refer to the preliminary review opinions generated by AI, give the final review results, and enter the final review opinions into the system. S6: In the post-reimbursement inquiry phase, employees can use the AI ​​reimbursement assistant to check the reimbursement progress and review results; S7: In the closed-loop management phase, the financial intelligent audit system summarizes and analyzes the data from the entire process, updates the reimbursement policy knowledge base and the training data of the AI ​​audit model, and optimizes the audit rules.

2. The financial intelligent auditing method according to claim 1, characterized in that, S1 specifically includes: S11: Quickly create travel requests. Employees can send travel request requests to the AI ​​expense reimbursement assistant using natural language or preset templates. The AI ​​expense reimbursement assistant extracts key information from the request and generates a travel request form. S12: Constructing travel itineraries. The AI ​​expense reimbursement assistant automatically captures or assists employees in entering transportation and accommodation information based on their travel needs, constructs a complete travel itinerary, and associates it with the corresponding project or customer information. S13: Policy consultation interaction. Employees consult the AI ​​expense reimbursement assistant about travel expense reimbursement policies through natural language. The AI ​​expense reimbursement assistant provides real-time feedback on the corresponding policy content based on a preset expense reimbursement policy knowledge base.

3. The financial intelligent auditing method according to claim 2, characterized in that, S2 specifically includes: AI-powered expense reimbursement assistant helps employees verify and supplement expense documents, ensuring the completeness of expense reimbursement information, and uploads the complete expense data to the financial intelligent auditing system.

4. The financial intelligent auditing method according to claim 3, characterized in that, S3 specifically includes: S31: Basic information review, verify the completeness and format of expense reimbursement documents, and confirm the authenticity and validity of the associated project / customer information; S32: Defect and problem identification, through multi-dimensional data comparison, to identify typical abnormal problems in travel expense reimbursement; S33: Expense standard review, compare the reimbursement expense details with the preset travel expense standard, and identify expenses exceeding the standard; S34: Verification of voucher authenticity. Using OCR and image recognition technologies, verify the authenticity and completeness of reimbursement vouchers and identify false or tampered vouchers.

5. The financial intelligent auditing method according to claim 4, characterized in that, S32 specifically includes: a) Duplicate reimbursements within the same time period: Compare the time of expense occurrence in the reimbursement documents with the approved / pending reimbursement documents to identify duplicate reimbursements for the same expense item within the same time period; b) Hotel and off-site transportation conflicts: Compare the hotel check-in / check-out times in the travel itinerary with the departure / arrival times of off-site transportation to identify time logic conflicts; c) Inconsistency between the place of expense occurrence and the place of attendance: Obtain employee attendance data during business trips, compare the place of expense occurrence with the place of attendance, and identify irregular expense claims outside of business trips; d) Hospitality expenses incurred during travel: Compare the travel time periods of the transportation itinerary with the time when hospitality expenses were incurred to identify any unauthorized hospitality expense claims during the travel period; e) Compliance review of out-of-town hospitality expenses: Verify the approval process for out-of-town hospitality expenses, the relevance between the hospitality recipients and the reason for the trip, and identify non-compliant out-of-town hospitality expenses.

6. The financial intelligent auditing method according to claim 5, characterized in that, S4 specifically includes: S41: If the expense report is normal, it is considered a compliant report and will be automatically approved and proceed to the next payment process. S42: If there are any abnormalities in the expense reimbursement documents, they will be judged as abnormal documents. The AI ​​audit model will automatically generate preliminary audit opinions and transfer the abnormal documents and preliminary audit opinions to the manual audit stage.

7. The financial intelligent auditing method according to claim 6, characterized in that, The preset audit rules mentioned in S3 include basic information verification rules, expense standard rules, business logic rules, and relationship rules, which can be configured and updated by the company's financial personnel according to actual business needs. The AI ​​audit model is trained by a large amount of historical reimbursement data (including compliant documents, abnormal documents, and corresponding audit opinions), has self-learning capabilities, and can continuously optimize the recognition accuracy based on the summary and analysis results of the entire process data.

8. The financial intelligent auditing method according to claim 7, characterized in that, In S12, the AI ​​expense reimbursement assistant can automatically retrieve flight tickets, high-speed rail tickets, and hotel order information by connecting to third-party transportation and accommodation platform interfaces.

9. The financial intelligent auditing method according to claim 8, characterized in that, The multi-dimensional audit in S3 also includes the matching analysis of travel expenses with related projects / clients. Combined with project budget data, it audits whether travel expenses are within the project budget, providing support for project cost control.

10. A financial intelligent auditing system for implementing the above-mentioned financial intelligent auditing method, characterized in that, include: AI Interaction Module: Also known as AI Expense Reimbursement Assistant, it is used to interact with employees throughout the entire process, including travel application creation, travel itinerary construction, policy consultation, guidance on submitting expense reimbursement documents, and expense reimbursement progress inquiry. It supports natural language interaction and has functions such as information extraction, data verification, and real-time feedback. Data collection and integration module: Used to collect travel-related data, including employee expense reports, scanned copies of vouchers, related project / customer information, connect to the attendance system to obtain clock-in data, connect to third-party transportation and accommodation platforms to obtain itinerary order data, connect to the project management system to obtain project budget data, and standardize and integrate the collected data; AI Automatic Review Module: Based on preset review rules and trained AI review models, it performs multi-dimensional automated review of integrated expense reimbursement data, identifies various abnormal issues, and automatically generates preliminary review opinions. Review and triage module: Based on the results of AI-automated review, compliant documents are automatically transferred to the payment process, while abnormal documents and preliminary review opinions are triaged to the manual review module; Manual review module: Used by finance personnel to review abnormal documents, enter final review comments, and support the editing, modification and feedback of review comments; The query and feedback module is used by employees to query the reimbursement progress, review results and reasons for abnormalities, receive review comments from finance personnel, and support the resubmission of rectified documents. Closed-loop management and analysis module: used to summarize the entire process of review data, conduct multi-dimensional analysis, update the reimbursement policy knowledge base and the training data of the AI ​​review model, and optimize the review rules; Data storage module: Used to store data for the entire reimbursement process, including travel application data, itinerary data, reimbursement document data, audit result data, project / customer related data, policy knowledge base data, etc., to ensure data security and traceability.

Images