Platform for realizing accounting intelligent report risk management and control
The intelligent accounting report risk management platform utilizes large language models and natural language interaction to automatically monitor and handle report risks, solving the problem that existing accounting report systems struggle to detect and manage report risks, and improving data accuracy and management efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING NANTIAN SOFTWARE
- Filing Date
- 2025-12-12
- Publication Date
- 2026-05-05
AI Technical Summary
In existing technologies, accounting reporting systems struggle to effectively identify and manage reporting risks, leading to inaccurate data and errors in business data, which impact decision-making and oversight. Furthermore, their reliance on manual processing results in low efficiency.
A platform for intelligent accounting report risk management is adopted, which integrates a back-end support module, an application service module, an accounting knowledge base, multiple MCP servers, an intelligent agent main control module, and a client module. It uses a large language model and natural language interaction to automatically monitor and handle report risks.
It enables early detection and management of reporting risks, improves the work efficiency of accounting personnel, reduces skill requirements, and enhances data accuracy and system effectiveness.
Smart Images
Figure CN121981518A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of accounting technology, specifically relating to a platform for implementing intelligent accounting report risk management. Background Technology
[0002] Financial institutions' accounting reporting systems (or accounting reporting modules, etc., providing accounting report management functions) may present two main categories of risks that require handling by accounting personnel (we call these reporting risks). One category is inaccurate report data, where there are contradictions or unreasonable data relationships within a single report or between multiple reports. This could be due to problems during manual data adjustments, bugs in the report generation program, or other reasons. The other category is problems with the business data used to generate the reports, which could be due to errors in the original business data used to generate the reports, or other reasons.
[0003] Accountants need to identify and resolve these issues promptly, otherwise it could cause losses to financial institutions, such as affecting leaders' understanding of the actual situation and making correct decisions, or leading to penalties from regulatory agencies due to problems with data reporting to regulators.
[0004] Typically, reporting systems provide built-in validation functions for intra-table and inter-table relationships. However, these validation functions can be ineffective or incomplete. For example, someone might manually adjust certain report data, the reporting system might have bugs, or the current version of the validation program might not perform calculations or validations for certain implicit relationships. The most likely source of problems is incorrect raw business data used to generate the reports, and the reporting system itself cannot detect these issues.
[0005] The current situation regarding the discovery, analysis, and handling of risks in financial statements is as follows.
[0006] 1. Improve the data quality of the raw business data used to generate reports through technical and management means; 2. Continuously improve the data validation relationships in reports and achieve automatic auditing and validation through technical means; 3. For important reports, the accounting staff will conduct a manual check before submitting them. When doing this, the accounting staff mainly rely on their own experience or some statistical information obtained from the data analysis system.
[0007] 4. When identifying risks in financial statements and conducting risk analysis, accountants will consult relevant information in the accounting reporting system and accounting system for risk analysis.
[0008] 5. Handling report risks may be a complex systemic project depending on the specific problems found, including management and technical measures. Among the technical measures, it may be to improve system programs or improve data control methods.
[0009] The most critical issue in the entire process of financial statement risk management is to identify financial statement risks as early as possible, but currently this mainly relies on manual processing by accounting personnel.
[0010] Therefore, overcoming the shortcomings of existing technologies is an urgent problem to be solved in the field of accounting technology. Summary of the Invention
[0011] The purpose of this invention is to address the shortcomings of existing technologies and provide a platform for implementing intelligent reporting and risk management in accounting.
[0012] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention provides a platform for realizing intelligent reporting and risk management of accounting, including a backend support module, an application service module, an accounting knowledge base, multiple MCP servers, an intelligent agent main control module, and a client module; in: The backend support module includes a monitoring system and a large language model; the monitoring system is used to monitor the report data in the accounting report system and output alarm information based on its built-in report risk monitoring model. The application service module includes an accounting system and an accounting reporting system; the accounting system processes accounting data; the accounting reporting system generates reports based on the accounting data. The accounting knowledge base is a vector database that stores accounting knowledge and its metadata. Multiple MCP servers, including: The accounting system MCP server is used to encapsulate and call the service interfaces of the accounting system. The accounting reporting system MCP server is used to encapsulate and call the service interfaces of the accounting reporting system. The monitoring system's MCP server is used to transmit alarm information, query requests, and new reports and risk monitoring models between the monitoring system and the intelligent agent's main control module. The knowledge base MCP server is used to respond to query requests and retrieve information from the accounting knowledge base; The intelligent agent master control module is used to control the conversation flow. The intelligent agent master control module is used to: receive the user's interactive request through natural language input, organize prompt words containing callable function information and submit them to the large language model, receive the operation instructions returned by the large language model, and call the corresponding MCP server according to the operation instructions to perform monitoring, query or handling operations of report risks. The client module provides a user interface for receiving user input, displaying session results, and receiving alarm notifications.
[0013] Furthermore, preferably, the agent's main control module integrates a program execution environment; when the large language model returns automatically generated program code for a target function that does not have a corresponding function interface, the program code is executed in the program execution environment, and the program code can call the functions provided by the corresponding MCP server or the agent's main control module.
[0014] Furthermore, preferably, the accounting knowledge base stores content structure description information for various reports; the content structure description information includes semantic descriptions of report data items and their corresponding resource names in the accounting report system; The accounting knowledge base stores report content structure description information in JSON format text, which is used to establish a mapping relationship between accounting terms and operable data items in the accounting report system.
[0015] Furthermore, preferably, the initial version of the report content structure description information is automatically generated by the report format file of the accounting report system combined with data dictionary information, and can be updated and stored in the accounting knowledge base after being adjusted by accounting personnel.
[0016] Furthermore, preferably, the monitoring system obtains report data through a monitoring agent deployed on the accounting reporting system side, and supports two data acquisition strategies: timed triggering or triggering by new report generation events in the accounting reporting system.
[0017] Furthermore, preferably, the client module includes an accounting intelligent assistant client and a notification receiving device; the accounting intelligent assistant client obtains and displays alarm information through polling; the notification receiving device receives alarm information via SMS or email.
[0018] A second aspect of this invention provides a method for generating a report risk monitoring model based on the aforementioned platform for realizing intelligent reporting risk control in accounting, comprising the following steps: The client module receives user reports and risk monitoring requests input in natural language. The intelligent agent main control module calls the knowledge base MCP server to obtain the report content structure description information related to the requirements; The intelligent agent main control module organizes the required information, the report content structure description information, and prompt words of the preset program template, and submits them to the large language model; Generate executable report risk monitoring model programs from large language models; The intelligent agent main control module submits the generated report risk monitoring model program to the monitoring system through the monitoring system MCP server.
[0019] Furthermore, preferably, before submitting the generated model program to the monitoring system, the method further includes: The intelligent agent main control module returns the requirement document and model program generated by the large language model to the client module for user review; Based on the user's review feedback, adjust the input requirement description and repeat the generation steps until the user confirms.
[0020] A third aspect of the present invention provides a method for handling report risks based on the aforementioned platform for realizing intelligent reporting risk management in accounting, characterized by comprising the following steps: The client module receives risk management requests from users, input in natural language, for specific reports. The intelligent agent main control module calls the knowledge base MCP server to obtain the content structure description information of the specific report; The intelligent agent main control module organizes a prompt word containing the processing request and the resource name of the report data item, submits it to the large language model, and receives the calling instruction returned by the large language model; The intelligent agent main control module calls the parameter management interface of the accounting report system MCP server according to the calling instruction, so as to modify or add data verification rules in the accounting report system.
[0021] A fourth aspect of the present invention provides a method for querying report risk information based on the aforementioned platform for realizing intelligent reporting risk control in accounting, characterized by comprising the following steps: The client module receives user-triggered report risk details query requests; The intelligent agent's main control module calls the monitoring system's MCP server to obtain detailed alarm information; Based on the user's further query intent, the intelligent agent main control module calls the large language model to determine the data source to be queried. The large language model returns an instruction to call the accounting statement system MCP server or the accounting system MCP server for querying. The intelligent agent main control module calls the corresponding model context protocol server according to the query command to obtain the associated report data or accounting raw data, and returns it to the client module for display.
[0022] This invention provides a platform for intelligent risk management of accounting reports. Targeting the report risks that accountants need to discover, analyze, and handle, it is based on the MCP protocol, uses a large language model as its intelligent core, and is supported by an accounting knowledge base. This allows accountants to construct report risk monitoring models within the monitoring system through a natural language interface, enabling them to discover report risks in a timely manner; query information related to report risks in the monitoring system, accounting report system, and accounting system to complete the analysis of specific report risks; and allow accountants to perform parameter adjustments and modifications to the accounting report system and accounting system through natural language interaction, supporting the handling of such report risks. This achieves early detection of accounting report risks, improves the efficiency of accountants' report risk management, and enhances the data validity and accuracy of the accounting report system.
[0023] Compared with the prior art, the beneficial effects of this invention are as follows:
[0024] 1. For report risks requiring monitoring, accounting personnel can design monitoring models for these risks through natural language interaction, submit them to the monitoring system, and after analysis and confirmation by technical staff, deploy them to implement the new report risk monitoring function. With the support of the report data detection module in the monitoring system, the report risk monitoring model can obtain newly generated report information in real-time or near real-time. Once a potential report risk is identified, an alarm is immediately sent to notify accounting personnel for further analysis and handling. This minimizes the time required to discover report risk events. Supported by large language models, intelligent agents, and knowledge bases, accounting personnel do not need to understand the specific technical implementation; technical personnel directly interact with the pre-implemented report risk monitoring model, significantly reducing the skill requirements for both accounting and technical personnel and shortening the deployment time of the report risk monitoring model. This ensures that the implementation of the new risk monitoring model can support the report risk management needs faced by accounting operations.
[0025] 2. Accounting personnel can interact through natural language and view detailed information related to risk events in the monitoring system's reports via the client. They can query report information and basic transaction information related to risk events in the accounting reporting system and accounting system, which can be used to analyze and locate problems and formulate solutions.
[0026] 3. Accounting personnel can use natural language interaction to adjust parameters in the accounting reporting system and accounting system, providing support for parameter adjustment in both systems for handling reporting risks. Attached Figure Description
[0027] Figure 1 This is a schematic diagram of the platform architecture for implementing intelligent accounting report risk management according to the present invention. Detailed Implementation
[0028] The present invention will now be described in further detail with reference to the embodiments.
[0029] Those skilled in the art will understand that the following embodiments are for illustrative purposes only and should not be construed as limiting the scope of the invention. Where specific techniques or conditions are not specified in the embodiments, they are performed in accordance with the techniques or conditions described in the literature in the field or according to the product instructions. Materials or equipment whose manufacturers are not specified are all conventional products that can be obtained by purchase. Example 1
[0030] like Figure 1 As shown, a platform for realizing intelligent reporting and risk management in accounting includes a backend support module, an application service module, an accounting knowledge base, multiple MCP servers, an intelligent agent main control module, and a client module. in: The backend support module includes a monitoring system and a large language model; the monitoring system is used to monitor the report data in the accounting report system and output alarm information based on its built-in report risk monitoring model. The application service module includes an accounting system and an accounting reporting system; the accounting system processes accounting data; the accounting reporting system generates reports based on the accounting data. The accounting knowledge base is a vector database that stores accounting knowledge and its metadata. Multiple MCP servers, including: The accounting system MCP server is used to encapsulate and call the service interfaces of the accounting system. The accounting reporting system MCP server is used to encapsulate and call the service interfaces of the accounting reporting system. The monitoring system's MCP server is used to transmit alarm information, query requests, and new reports and risk monitoring models between the monitoring system and the intelligent agent's main control module. The knowledge base MCP server is used to respond to query requests and retrieve information from the accounting knowledge base; The intelligent agent master control module is used to control the conversation flow. The intelligent agent master control module is used to: receive the user's interactive request through natural language input, organize prompt words containing callable function information and submit them to the large language model, receive the operation instructions returned by the large language model, and call the corresponding MCP server according to the operation instructions to perform monitoring, query or handling operations of report risks. The client module provides a user interface for receiving user input, displaying session results, and receiving alarm notifications. Example 2
[0031] like Figure 1As shown, a platform for realizing intelligent reporting and risk management in accounting includes a backend support module, an application service module, an accounting knowledge base, multiple MCP servers, an intelligent agent main control module, and a client module. in: The backend support module includes a monitoring system and a large language model; the monitoring system is used to monitor the report data in the accounting report system and output alarm information based on its built-in report risk monitoring model. The application service module includes an accounting system and an accounting reporting system; the accounting system processes accounting data; the accounting reporting system generates reports based on the accounting data. The accounting knowledge base is a vector database that stores accounting knowledge and its metadata. Multiple MCP servers, including: The accounting system MCP server is used to encapsulate and call the service interfaces of the accounting system. The accounting reporting system MCP server is used to encapsulate and call the service interfaces of the accounting reporting system. The monitoring system's MCP server is used to transmit alarm information, query requests, and new reports and risk monitoring models between the monitoring system and the intelligent agent's main control module. The knowledge base MCP server is used to respond to query requests and retrieve information from the accounting knowledge base; The intelligent agent master control module is used to control the conversation flow. The intelligent agent master control module is used to: receive the user's interactive request through natural language input, organize prompt words containing callable function information and submit them to the large language model, receive the operation instructions returned by the large language model, and call the corresponding MCP server according to the operation instructions to perform monitoring, query or handling operations of report risks. The client module provides a user interface for receiving user input, displaying session results, and receiving alarm notifications.
[0032] The agent's main control module integrates a program execution environment. When the large language model returns automatically generated program code for a target function that does not have a corresponding function interface, the program code is executed in the program execution environment. The program code can call the corresponding MCP server or the functions provided by the agent's main control module.
[0033] The accounting knowledge base stores content structure description information for various reports; the content structure description information includes semantic descriptions of report data items and their corresponding resource names in the accounting reporting system; The accounting knowledge base stores report content structure description information in JSON format text, which is used to establish a mapping relationship between accounting terms and operable data items in the accounting report system.
[0034] The initial version of the report content structure description information is automatically generated by the report format file of the accounting report system combined with data dictionary information, and can be updated and stored in the accounting knowledge base after being adjusted by accounting personnel.
[0035] The monitoring system obtains report data through a monitoring agent deployed on the accounting reporting system side, and supports two data acquisition strategies: timed triggering or triggering by new report generation events in the accounting reporting system.
[0036] The client module includes an accounting intelligent assistant client and a notification receiving device; the accounting intelligent assistant client obtains and displays alarm information through polling; the notification receiving device receives alarm information via SMS or email. Example 3
[0037] A method for generating a report risk monitoring model based on the platform for intelligent reporting risk control in accounting, as described in Embodiment 1 or Embodiment 2, includes the following steps: The client module receives user reports and risk monitoring requests input in natural language. The intelligent agent main control module calls the knowledge base MCP server to obtain the report content structure description information related to the requirements; The intelligent agent main control module organizes the required information, the report content structure description information, and prompt words of the preset program template, and submits them to the large language model; Generate executable report risk monitoring model programs from large language models; The intelligent agent main control module submits the generated report risk monitoring model program to the monitoring system through the monitoring system MCP server. Example 4
[0038] A method for generating a report risk monitoring model based on the platform for intelligent reporting risk control in accounting, as described in Embodiment 1 or Embodiment 2, includes the following steps: The client module receives user reports and risk monitoring requests input in natural language. The intelligent agent main control module calls the knowledge base MCP server to obtain the report content structure description information related to the requirements; The intelligent agent main control module organizes the required information, the report content structure description information, and prompt words of the preset program template, and submits them to the large language model; Generate executable report risk monitoring model programs from large language models; The intelligent agent main control module submits the generated report risk monitoring model program to the monitoring system via the monitoring system MCP server. Before submitting the generated model program to the monitoring system, the following steps are also included: The intelligent agent main control module returns the requirement document and model program generated by the large language model to the client module for user review; Based on the user's review feedback, adjust the input requirement description and repeat the generation steps until the user confirms. Example 5
[0039] A method for handling reporting risks based on a platform for intelligent accounting reporting risk management as described in Embodiment 1 or Embodiment 2 includes the following steps: The client module receives risk management requests from users, input in natural language, for specific reports. The intelligent agent main control module calls the knowledge base MCP server to obtain the content structure description information of the specific report; The intelligent agent main control module organizes a prompt word containing the processing request and the resource name of the report data item, submits it to the large language model, and receives the calling instruction returned by the large language model; The intelligent agent main control module calls the parameter management interface of the accounting report system MCP server according to the calling instruction, so as to modify or add data verification rules in the accounting report system. Example 6
[0040] A method for querying report risk information based on a platform for intelligent accounting report risk management as described in Embodiment 1 or Embodiment 2, characterized by comprising the following steps: The client module receives user-triggered report risk details query requests; The intelligent agent's main control module calls the monitoring system's MCP server to obtain detailed alarm information; Based on the user's further query intent, the intelligent agent main control module calls the large language model to determine the data source to be queried. The large language model returns an instruction to call the accounting statement system MCP server or the accounting system MCP server for querying. The intelligent agent main control module calls the corresponding model context protocol server according to the query command to obtain the associated report data or accounting raw data, and returns it to the client module for display. Example 7
[0041] like Figure 1 As shown, a platform for intelligent reporting and risk management in accounting can be broadly categorized into six types, as follows: The first category is backend support modules, including monitoring systems and large language models; The second category is application service modules, including accounting reporting systems and accounting systems; The third category is accounting knowledge base; The fourth category is MCP servers that interact with various systems; The fifth category is the intelligent agent main control module; The sixth category is client-side modules.
[0042] The descriptions of each system are as follows.
[0043] 1. Backend support modules 1.1 Monitoring System The monitoring system here specifically refers to an application monitoring system used to monitor the report data generated by the accounting reporting system. The monitoring system provides both event-driven and timed accounting report information acquisition capabilities, supports various report risk monitoring models, and will issue alarm information after identifying report risk events. The monitoring system has a built-in report data detection module, and the report risk monitoring model can obtain newly generated report information in real-time or near real-time.
[0044] 1.2 Large Language Model Large language models refer to generative large language models such as Deepseek, Qianwen, and Doubao. These models can be provided through vendor-provided services or through local deployment. Local deployment ensures data security but is more expensive.
[0045] 2. Application service modules 2.1 Accounting Reporting System An accounting reporting system refers to the internal production system of an enterprise (especially financial institutions such as banks), or it may be an accounting reporting module, providing functions for generating and managing accounting reports. To support the implementation of the accounting intelligent assistant's report risk management function, the system needs to be configured and adjusted to provide service call support for the reporting system's MCP server. This mainly includes two types of transaction service functions: one is accounting report query transaction services, used by accounting personnel to query data related to report risk management; the other is accounting report system parameter (including data validation formulas) management transaction services, executed when accounting personnel need to adjust the parameters of the reporting system to handle report risks.
[0046] 2.2 Accounting System An accounting system refers to the internal production system of an enterprise (especially financial institutions such as banks). It may be called an accounting middle platform or various other names, and its main function is accounting. The raw business data required for the accounting reporting system to generate reports is managed by the accounting system. To support the implementation of these functions, configuration adjustments are needed to provide service call support for the accounting system's MCP server. This mainly includes two types of transaction service functions: one is the raw accounting data query transaction service, used by accounting personnel to query data related to report risk management; the other is the accounting system parameter management transaction service, executed when accounting personnel need to adjust accounting system parameters to standardize the raw business data used to generate reports when handling report risks.
[0047] 3. Accounting Knowledge Base The Accounting Knowledge Base is a vector database storing accounting knowledge. Besides knowledge documents, it also stores corresponding metadata, such as knowledge type, source (filename, large language model, etc.), location, publisher, and publication time. Its content includes policies, standards, and regulations related to finance and accounting, as well as content added by accounting personnel. This content is divided into two main categories: management and knowledge. In addition, it includes accounting rule knowledge, which is loaded from the accounting rule base of the accounting system.
[0048] The accounting knowledge base also stores the content structure of each type of report in the accounting reporting system and the mapping information for the resource names of report data items. Each type of report content structure is described using a JSON-formatted text file, which includes the corresponding report data item resource name (used to reference reports and their data items in the accounting reporting system when automatically generating report risk monitoring models). The two-dimensional table illustrates the report structure through a tree-like sequence describing the columns. Describing the report content structure primarily establishes the correspondence between report content and information in the accounting system, providing semantic support for querying data in accounting information. The additional description of the report data item resource names establishes the correspondence between the textual description of the report content and the references to accounting report data items. The initial version of the report content structure is automatically generated from the report format file in the accounting reporting system, with the corresponding information of the report data items in the data dictionary appended. It is then saved after adjustments by accounting personnel.
[0049] 4. MCP server The MCP server primarily provides three types of functions: tools (function interfaces that can be called), resources (data sources that can be queried), and hints (predefined instruction templates).
[0050] MCP servers include: 4.1 Reporting System MCP Server The reporting system's MCP server encapsulates the two types of transaction service interfaces provided by the accounting reporting system. It stores the reporting resource information and parameter resource information of the accounting reporting system.
[0051] 4.2 Accounting System MCP Server This is an MCP server for the accounting system, which encapsulates the two types of transaction service interfaces provided by the accounting system. It also implements the function of converting between the intelligent agent's JSON data format and the accounting system's transaction data format.
[0052] 4.3 Monitoring System MCP Server The monitoring system's MCP server provides the transaction interface between the monitoring system and the intelligent agent, offering three types of functions: first, transmitting report risk event alarm information to the intelligent agent; second, supporting the intelligent agent to query detailed information related to report risk events from the monitoring server; and third, supporting the intelligent agent to transmit new report risk monitoring model information to the monitoring system. The new report risk monitoring model needs to undergo review, adjustment, testing, verification, and release within the monitoring system. These functions are provided by the monitoring system and will not be discussed further in this invention.
[0053] 4.4 Knowledge Base MCP Server The knowledge base MCP server provides knowledge base query functionality. It supports natural language query requests and metadata-based query requests. Multiple query results can be provided.
[0054] 5. Intelligent Agent Main Control Module The accounting intelligent assistant's main control function is implemented by the intelligent agent. It is used for overall control of the session flow. Sessions are initiated by alarm information from the accounting intelligent assistant client, the monitoring system's MCP server, or custom events. Identity authentication, information security, process management, and error handling are implemented by the intelligent agent's main control module. The accounting intelligent assistant provides a Python runtime environment. In most cases, the large language model returns JSON format information calling functions provided by the MCP server or the accounting intelligent assistant. When functions not provided by existing program modules or without corresponding function call interfaces are involved, the accounting intelligent assistant requests the large language model to automatically generate and return a Python program. This Python program can use resources including callable function services provided by the existing MCP server or the accounting intelligent assistant to achieve the target function. This Python program is called by the accounting intelligent assistant.
[0055] 6. Client Module The client module includes an accounting intelligent assistant client and a notification receiving device; 6.1 Accounting Smart Assistant Client The accounting intelligent assistant client offers several client formats, including web pages, mobile apps, PC client programs, and application system plugins, serving as the primary channels for customers to use the accounting intelligent assistant. It mainly provides various functions through interactive sessions. These clients continuously check the accounting intelligent assistant's report risk alarm information through periodic polling, and notify the user via pop-up windows upon receiving alarm information.
[0056] 6.2 Notification to Receiving Equipment The notification receiving device is used to receive report risk alarm information. In addition to the client, alarm information will also be sent to users via SMS, email and other means.
[0057] The specific implementation method of the present invention will be illustrated by taking several typical application scenarios as examples.
[0058] I. Accounting personnel can monitor the risks of new financial statements through natural language interaction. To effectively monitor financial statement risks, a new financial statement risk monitoring model needs to be developed. This model monitors the content of newly generated financial statements specified in the accounting reporting system, performs model calculations, and issues an alert once a risk identified by the model is detected. One risk monitoring model monitors one type of risk, while multiple monitoring models can run simultaneously.
[0059] The report data detection module, which acquires report information, runs on the server side of the monitoring system and deploys a monitoring agent on the machine of the accounting report system. The client side of the report data detection module runs through the monitoring agent. For event-triggered modes (such as triggering a new report event when a new report is generated and saved for the monitored report type), the accounting report system is required to actively transmit the new report event information so that the client side of the report data detection module can obtain the event information in a timely manner and complete the subsequent report acquisition operation.
[0060] Since the report risk monitoring model largely utilizes trend analysis and related analysis results from the big data platform, some report risk monitoring models require querying multiple data records of a specified set of fields in a specified report to run. We have also encapsulated the query function into the report data detection module, and treated the query interface and parameters as resources, managed by the resource management function in the monitoring system's MCP server, for the report risk monitoring model to call.
[0061] (a) Prerequisites: 1. The monitoring system needs to have a management service platform for report risk monitoring models. The developed report risk monitoring models should be able to be described in a fully readable text format (supporting the transfer and operation of newly generated monitoring models).
[0062] 2. The callable resources of the monitoring system (including various service interfaces and report data monitoring modules) are described in the resource information of the monitoring system's MCP server. These resources can all be called using the Python language (the situation is similar when using other programming languages; this explanation only uses Python) and can support the operation of the report risk monitoring model.
[0063] 3. The report data detection module provides two report information acquisition strategies: timed and event-triggered. The event-triggered strategy requires the support of the accounting report system.
[0064] 4. The current user has the authority to call the monitoring system's MCP server to transmit reports and risk monitoring models to the monitoring system.
[0065] 5. Accountants preparing to customize a report risk monitoring model have already mastered the data analysis results obtained from big data platforms or data analysis platforms. These mainly fall into two categories: one is the trend analysis results of key and sensitive report data items, and the other is the results of correlation between multiple report data items. Based on these results, a report risk monitoring model is constructed.
[0066] (II) Process Description: 1. When a user establishes a new session in the accounting intelligent assistant client, they use natural language and accounting terminology to input the purpose and operation steps for creating a report risk monitoring model. For trend-related models, the user explains the range of normal values; for correlation-related models, the user explains the range of normal relationship values; and for other types, the user explains as appropriate. If the value exceeds the normal range, an alarm will be triggered. The accounting intelligent assistant client will then send a request to the accounting intelligent assistant intelligent agent. 2. The accounting intelligent assistant organizes prompt words, including information on callable functions (including information on all functions provided by the MCP server and functions supported by the accounting intelligent assistant, as well as resources and prompt information), and submits user input information to the large language model; 3. The large language model returns the function and parameters to be called next; 4. The accounting intelligent assistant calls the knowledge base MCP server to retrieve the content structure information of the specified report. This information, along with other prompts, is sent to the large language model. The large model returns translation information, converting the terminology in the model requirements described by the accountant into the accurate content structure information of the report, including the names of the report data items. The accounting intelligent assistant also calls the prompts from the monitoring system MCP server, providing the large language model with a template for generating a report risk monitoring model program. Standardized requirements information is then added and passed to the large language model. The large language model returns the generated content, including a standardized requirements document based on the requirements described by the accountant in the interaction, a design document summarized based on the monitoring program implementation, and a report risk monitoring model program that conforms to the monitoring system's rules. 5. The accounting intelligent assistant will return the generated content to the accounting intelligent assistant client for review by the accounting staff, primarily by reviewing the requirements document. If any issues are found, the requirements will be adjusted via a chat session. If the requirements are adjusted, steps 1-5 are essentially repeated until the accounting staff is satisfied. 6. Accounting personnel are required to submit the confirmed financial statement risk monitoring model to the monitoring system; 7. The accounting intelligent assistant submits the prompt words to the large language model; 8. The large language model returns the function and parameters provided by the monitoring system's MCP server to transmit report risk monitoring model information to the monitoring system.
[0067] 9. The accounting intelligent assistant calls the interface provided by the monitoring system's MCP server to transmit the risk monitoring model of the report based on the information returned by the large language model.
[0068] 10. The monitoring system's MCP server transmits the report risk monitoring model to the monitoring system. The successful transmission return information is transmitted from the monitoring system's MCP server and the accounting intelligent assistant intelligent agent to the accounting intelligent assistant client and presented to the user.
[0069] 11. The newly generated report risk monitoring model is reviewed, modified, adjusted, and tested by technical personnel in the monitoring system before being deployed online. This part will not be described in detail in this invention.
[0070] II. Sending report risk alert information Preliminary steps: The accounting intelligent assistant has already established user accounts for accountants and saved information such as mobile phone numbers and email addresses; the accounting intelligent assistant has also been configured to specify which types of report risk alarm information need to be sent to which users and through which channels.
[0071] 1. The report risk monitoring model in the monitoring system detects report risks, and the monitoring system sends alarm information to the monitoring system MCP server; 2. The monitoring system's MCP server forwards alarm information to the accounting intelligent assistant. 3. The accounting intelligent assistant sends alarm information to the target user through the accounting intelligent assistant client based on the configuration and logged-in user information; it also sends SMS and emails to the target user according to the configuration. 4. The target user's accounting intelligent assistant client polls for alarm information in the background and notifies the user via pop-up window when an alarm is detected; the target user receives alarm SMS and alarm email.
[0072] III. Accounting personnel review detailed information related to risks in the monitoring system, accounting reporting system, and accounting system. Process description: 1. In a new session created by the accounting intelligent assistant client, the user enters: "Query the detailed information of the alarm information just sent to me." The accounting intelligent assistant client will then send the request to the accounting intelligent assistant agent. 2. The accounting intelligent assistant will submit the prompt information, including information on callable functions (information on all functions provided by the MCP server and functions supported by the accounting intelligent assistant) and user input information to the large language model; 3. Based on the prompt words, the large language model returns the function and parameters to be called next; 4. The accounting intelligent assistant agent returns information based on the large language model, searches for the current user's alarm information, and calls the transaction provided by the monitoring system MCP server to query detailed alarm information based on the alarm category and identifier in the alarm information. The monitoring system MCP server calls the monitoring system query transaction to obtain detailed alarm information and returns it to the accounting intelligent assistant agent. 5. The accounting intelligent assistant will return detailed alarm information to the accounting intelligent assistant client; 6. Users can view the detailed information returned by the monitoring system; 7. Based on the detailed information, the user uses natural language to input their requirements for querying relevant reports in the accounting reporting system and sends them to the accounting intelligent assistant. 8. The accounting intelligent assistant will submit the prompt information, including information on callable functions (information on all functions provided by the MCP server and functions supported by the accounting intelligent assistant, including information on resources available for operation by the reporting system MCP server) and user input information to the large language model; 9. The large language model returns JSON information for querying the specified report from the MCP server of the reporting system; 10. The accounting intelligent assistant, based on the information returned by the large language model, invokes the query transaction provided by the reporting system's MCP server. The reporting system's MCP server forwards the query transaction to the accounting reporting system to complete the query transaction. The report information is then returned to the accounting intelligent assistant. 11. The accounting intelligent assistant will format the report information and return it to the accounting intelligent assistant client. 12. When a user wants to view report information and needs to query relevant information from the accounting system, the user can use natural language to input the request for querying relevant data from the accounting system based on the report they are interested in, and send it to the accounting intelligent assistant. 13. The accounting intelligent assistant will submit the prompt information, including the information of the functions that can be called (information on all functions provided by the MCP server and the functions supported by the accounting intelligent assistant, including information on the resources available for operation by the accounting system MCP server) and the user input information to the large language model; 14. The large language model returns JSON information for querying specified information from the accounting system's MCP server; 15. The accounting intelligent assistant agent invokes the query transaction provided by the accounting system MCP server based on the information returned by the large language model. The accounting system MCP server forwards the query transaction to the accounting system to complete the query transaction. The queried information is returned to the accounting intelligent assistant agent via the original path. 16. The accounting intelligent assistant will format the report information and return it to the accounting intelligent assistant client.
[0073] IV. Accounting personnel handle financial statement risks through natural language interaction Prerequisite: The current user has permission to access transaction services related to the parameter management of the accounting report system.
[0074] Process description: 1. Users can create a new session in the accounting intelligent assistant client and use natural language to input their requirements for querying the specified report content structure in the accounting knowledge base; 2. The accounting intelligent assistant will submit the prompt information, including information on callable functions (information on all functions provided by the MCP server and functions supported by the accounting intelligent assistant) and user input information to the large language model; 3. The large language model returns the transaction functions and parameter information for querying the accounting knowledge base; 4. The accounting intelligent assistant calls the knowledge base MCP server to query the specified report content structure information; 5. The knowledge base MCP server queries the accounting knowledge base, obtains the specified report content structure information (including the resource name of report data items), and then returns it to the accounting intelligent assistant. 6. The accounting intelligent assistant will return the report content structure information (including the report data item resource name) to the accounting intelligent assistant client; 7. The user refers to the report content structure information (including the report data item resource name) and specifies that a validation formula should be added to the report. The variables in the validation formula are implemented by referencing the report data item resource name; the accounting intelligent assistant client will send the request to the accounting intelligent assistant intelligent agent; 8. The accounting intelligent assistant will submit the prompt information, including information on callable functions (information on all functions provided by the MCP server and functions supported by the accounting intelligent assistant, including information on resources available for operation by the reporting system MCP server) and user input information to the large language model; 9. The large language model returns the functions and parameters for calling the transaction services of the accounting report system parameter management class in the MCP server of the reporting system; 10. The accounting intelligent assistant calls the reporting system's MCP server to add a validation formula to a specified report. After the transaction is successful, the success information is returned to the accounting intelligent assistant client. 11. In the accounting report system, a report verification formula to be processed has been added. It will be put into production after being reviewed and tested by technical personnel. This part will not be described in detail in this invention.
[0075] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.
Claims
1. A platform for implementing intelligent reporting and risk management in accounting, characterized in that, It includes a backend support module, an application service module, an accounting knowledge base, multiple MCP servers, an intelligent agent control module, and a client module; in: The backend support module includes a monitoring system and a large language model; the monitoring system is used to monitor the report data in the accounting report system and output alarm information based on its built-in report risk monitoring model. The application service module includes an accounting system and an accounting reporting system; the accounting system processes accounting data; the accounting reporting system generates reports based on the accounting data. The accounting knowledge base is a vector database that stores accounting knowledge and its metadata. Multiple MCP servers, including: The accounting system MCP server is used to encapsulate and call the service interfaces of the accounting system. The accounting reporting system MCP server is used to encapsulate and call the service interfaces of the accounting reporting system. The monitoring system's MCP server is used to transmit alarm information, query requests, and new reports and risk monitoring models between the monitoring system and the intelligent agent's main control module. The knowledge base MCP server is used to respond to query requests and retrieve information from the accounting knowledge base; The intelligent agent master control module is used to control the conversation flow. The intelligent agent master control module is used to: receive the user's interactive request through natural language input, organize prompt words containing callable function information and submit them to the large language model, receive the operation instructions returned by the large language model, and call the corresponding MCP server according to the operation instructions to perform monitoring, query or handling operations of report risks. The client module provides a user interface for receiving user input, displaying session results, and receiving alarm notifications.
2. The platform for realizing intelligent reporting and risk control in accounting as described in claim 1, characterized in that, The agent's main control module integrates a program execution environment. When the large language model returns automatically generated program code for a target function that does not have a corresponding function interface, the program code is executed in the program execution environment. The program code can call the corresponding MCP server or the functions provided by the agent's main control module.
3. The platform for realizing intelligent reporting and risk control in accounting as described in claim 1, characterized in that, The accounting knowledge base stores content structure description information for various reports; the content structure description information includes semantic descriptions of report data items and their corresponding resource names in the accounting reporting system; The accounting knowledge base stores report content structure description information in JSON format text, which is used to establish a mapping relationship between accounting terms and operable data items in the accounting report system.
4. The platform for realizing intelligent reporting and risk control in accounting as described in claim 3, characterized in that, The initial version of the report content structure description information is automatically generated by the report format file of the accounting report system combined with data dictionary information, and can be updated and stored in the accounting knowledge base after being adjusted by accounting personnel.
5. The platform for realizing intelligent reporting and risk control in accounting as described in claim 1, characterized in that, The monitoring system obtains report data through a monitoring agent deployed on the accounting reporting system side, and supports two data acquisition strategies: timed triggering or triggering by new report generation events in the accounting reporting system.
6. The platform for realizing intelligent reporting and risk control in accounting as described in claim 1, characterized in that, The client module includes an accounting intelligent assistant client and a notification receiving device; the accounting intelligent assistant client obtains and displays alarm information through polling; the notification receiving device receives alarm information via SMS or email.
7. A method for generating a report risk monitoring model based on a platform for realizing intelligent reporting risk control in accounting, as described in any one of claims 1 to 4, characterized in that, Includes the following steps: The client module receives user reports and risk monitoring requests input in natural language. The intelligent agent main control module calls the knowledge base MCP server to obtain the report content structure description information related to the requirements; The intelligent agent main control module organizes the required information, the report content structure description information, and prompt words of the preset program template, and submits them to the large language model; Generate executable report risk monitoring model programs from large language models; The intelligent agent main control module submits the generated report risk monitoring model program to the monitoring system through the monitoring system MCP server.
8. The method for generating a report risk monitoring model according to claim 7, characterized in that, Before submitting the generated model program to the monitoring system, the process also includes: The intelligent agent main control module returns the requirement document and model program generated by the large language model to the client module for user review; Based on the user's review feedback, adjust the input requirement description and repeat the generation process until the user confirms.
9. A method for handling reporting risks based on a platform for intelligent reporting risk management of accounting as described in any one of claims 1 to 4, characterized in that, Includes the following steps: The client module receives risk management requests from users, input in natural language, for specific reports. The intelligent agent main control module calls the knowledge base MCP server to obtain the content structure description information of the specific report; The intelligent agent main control module organizes a prompt word containing the processing request and the resource name of the report data item, submits it to the large language model, and receives the calling instruction returned by the large language model; The intelligent agent main control module calls the parameter management interface of the accounting report system MCP server according to the calling instruction, so as to modify or add data verification rules in the accounting report system.
10. A method for querying report risk information based on a platform for implementing intelligent reporting risk control in accounting, as described in any one of claims 1 to 4, characterized in that, Includes the following steps: The client module receives user-triggered report risk details query requests; The intelligent agent's main control module calls the monitoring system's MCP server to obtain detailed alarm information; Based on the user's further query intent, the intelligent agent main control module calls the large language model to determine the data source to be queried. The large language model returns an instruction to call the accounting statement system MCP server or the accounting system MCP server for querying. The intelligent agent main control module calls the corresponding model context protocol server according to the query command to obtain the associated report data or accounting raw data, and returns it to the client module for display.