Due diligence report generation method

Abstract

The disclosure provides a due diligence report generation method, which comprises the following steps: obtaining a due diligence task instruction through a digital human interaction module, and analyzing the due diligence task instruction to obtain key due diligence parameters of the due diligence task; task planning is performed based on the key due diligence parameters by a scheduling agent to obtain a task execution process of the due diligence task; wherein the task execution process comprises a plurality of subtask nodes; the scheduling agent schedules an execution agent matching a subtask type indicated by each subtask node according to the execution order between the plurality of subtask nodes to execute a corresponding subtask execution strategy, so as to obtain a target task execution result of the due diligence task; and a report generation agent generates a due diligence report of the due diligence task according to the target task execution result. The technical problems of low artificial docking efficiency, large instruction understanding deviation, lack of task allocation pertinence and long report generation period in the traditional due diligence processing process can be effectively solved.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, and in particular to a method for generating due diligence reports. Background Technology

[0002] In some industries (such as banking and insurance), due diligence processes often rely on manual cross-system queries and offline collaboration among multiple roles. However, in practice, staff frequently switch between multiple heterogeneous systems, manually entering keywords to retrieve relevant information about target entities (such as individuals and organizations), making each due diligence session time-consuming. In multi-role collaboration, different roles need to work together via email, meetings, and paper documents, involving numerous approval steps on average. This process is not only lengthy but also prone to information transmission discrepancies. Furthermore, different staff members have significantly different assessment criteria for risk indicators, with subjective judgment having a significant impact. This can lead to a lack of unified quantitative standards for due diligence conclusions, making it difficult to support informed decision-making. Summary of the Invention

[0003] This disclosure aims to at least partially address one of the technical problems in the related art.

[0004] One aspect of this disclosure proposes a due diligence report generation method. This method utilizes a digital human interaction module to acquire and parse due diligence task instructions. It relies on a scheduling agent to scientifically plan the task execution process and intelligently schedule the execution agents, matching corresponding execution agents to perform each sub-task. Finally, a report generation agent outputs a standardized due diligence report. This effectively solves the technical problems of low efficiency due to manual interaction, large errors in instruction understanding, lack of targeted task allocation, and long report generation cycles in traditional due diligence processes. Its refined management mode based on sub-task nodes maintains a high degree of compatibility between execution strategies and sub-task types at each stage, and enables automated and intelligent operation of the entire due diligence process. This improves the execution efficiency and accuracy of due diligence tasks while reducing labor costs and operational error rates. Furthermore, the introduction of the digital human interaction module enhances the convenience and flexibility of instruction interaction, adapting to due diligence needs in different scenarios and improving the versatility and scalability of the due diligence report generation method.

[0005] A first aspect of this disclosure provides a method for generating a due diligence report, the method comprising: The due diligence task instructions are obtained through the digital human interaction module, and the due diligence task instructions are parsed to obtain the key due diligence parameters of the due diligence task. The task execution flow of the due diligence task is obtained by scheduling an intelligent agent to plan the task based on the key due diligence parameters; wherein, the task execution flow includes multiple sub-task nodes, and the sub-task nodes are used to indicate the sub-task type and sub-task execution strategy. The scheduling agent schedules execution agents that match the subtask type indicated by each subtask node to execute the corresponding subtask execution strategy according to the execution order among the multiple subtask nodes, so as to obtain the target task execution result of the due diligence task. The report-generating agent generates a due diligence report for the due diligence task based on the execution results of the target task.

[0006] A second aspect of this disclosure provides a due diligence report generation apparatus, the apparatus comprising: The processing module is used to obtain due diligence task instructions through the digital human interaction module and parse the due diligence task instructions to obtain key due diligence parameters of the due diligence task. The planning module is used to perform task planning based on the key due diligence parameters by a scheduling agent to obtain the task execution flow of the due diligence task; wherein, the task execution flow includes multiple sub-task nodes, and the sub-task nodes are used to indicate the sub-task type and sub-task execution strategy. The first execution module is used to schedule execution agents that match the subtask types indicated by each of the subtask nodes to execute corresponding subtask execution strategies according to the execution order among the multiple subtask nodes, so as to obtain the target task execution result of the due diligence task. The generation module is used to generate a due diligence report for the due diligence task based on the execution result of the target task through the report generation agent.

[0007] A third aspect of this disclosure provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the due diligence report generation method as proposed in the first aspect of this disclosure.

[0008] The fourth aspect of this disclosure provides a non-transitory computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the due diligence report generation method as described in the first aspect of this disclosure.

[0009] A fifth aspect of this disclosure provides a computer program product in which, when instructions are executed by a processor, the due diligence report generation method described in a first aspect of this disclosure is performed.

[0010] The technical solutions provided by the above embodiments of this disclosure bring at least the following beneficial effects: The due diligence task instructions are obtained through a digital human interaction module and parsed to obtain key due diligence parameters. A scheduling agent then plans the task based on these parameters to obtain the task execution flow. This flow includes multiple sub-task nodes, each indicating the sub-task type and execution strategy. The scheduling agent, following the execution order of these sub-task nodes, schedules execution agents matching the sub-task type to execute the corresponding strategies, thus obtaining the target task execution result. Finally, a report generation agent generates a due diligence report based on the target task execution result. Therefore, by using a digital human interaction module to acquire and parse due diligence task instructions, and relying on a scheduling agent to scientifically plan the task execution process and intelligently schedule the execution agents, corresponding execution agents are matched to perform each sub-task. Finally, a report generation agent outputs a standardized due diligence report, effectively solving the technical problems of low efficiency of manual interaction, large deviations in instruction understanding, lack of targeted task allocation, and long report generation cycle in traditional due diligence processes. Its refined management and control mode based on sub-task nodes can maintain a high degree of adaptability between the execution strategies of each link and the sub-task types, and realize the automated and intelligent operation of the entire due diligence process, improving the execution efficiency and accuracy of due diligence tasks, and reducing labor costs and operational error rates. At the same time, the introduction of the digital human interaction module can enhance the convenience and flexibility of instruction interaction, adapt to the due diligence needs in different scenarios, improve the universality and scalability of the due diligence report generation method, and provide reliable technical support for the efficient development of various due diligence businesses.

[0011] Additional aspects and advantages of this disclosure will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this disclosure. Attached Figure Description

[0012] The above and / or additional aspects and advantages of this disclosure will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, in which: Figure 1 A flowchart illustrating a due diligence report generation method provided in an embodiment of this disclosure; Figure 2 A flowchart illustrating another due diligence report generation method provided in this disclosure embodiment; Figure 3 This is a schematic diagram illustrating the execution order between subtask execution nodes provided in an embodiment of this disclosure; Figure 4 This is a flowchart illustrating the due diligence processing method based on multi-agent collaboration provided in an embodiment of this disclosure. Figure 5 This is a schematic diagram of the structure of a due diligence report generation device provided in an embodiment of the present disclosure; Figure 6 This is a schematic diagram of the structure of an electronic device shown in an exemplary embodiment of the present disclosure. Detailed Implementation

[0013] Embodiments of this disclosure are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting this disclosure.

[0014] It should be noted that the acquisition, storage, use, and processing of data in this disclosed technical solution comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0015] It should also be noted that the information (including but not limited to user personal information, transaction information, etc.), data (including but not limited to data used for analysis, data stored, data displayed, etc.) and signals involved in this disclosure are all authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0016] This disclosure proposes a method for generating due diligence reports.

[0017] The due diligence report generation method of this disclosure is described below with reference to the accompanying drawings.

[0018] Figure 1 This is a flowchart illustrating a due diligence report generation method provided in an embodiment of the present disclosure.

[0019] This disclosure illustrates the example of a due diligence report generation method configured in a due diligence report generation device. This due diligence report generation device can be applied to any electronic device so that the electronic device can perform the due diligence report generation function.

[0020] Among them, electronic devices can be any device with computing capabilities, such as personal computers, mobile terminals, servers, etc. Mobile terminals can be hardware devices with various operating systems, touch screens and / or displays, such as mobile phones, tablets, personal digital assistants, wearable devices, etc.

[0021] like Figure 1 As shown, the due diligence report generation method includes the following steps S110 to S130: Step S110: Obtain due diligence task instructions through the digital human interaction module, and parse the due diligence task instructions to obtain the key due diligence parameters of the due diligence task.

[0022] The digital human interaction module can interact with users in a multimodal manner (such as voice, text, visual, and tactile modes) using a human-like approach. Optionally, the digital human interaction module can integrate AI (Artificial Intelligence) vision, voice, and natural language capabilities.

[0023] Among them, the due diligence task instruction can be a specific description of the requirements for the due diligence task (also known as the due diligence task).

[0024] Key due diligence parameters may include, but are not limited to: object attribute information of the object to be due diligence, due diligence type, due diligence constraints, etc.

[0025] The subject of due diligence can be the target entity to be investigated, such as, but not limited to, individuals or enterprises.

[0026] The object attribute information may include the identity information of the object to be investigated (such as name, identity identifier (such as ID card, passport number, employee ID (Identity Document) etc.), location information (such as residential address, work address, etc.). It should be noted that the object attribute information of the object to be investigated can be set and added as needed, and this disclosure does not restrict this.

[0027] The due diligence types can include, but are not limited to, personal housing loan due diligence, consumer loan due diligence, and business loan due diligence. It should be noted that the due diligence report generation method disclosed herein can be applied to scenarios such as financial loan due diligence, pre-employment background checks, insurance underwriting due diligence, and investment compliance due diligence. The due diligence types corresponding to different scenarios can be the same or different; this disclosure does not impose any restrictions on this.

[0028] The due diligence constraints may include time constraints, loan amount constraints, etc., and can be set as needed. This disclosure does not impose any restrictions on them.

[0029] In this embodiment of the disclosure, due diligence task instructions can be obtained through the digital human interaction module, and the due diligence task instructions can be parsed through the digital human interaction module to obtain the key due diligence parameters of the due diligence task.

[0030] As an example, the digital human interaction module may include a digital human interaction interface (also known as a visualization unit), a speech recognition engine (also known as a speech recognition unit), and a natural language processing unit. Suppose a user inputs a due diligence task instruction via the digital human interaction interface in the digital human interaction module: "Conduct due diligence on Zhang San (ID number: 11010XXXXXX) for a housing loan, with a disclosed amount of 1 million yuan and a due diligence period of 1 working day." The speech recognition engine in the digital human interaction module recognizes the acquired speech signal to obtain the instruction text of the due diligence task. Then, the natural language processing unit in the digital human interaction module can parse the instruction text to obtain the key due diligence parameters of the due diligence task, such as name, ID number, due diligence type (e.g., housing loan due diligence), and due diligence constraints (e.g., 1 working day and 1 million yuan loan amount).

[0031] It should be noted that the digital human interactive interface can also support input methods such as text input and gesture input, and this disclosure does not impose any restrictions on this.

[0032] It is understandable that there may be cases where key due diligence parameters obtained through parsing are missing. Therefore, in one possible implementation of this disclosure, when a key due diligence parameter is missing a value, the user can interact with the digital human interaction module to supplement the missing value, thereby obtaining complete key due diligence parameters. As an example, assuming that the address information of the object to be investigated is missing from the key due diligence parameters, the address information of the object to be investigated can be queried using a digital human virtual avatar rendering unit in the digital human interaction module, and the key due diligence parameters can be supplemented through multiple rounds of dialogue. Thus, by using the digital human interaction module to address the problem of missing key due diligence parameters, an efficient and convenient parameter completion mechanism is constructed, effectively avoiding the risk of due diligence tasks stalling, process bottlenecks, or distorted conclusions due to incomplete parameters. Compared to traditional methods like in-person inquiries and text messages, digital human interaction connects with users through natural, human-like dialogue. This lowers the barrier to entry for users to supplement information, aligns with everyday communication habits, and provides real-time responses and guidance to help users complete missing parameters, shortening the parameter completion cycle and avoiding efficiency losses due to information transmission delays. Simultaneously, the digital human can accurately locate key missing due diligence parameters and initiate targeted inquiries, avoiding redundant questions and ensuring the accuracy and completeness of the supplemented information. This lays a solid foundation for subsequent processing based on complete parameter matching and the execution of sub-tasks. Furthermore, this interactive completion mechanism records all communication content and supplementary information, forming a traceable parameter completion chain. This facilitates subsequent quality checks and compliance verification, further enhancing the rigor and standardization of due diligence work, ensuring that due diligence tasks are not interrupted due to missing parameters, and improving the continuity and stability of the entire due diligence process.

[0033] Step S120: The task execution flow of the due diligence task is obtained by scheduling the intelligent agent to perform task planning based on key due diligence parameters; wherein, the task execution flow includes multiple sub-task nodes, and the sub-task nodes are used to indicate the sub-task type and sub-task execution strategy.

[0034] The task execution process can include multiple subtask nodes, and any subtask node can be used to indicate the corresponding subtask type and subtask execution strategy.

[0035] The types of subtasks may include, but are not limited to: data collection subtasks, data preprocessing subtasks, qualification review subtasks, financial analysis tasks, risk assessment subtasks, etc. This disclosure does not impose any restrictions on them.

[0036] As an example, after obtaining key due diligence parameters, the digital human interaction module can send these parameters to the scheduling agent. The scheduling agent can then plan tasks based on these parameters, thereby obtaining the task execution flow for the due diligence task.

[0037] As one possible implementation, when the key due diligence parameters of a due diligence task include the due diligence type, a scheduling agent can retrieve historical execution flows that match (or are identical to) the due diligence type from historical execution data based on the due diligence type in the key due diligence parameters, and then determine the historical execution flow as the task execution flow for that due diligence task. Thus, by reusing the matching historical execution flow based on the due diligence type, a standardized and efficient due diligence process configuration system is constructed, effectively solving the problems of strong subjectivity, repetitive development, and low efficiency in traditional due diligence task process design. On the one hand, the historical execution flows are all formed based on practical experience of similar due diligence tasks and have been verified in actual business scenarios. The scheduling agent directly determines them as the execution flow for the current due diligence task, eliminating the need for manual design of the process framework from scratch and configuration of subtask connection logic. This effectively shortens the preparation cycle before task initiation, reduces the human resource cost and trial-and-error cost of process design, and ensures the rationality and compliance of the process, avoiding process omissions or efficiency losses caused by human design biases. On the other hand, precisely reusing similar historical processes based on due diligence type enables the standardization of execution for similar due diligence tasks. This ensures that different batches and different personnel handling the same type of due diligence tasks follow consistent process specifications and operating guidelines, improving the standardization level and comparability of results in due diligence work. This facilitates subsequent quality control and process optimization for due diligence tasks. Simultaneously, historical execution processes may contain optimal execution paths and accumulated experience for similar tasks. Reusing these processes not only reduces repetitive work but also indirectly reuses efficient strategies from past practices, further improving the efficiency and quality of current due diligence tasks.

[0038] As another possible implementation, when the key due diligence parameters of the due diligence task include the due diligence type, the scheduling agent can query the mapping rule base based on the due diligence type to obtain the processing flow that matches the due diligence type from the mapping rule base, and the processing flow that matches the due diligence type can be determined as the task execution flow of the due diligence task.

[0039] As an example, a correspondence between due diligence types and processing flows can be established in advance and saved in a mapping rule base. Thus, after determining the due diligence type of a due diligence task, the scheduling agent can query the mapping rule base based on the due diligence type to obtain the processing flow that matches the due diligence type from the mapping rule base, and can determine the processing flow that matches the due diligence type as the task execution flow of the due diligence task.

[0040] Therefore, the task execution process can be determined by scheduling intelligent agents to query the mapping rule base based on the due diligence type. The mapping rule base stores the correspondence between due diligence types and processing flows, all of which have been verified and standardized through business practice. This avoids the subjective bias and experience dependence of manually designed processes, and saves the tedious work of building processes from scratch, shortening the preparation cycle before the due diligence task is launched, and reducing labor costs and trial-and-error risks. At the same time, this mechanism can achieve unified adaptation of the execution flow of similar due diligence tasks, ensuring that the same type of due diligence work in different scenarios follows consistent operating procedures and sub-task connection logic, improving the comparability and compliance of due diligence results, and facilitating subsequent quality control and process optimization. In addition, the mapping rule base can flexibly adapt to business iteration needs. When due diligence standards are updated or new due diligence types are added, only the mapping relationship in the library needs to be optimized synchronously to achieve a unified upgrade of the entire system process, without the need to adjust individual tasks one by one, enhancing the scalability and flexibility of the technology.

[0041] In one possible implementation of this disclosure, a scheduling agent can set the task priority of each sub-task node according to the execution order among multiple sub-task nodes.

[0042] In one possible implementation of this disclosure, the deadline for each subtask node can be set by a scheduling agent.

[0043] Step S130: By scheduling the intelligent agent according to the execution order among multiple sub-task nodes, the execution intelligent agent that matches the sub-task type indicated by each sub-task node is scheduled to execute the corresponding sub-task execution strategy in order to obtain the target task execution result of the due diligence task.

[0044] It should be noted that the execution order of multiple subtask nodes can be serial, parallel, or a combination of serial and parallel execution, and this disclosure does not impose any restrictions on this.

[0045] It should also be noted that any execution agent can have a corresponding subtask type, and any execution agent can execute the subtask execution strategy indicated by the subtask node that matches its corresponding subtask type.

[0046] As one possible implementation, such as Figure 2 As shown, step S130 can be achieved using the following steps: Step S131: By scheduling the intelligent agent according to the execution order among multiple subtask nodes, the execution intelligent agent that matches the subtask type indicated by each subtask node is scheduled to execute the corresponding subtask execution strategy, so as to obtain the subtask execution result corresponding to each execution intelligent agent.

[0047] For example, the scheduling agent schedules execution agents that match the subtask types indicated by each subtask node in the execution order among the multiple subtask nodes. Thus, for any given execution agent, that task agent can execute the subtask execution strategy indicated by the subtask node matching its corresponding subtask type, obtaining the subtask execution result corresponding to that execution agent. For instance, suppose the multiple subtask nodes include node 1, node 2, node 3, and node 4, and the execution order among these nodes is as follows: Figure 3 As shown, the output of node 1 serves as the input of node 2, and the output of node 2 can serve as the input of nodes 3 and 4. The scheduling agent can, according to the execution order among the nodes, first schedule the execution agent a that matches the subtask type indicated by node 1. Execution agent a can execute the subtask execution strategy indicated by node 1 to obtain the corresponding subtask execution result and send the corresponding subtask execution result to the scheduling agent. Second, the scheduling agent sends the subtask execution result corresponding to execution agent a to execution agent b that matches the subtask type indicated by node 2. Execution agent b can, based on the subtask execution result corresponding to execution agent a, execute the subtask execution strategy indicated by node 2 to obtain the corresponding subtask execution result and send the corresponding subtask execution result to the scheduling agent. Finally, the scheduling agent can simultaneously send the subtask execution results corresponding to the executing agent b to the executing agent c1 (whose subtask type matches that indicated by node 3) and the executing agent c2 (whose subtask type matches that indicated by node 4). Executing agent c1 can execute the subtask execution strategy indicated by node 3 based on the subtask execution results corresponding to the executing agent b, and executing agent c2 can execute the subtask execution strategy indicated by node 4 based on the subtask execution results corresponding to the executing agent b. Thus, executing agents c1 and c2 can obtain the corresponding subtask execution results and send them to the scheduling agent. Correspondingly, the scheduling agent can receive the subtask execution results sent by executing agents c1 and c2.

[0048] In one possible implementation of this disclosure, when the key due diligence parameters include object attribute information of the object to be due diligence, and the subtask type includes a data acquisition subtask, the subtask execution strategy corresponding to the data acquisition subtask, i.e., the subtask execution strategy under the data acquisition subtask, may include: obtaining multi-dimensional due diligence data as the result of subtask execution from multiple target data systems based on the object attribute information of the object to be due diligence.

[0049] The target data system can refer to a data platform that provides authoritative data in specific dimensions, which needs to be connected to in order to comprehensively evaluate the subject of due diligence. Such platforms may include, but are not limited to, identity verification systems, personal credit reporting platforms, bank statement systems, personal tax systems, public opinion information systems, etc. This disclosure does not impose any restrictions on this.

[0050] As an example, the execution agent corresponding to the data acquisition subtask, such as the data acquisition agent, can, based on the identity information in the object attribute information of the object to be investigated, combined with the data items to be collected corresponding to the data acquisition task, use an interface module and the MCP (Model Context Protocol) service as a standardized connection framework to connect to multiple target data systems and other heterogeneous data sources. Then, the execution agent corresponding to the data acquisition subtask can obtain the data content corresponding to the data items to be collected that match the identity information, and use the data content of these data items to be collected as multi-dimensional due diligence data. Furthermore, the multi-dimensional due diligence data can be used as the execution result of the corresponding subtask.

[0051] It should be noted that any data acquisition subtask under any due diligence type can have a corresponding data item to be collected. As one possible implementation, a correspondence between due diligence types and data items to be collected can be established in advance and saved to a relevant database. When the due diligence type is determined and the task type indicated by the subtask node includes a data acquisition subtask, the corresponding data item to be collected is determined based on the due diligence type.

[0052] Therefore, by acquiring multi-dimensional due diligence data from multiple target data systems based on the object attribute information of the subject under investigation, the accuracy and targeting of data collection are improved. Compared with the indiscriminate data collection methods in traditional due diligence, this method can filter and match corresponding data from target data systems based on the object attribute information of the subject under investigation, effectively filtering redundant and invalid data, improving the efficiency and quality of data collection, and reducing the redundancy cost of data processing. At the same time, the acquisition of multi-dimensional due diligence data covers multiple core aspects of the subject under investigation, which can present the true situation of the subject under investigation in a comprehensive and three-dimensional way, avoiding the one-sidedness and limitations brought by single-dimensional data, and providing a comprehensive and solid data foundation for the subsequent writing of the due diligence report.

[0053] As one possible approach, after acquiring multi-dimensional due diligence data, at least one of the following can be recorded: the data source and the collection time. Retaining the data source information clarifies the data's origin and level of authority, facilitating subsequent tiered assessment of data reliability. Recording the collection time precisely anchors the corresponding time point, adapting to the status analysis needs of different stages of the due diligence object, and providing crucial evidence for dynamic adjustment and retrospective review of due diligence conclusions. By recording at least one key piece of information—the data source and the collection time—each piece of due diligence data is given a traceable and verifiable identity. Compared to traditional due diligence methods that only retain data content without recording the source and time, this design clearly reconstructs the data acquisition path and timeliness, effectively solving the problems of difficulty in verifying data authenticity and defining timeliness, thus enhancing the credibility and compliance of due diligence data.

[0054] Optionally, compliance verification codes for due diligence data can also be generated. As an example, the identity information of the entity to be investigated, the data source, and the collection time can be concatenated according to a preset format to obtain the corresponding compliance verification code. This creates a unique and tamper-proof compliance credential for each piece of due diligence data, strengthening the compliance defenses of the due diligence process from the data source. Compared to traditional due diligence methods that rely solely on manual recording to verify data compliance, this verification code can quickly achieve data traceability and authenticity verification, effectively avoiding the risks of data tampering, forgery, or origination of unknown sources. It enhances the compliance and credibility of due diligence data, perfectly adapting to due diligence scenarios with stringent data compliance requirements, such as those in finance and law.

[0055] As one possible implementation, the target data system can be obtained in the following way: when the key due diligence parameters include the due diligence type, a data acquisition agent determines the target data system that matches the due diligence type from multiple primary data systems based on the due diligence type. Thus, by using a data acquisition agent to accurately select a matching target data system from multiple primary data systems based on the due diligence type in the key due diligence parameters, intelligent and scenario-based adaptation of data sources is achieved. Compared to the traditional method of manually selecting data sources in due diligence, this design can automatically match target data systems with corresponding data dimensions and data authority based on the differentiated needs of due diligence types (such as financial due diligence, compliance due diligence, business due diligence, etc.), effectively avoiding the subjective bias and efficiency bottlenecks in the manual selection process, and improving the accuracy and timeliness of data source selection. Meanwhile, this targeted matching mode based on due diligence type ensures that the acquired due diligence data is highly consistent with the core requirements of the current due diligence task, reducing the amount of irrelevant data collected from the source and lowering the computing power and time costs of subsequent data processing and result fusion. In addition, this mechanism can flexibly adapt to diverse due diligence scenarios. When the due diligence type changes, the data collection agent can quickly adjust the screening strategy of the target data system, enhancing the scenario adaptability and flexible scalability of the solution, and laying a solid data source foundation for the efficient implementation of different types of due diligence tasks.

[0056] In one possible implementation of this disclosure, the subtask type may further include at least one of qualification review subtask, financial analysis subtask, and risk assessment subtask. It should be noted that the subtask nodes in the task execution flow of the due diligence type can also be used to indicate other subtask types, and can be set as needed; this disclosure does not impose any limitations on this.

[0057] Optionally, in any embodiment of this disclosure, the subtask execution strategy under the qualification review subtask may include: conducting qualification review on the subject to be reviewed based on multi-dimensional due diligence data, so as to determine the qualification review result as the result of the corresponding subtask execution.

[0058] The qualification review results can be used, for example, to indicate whether the entity to be investigated is compliant. It should be noted that this disclosure does not limit the scope of information indicated by the qualification review results.

[0059] As an example, the execution agent corresponding to the qualification review sub-task can be a qualification review agent, which can review the identity validity, compliance records (such as no criminal record, no malicious overdue record, etc.), continuity of social security and housing provident fund payments, and completeness of tax declarations of the subject to be reviewed based on multi-dimensional due diligence data, so as to determine the qualification review result as the execution result of the corresponding sub-task based on the review results.

[0060] Optionally, in any embodiment of this disclosure, if the qualification review result indicates non-compliance, the qualification review result can be marked with data items that have not passed the review in the multi-dimensional due diligence data. For example, when the qualification review of the subject to be reviewed is carried out based on multi-dimensional due diligence data, if the subject to be reviewed has a "credit card overdue record", the qualification review result can indicate non-compliance, and the corresponding data items that have not passed the review can be marked with the qualification review result, such as "a credit card overdue record existed in May 2024".

[0061] Optionally, in any embodiment of this disclosure, the subtask execution strategy under the financial analysis subtask may include: determining the indicator values ​​of the subject to be investigated in multiple preset financial indicators based on multi-dimensional due diligence data, and generating a financial status analysis result as the execution result of the corresponding subtask based on the indicator values ​​of the subject to be investigated in multiple preset financial indicators.

[0062] The preset financial indicators can be pre-defined, such as the revenue-to-debt ratio, average monthly surplus ratio, and revenue stability coefficient, etc., and this disclosure does not impose any restrictions on them. The revenue-to-debt ratio can be determined based on the ratio between total monthly liabilities and total monthly revenue; the average monthly surplus ratio can be determined based on the ratio between average monthly surplus and average monthly revenue; and the revenue stability coefficient can be determined by calculating the standard deviation of monthly revenue fluctuations and comparing it with average monthly revenue.

[0063] As an example, the execution agent corresponding to the financial analysis subtask, such as a financial analysis agent, can perform weighted fusion of the indicator values ​​of multiple preset financial indicators on the subject of due diligence to obtain a target score; determine the corresponding financial status based on the score range or score level to which the target score belongs; and generate a financial status analysis result based on the financial status and the indicator values ​​of multiple preset financial indicators. For example, if the financial status is "stable financial status" and the multiple preset financial indicators are "average monthly income of xx yuan and income-to-debt ratio of y", then the financial status analysis result is "average monthly income of xx yuan, income-to-debt ratio of y, stable financial status".

[0064] It should be noted that a correspondence between score ranges or score levels and corresponding financial statuses can be established in advance and saved. Then, after determining the score range or score level to which the target score belongs, the above correspondence can be queried to determine the corresponding financial status.

[0065] As another example, the execution agent corresponding to the financial analysis subtask, such as a financial analysis agent, can employ a recognition model to determine the financial status analysis results of the object under investigation based on the indicator values ​​of multiple preset financial indicators. For instance, the indicator values ​​of the object under investigation at multiple preset financial indicators can be input into the recognition model, and the financial status analysis results of the object under investigation can be obtained in response to the output of the recognition model.

[0066] Optionally, in any embodiment of this disclosure, the subtask execution strategy under the risk assessment subtask may include: analyzing at least one or more combinations of the qualification review results, financial status analysis results, and object public opinion data in the due diligence data to obtain risk assessment information as the result of the corresponding subtask execution.

[0067] The public opinion data of the target may include negative information such as debt disputes and administrative penalties. The public opinion data of the target can be set as needed, and this disclosure does not impose any restrictions on it.

[0068] The risk assessment information may include, for example, risk level and risk point descriptions, which are not limited in this disclosure. Risk point descriptions refer to a written description of the risk characteristics of the entity to be investigated, the conditions that trigger the risk, the potential impact of the risk, and the basis for determining the existence of the risk. It should be noted that risk point descriptions can be positive (e.g., no overdue payments on credit records) or negative (e.g., a debt-to-income ratio exceeding a set threshold (e.g., 45%)), which are not limited in this disclosure.

[0069] As an example, the execution agent corresponding to the risk assessment sub-task, such as the risk assessment agent, can use a logistic regression model to obtain risk assessment information as the execution result of the corresponding sub-task based on the qualification review results, financial status analysis results, and object public opinion data in the due diligence data. For example, the qualification review results, financial status analysis results, and object public opinion data in the due diligence data can be input into the logistic regression model, and the risk assessment information as the execution result of the corresponding sub-task can be obtained in response to the output of the logistic regression model.

[0070] It should be noted that the logistic regression model can be obtained by training the model using historical due diligence data.

[0071] As another example, the execution agent corresponding to the risk assessment sub-task, such as the risk assessment agent, can match the qualification review results, financial status analysis results, and object public opinion data in the due diligence data with expert rules, and determine the risk assessment information as the execution result of the corresponding sub-task based on the matching results.

[0072] Among them, the expert rules can be pre-set. For example, if the income-to-debt ratio in the financial analysis results is not less than a set threshold, it is judged as a high-risk level; or if the qualification review results indicate that the subject of due diligence has a record of dishonesty, it is judged as a high-risk level, and so on. This disclosure does not restrict the setting of expert rules.

[0073] As another example, the execution agent corresponding to the risk assessment sub-task can use a logistic regression model and expert rules to determine the risk assessment information as the execution result of the corresponding sub-task based on the qualification review results, financial status analysis results, and public opinion data of the target in the due diligence data.

[0074] Therefore, the qualification review sub-task relies on multi-dimensional due diligence data for verification, abandoning the review model that depends on single materials or partial information, effectively improving the authenticity and credibility of the qualification review conclusions, and avoiding misjudgments due to incomplete information; the financial analysis sub-task focuses on the quantitative calculation and systematic analysis of preset financial indicators, outputting standardized and comparable financial status conclusions, providing accurate quantitative basis for judging the financial health of the due diligence target, and reducing the subjectivity of financial risk identification; the risk assessment sub-task integrates multi-source information such as qualification review results, financial analysis conclusions, and public opinion data of the target for cross-validation and comprehensive judgment, breaking through the limitations of single-dimensional risk analysis, and can more comprehensively and keenly identify various hidden dangers such as qualification risks, credit risks, and public opinion risks of the due diligence target. In summary, by incorporating qualification review, financial analysis, and risk assessment into the due diligence sub-task system, and designing targeted sub-task execution strategies for each sub-task type, a three-dimensional and full-process due diligence investigation of the due diligence target can be achieved, improving the standardization, comprehensiveness, and accuracy of the due diligence work, thereby improving the accuracy and reliability of the subsequently generated due diligence report.

[0075] It is understandable that after obtaining multi-dimensional due diligence data, data preprocessing can be performed on the multi-dimensional due diligence data. Therefore, in one possible implementation of this disclosure embodiment, the subtask type includes a data preprocessing subtask; correspondingly, the subtask execution strategy under the data preprocessing subtask may include: preprocessing the multi-dimensional due diligence data; wherein, the preprocessing includes at least one of data deduplication, filling missing values, standardization, removing outliers, and desensitization processing.

[0076] As an example, the execution agent corresponding to the data preprocessing subtask, such as the data cleaning and integration agent, can perform the following data preprocessing on the collected multi-dimensional due diligence data: 1. Data deduplication, such as removing duplicate identity information, transaction records, and other data; 2. Fill in missing values. For example, for missing non-critical fields (such as contact information), reserve information can be used to fill in the missing values. For critical fields (such as credit records), the data collection agent can be triggered to re-collect the data. 3. Standardization, such as converting data from different data sources in due diligence into the same date format and monetary unit (e.g., unifying it to "YYYY-MM-DD" date format and "yuan" monetary unit). 4. Remove outliers, such as identifying abnormal transactions in the transaction log using the standard deviation method, marking them and retaining the original data for subsequent verification; 5. De-identification processing, such as de-identifying sensitive data in due diligence data, and displaying the complete data only in authorized scenarios.

[0077] It should be noted that this disclosure does not restrict the preprocessing methods; in practical applications, settings can be made as needed.

[0078] Therefore, by deduplicating data, filling in missing values, and removing outliers, redundant, incomplete, and distorted information in multi-dimensional due diligence data can be effectively filtered, improving the completeness and accuracy of the due diligence data and preventing analytical biases or erroneous conclusions in subsequent sub-tasks due to poor-quality data. Data standardization can convert due diligence data from different sources, in different formats, and with different dimensions into a unified and comparable standard form, eliminating the interference of data dimensional differences on the analysis results and maintaining the scientific rigor and rationality of each subsequent sub-task execution stage. Simultaneously, anonymization can effectively protect sensitive information in the due diligence data without affecting the value of data analysis, meeting both business analysis needs and relevant data security and privacy protection regulations. In summary, by implementing a pre-processing data workflow, the quality of due diligence data is improved from the source, thereby maintaining the reliability, standardization, and security of the entire due diligence process.

[0079] Step S132: By scheduling the agent, the execution results of at least some of the subtasks corresponding to the executing agents are fused to obtain the execution result of the target task.

[0080] As an example, the scheduling agent can fuse the execution results of subtasks corresponding to some of the executing agents to obtain the execution result of the target task. Using the example above, the scheduling agent can fuse the subtask execution results sent by executing agents c1 and c2 to obtain the execution result of the target task.

[0081] As another example, the scheduling agent can merge the subtask execution results of all executing agents corresponding to the task execution flow to obtain the target task execution result. Using the example above, the scheduling agent can merge the subtask execution results sent by executing agents a, b, c1, and c2 to obtain the target task execution result.

[0082] Therefore, by scheduling agents to accurately match and schedule corresponding types of execution agents to execute specific sub-task execution strategies according to the execution order of sub-task nodes, specialized division of labor and targeted control of sub-task execution are achieved. Compared with the traditional model of manually assigning tasks or a single agent handling the entire process, different types of execution agents can focus on their areas of expertise in sub-tasks, effectively avoiding execution deviations caused by mismatched capabilities and improving the accuracy and reliability of each sub-task execution result. At the same time, the scheduling agent performs unified fusion processing on at least some sub-task execution results, breaking down information silos between execution agents, enabling complementary verification and value mining of multi-source heterogeneous data, avoiding the fragmentation problem of scattered results, and making the final output task execution results more complete and consistent. This provides high-quality data support for the generation of subsequent due diligence reports, improving the accuracy and reliability of due diligence report generation.

[0083] It is understandable that data conflicts may occur when fusing the execution results of subtasks corresponding to each executing agent. In this case, in one possible implementation of this disclosure, a preset conflict arbitration rule can be used to arbitrate conflicting data in the execution results of each subtask. For example, assuming the name of the object to be processed in subtask execution result 1 is 'ss' and the name of the object to be processed in subtask execution result 2 is 'aa', and the preset conflict arbitration rule is to determine the name according to the priority level of the executing agents, then the name of the object to be processed corresponding to the executing agent with the higher priority among the executing agents corresponding to each subtask execution result can be determined as the final name of the object to be processed in the target task execution result. It should be noted that the above example of conflict arbitration rule is merely exemplary, and other rules can be used in actual applications, which can be set as needed. Thus, by using the preset conflict arbitration rule to perform targeted arbitration processing on data conflicts in the execution results of each subtask, the contradiction problem of multi-source due diligence data and multi-subtask analysis conclusions can be effectively solved.

[0084] Step S140: The report generation agent generates a due diligence report for the object to be investigated based on the results of the target task execution.

[0085] As one possible implementation, when the key due diligence parameters include the due diligence type, a report generation agent can determine a target report template that matches the due diligence type from multiple preset report templates based on the due diligence type; the report generation agent then fills the target report template with the results of the target task execution to obtain the due diligence report.

[0086] The preset report template can be pre-set, and this disclosure does not restrict the setting of the preset report template.

[0087] As an example, a correspondence between due diligence types and preset report templates can be established in advance. Then, the report generation agent can query the above correspondence based on the due diligence type to determine the report template corresponding to the due diligence type and set the report template as the target report template.

[0088] For example, assuming the due diligence type is personal credit due diligence, the report generation agent can call the personal credit due diligence report template (referred to as the target report template in this disclosure) that matches the personal credit due diligence. The agent will then fill in the qualification review results, compliance records, financial status analysis results, risk assessment information, etc. from the target task execution results into the qualification review, compliance records, financial status, risk assessment, and other chapters in sequence to generate a structured due diligence report in PDF (Portable Document Format) format.

[0089] It should be noted that the above example only uses a PDF format for the due diligence report. In practical applications, the due diligence report can also be in other formats, and this disclosure does not impose any restrictions on this.

[0090] Therefore, the report generation agent can select matching target report templates from a preset template library based on the due diligence type, avoiding the subjective bias of manual template selection. This ensures that reports output from different types of due diligence work conform to business specifications in terms of structure, dimensions, and core elements, improving the uniformity and professionalism of the reports. On the other hand, the report generation agent can automatically fill the corresponding positions in the template with the results of the target task execution, replacing the tedious process of traditional manual data organization and entry. This effectively shortens the report generation cycle, reduces the error rate of manual operation, and improves the overall efficiency of due diligence work. At the same time, the standardized template filling and generation process can ensure the completeness and logical coherence of the report content, ensuring that the core analytical data and conclusions in the due diligence process are presented in the final report without omission or mismatch, providing clear and intuitive data basis for relevant personnel to quickly obtain key information and make scientific decisions.

[0091] In order to improve at least one of the credibility, compliance, and traceability of due diligence reports, in one possible implementation of this disclosure, a report generation agent can mark at least one of the data source (or data origin) and compliance verification code of key data in the due diligence report.

[0092] In one possible implementation of this disclosure, the due diligence report can be displayed through a digital human interaction module. As an example, after receiving the due diligence report, the report-generating agent can send it to the digital human interaction module, which can then display the report on the digital human interaction interface. This visualization approach allows relevant users to intuitively understand the information in the due diligence report, facilitating subsequent processing.

[0093] Optionally, in some embodiments, the user's query task for the due diligence report can be obtained through the digital human interaction module, and the corresponding response information for the query task can be generated through the digital human interaction module.

[0094] The query task can be used to obtain relevant information from the due diligence report. For example, the query task could be "What is the individual's risk level?" It should be noted that the forms of obtaining the query task include, but are not limited to, voice input, text input, touch input, etc., and this disclosure does not impose any restrictions on them.

[0095] The response information can be the answer or feedback provided for the query task.

[0096] As an example, the digital human interaction module can respond to the user's voice input operation on the digital human interaction interface to obtain a query task for the due diligence report. Then, the digital human interaction module can generate the corresponding response information for the query task based on the query task and the due diligence report.

[0097] Therefore, on the one hand, the digital human interaction module, with its human-like interactive style, accepts users' query needs. Compared to traditional text retrieval or manual interaction, it better aligns with users' natural interaction habits, lowers the operational threshold for users to obtain key information from due diligence reports, and improves the friendliness and ease of use of the service. On the other hand, this module can directly extract corresponding data, conclusions, and analytical basis from the generated due diligence report to quickly generate accurate response information, avoiding the tedious process of users searching page by page in lengthy reports and shortening the time cost of information acquisition. In summary, by setting up the digital human interaction module, intelligent reception and accurate feedback on users' due diligence report query tasks are achieved, improving the interactive experience of due diligence services and enhancing information delivery efficiency.

[0098] It is understandable that after the due diligence report is generated, there may be changes to the relevant data of the object to be due diligence in some target data systems. That is, the due diligence data of the object to be due diligence is updated. Therefore, in one possible implementation of this disclosure, when the due diligence data corresponding to the object to be due diligence is updated, a target agent can be determined from multiple execution agents by a scheduling agent; the target agent executes the corresponding sub-task execution strategy based on the updated due diligence data to obtain the updated sub-task execution result; and the due diligence report of the object to be due diligence is updated by the scheduling agent based on the updated sub-task execution result.

[0099] As an example, when the due diligence data corresponding to the object to be due diligence is updated, the scheduling agent can determine the target agent from multiple execution agents based on the changed target data in the due diligence data. For example, the execution agent whose subtask execution process is associated with the target data can be determined as the target agent. Then, the scheduling agent can schedule the target agent so that the target agent executes the corresponding subtask execution strategy based on the updated due diligence data to obtain the updated subtask execution results. Furthermore, the scheduling agent can merge the updated subtask execution results with the subtask execution results that have not been updated among the multiple subtask execution results to update the due diligence report of the object to be due diligence, thus obtaining the updated due diligence report.

[0100] It should be noted that this disclosure does not limit the number of target intelligent agents; there may be one or more.

[0101] Thus, on the one hand, when due diligence data is updated, the scheduling agent accurately selects the target agent without restarting the entire process of sub-tasks. It only executes the corresponding strategy for the links affected by the data update, reducing the repeated computing power consumption after the data update, improving the efficiency and targeting of sub-task iteration, and avoiding resource waste. On the other hand, the target agent executes sub-tasks based on the updated due diligence data, keeps the sub-task results synchronized with the latest data, and then updates the due diligence report through the scheduling agent. This achieves an automated closed loop from data update to report optimization, completely eliminating the cumbersome process of manually tracking data changes and manually correcting reports. It avoids omissions and errors that may occur with manual updates, greatly shortens the report update cycle, and ensures that the due diligence conclusions are always based on the latest and most accurate data. At the same time, this dynamic update mechanism can respond to data changes of the subject of due diligence in real time, enabling the due diligence report to have the ability to continuously iterate. It effectively avoids the limitation of traditional due diligence reports being "fixed as soon as they are issued," and effectively solves the technical problems of static data and insufficient timeliness of reports in traditional due diligence work. It improves the dynamic adaptability and reliability of due diligence services, and can provide dynamic and accurate reference for subsequent decision-making. It is especially suitable for due diligence scenarios with frequent data changes.

[0102] In one possible implementation of this disclosure, a digital human interaction module can broadcast voice prompts to indicate that the due diligence report for the target entity has been updated. Thus, the human-like voice interaction of the digital human can effectively attract the attention of relevant staff, effectively avoiding the problem of missed report updates due to information overload and monotonous notification formats, ensuring that due diligence personnel grasp the report dynamics as soon as possible, and improving the accuracy and timeliness of information delivery. Staff do not need to focus on looking at the screen to receive report update prompts, adapting to diverse work scenarios such as mobile office and multitasking, optimizing the human-computer interaction experience and operational convenience, and effectively reducing the operational threshold and time cost. The digital human interaction module can support customized voice tone and broadcast scripts, generating differentiated prompt content for different report update types for different targets, helping staff quickly distinguish the corresponding due diligence subject and update focus, improving information identification efficiency, and thus enhancing the recognizability and personalization of information transmission. For special scenarios such as visually impaired people and outdoor work, voice prompts can overcome the limitations of visual interaction, ensuring comprehensive delivery of due diligence report update information, that is, achieving barrier-free coverage of information transmission.

[0103] The due diligence report generation method of this disclosure involves obtaining due diligence task instructions through a digital human interaction module and parsing the instructions to obtain key due diligence parameters. A scheduling agent then plans the task based on these parameters to obtain the task execution flow. The task execution flow includes multiple sub-task nodes, each indicating a sub-task type and execution strategy. The scheduling agent, following the execution order of the sub-task nodes, schedules execution agents matching the sub-task type indicated by each node to execute the corresponding sub-task execution strategy, thereby obtaining the target task execution result. Finally, a report generation agent generates a due diligence report based on the target task execution result. Therefore, by using a digital human interaction module to acquire and parse due diligence task instructions, and relying on a scheduling agent to scientifically plan the task execution process and intelligently schedule the execution agents, corresponding execution agents are matched to perform each sub-task. Finally, a report generation agent outputs a standardized due diligence report, effectively solving the technical problems of low efficiency of manual interaction, large deviations in instruction understanding, lack of targeted task allocation, and long report generation cycle in traditional due diligence processes. Its refined management and control mode based on sub-task nodes can maintain a high degree of adaptability between the execution strategies of each link and the sub-task types, and realize the automated and intelligent operation of the entire due diligence process, improving the execution efficiency and accuracy of due diligence tasks, and reducing labor costs and operational error rates. At the same time, the introduction of the digital human interaction module can enhance the convenience and flexibility of instruction interaction, adapt to the due diligence needs in different scenarios, improve the universality and scalability of the due diligence report generation method, and provide reliable technical support for the efficient development of various due diligence businesses.

[0104] To clearly illustrate the due diligence report generation method disclosed herein, a detailed explanation is provided below with examples.

[0105] As an example, the due diligence report generation method of this disclosure is applied to a multi-agent collaborative due diligence processing platform for illustration. The due diligence processing platform may include a digital human interaction module, a collaborative scheduling agent (referred to as a scheduling agent in this disclosure), a data collection agent, a data cleaning and integration agent, a qualification review agent, a financial analysis agent, a risk assessment agent, and a report generation agent, correspondingly, such as... Figure 4 As shown, the following steps may be included: Step 401, Request Reception and Parsing Relevant staff can initiate due diligence requests (referred to as due diligence task instructions in this disclosure) through a digital human interactive interface (supporting voice and text input). For example, "Conduct due diligence on a housing loan for Zhang San (ID number: 11010XXXXXX), with a disclosed amount of xx million yuan and a due diligence period of 1 working day." The digital human interaction module can convert the voice-based due diligence request into text through a voice recognition engine, and then parse it through a natural language processing unit to extract key due diligence parameters, such as the target individual's name, ID number, due diligence type (e.g., housing loan due diligence), and constraints (e.g., 1 working day period, 1 million yuan loan amount matching analysis). If any parameters are found to be missing during the parsing process, such as the household registration address not being clearly stated, the digital human interaction module can use a digital human virtual avatar to ask questions via voice, and complete the missing parameters through multiple rounds of dialogue.

[0106] Step 402, Task Assignment The collaborative scheduling agent can have a built-in personal due diligence task mapping rule base (referred to as the mapping rule base in this disclosure). This rule base can match the corresponding processing flow according to the due diligence type (for example, for personal credit due diligence, financial analysis and credit verification should be emphasized), generate a detailed task list, and reasonably allocate the data collection tasks in the task list to the data collection agent, and allocate the data preprocessing tasks in the task list to the data cleaning and integration agent responsible for data cleaning and integration. In this way, the task allocation for each functional agent is completed sequentially, while setting clear task priorities and deadlines for each task.

[0107] Step 403, Multi-source data acquisition The data collection agent connects to heterogeneous data sources such as identity verification systems (verifying identity authenticity), personal credit reporting platforms, social security and housing provident fund systems, personal tax systems, bank statement systems, public opinion information systems (screening negative public opinion about individuals), and asset certificate filing systems (verifying real estate / vehicle filing information) through the MCP service via the interface module. It collects multi-dimensional data such as identity information, credit records, income data, compliance records, and asset information of the target individual according to the task list. During the collection process, it can automatically record the data source, timestamp, and compliance verification code.

[0108] Step 404, Data Cleaning and Integration The data cleaning and integration agent performs the following processing on the collected multi-dimensional personal data (referred to as due diligence data in this disclosure) to obtain standardized due diligence data: 1. Data deduplication, such as removing duplicate identity information, transaction records, and other data; 2. Fill in missing values. For example, for missing non-critical fields (such as contact information), reserve information can be used to fill in the missing values. For critical fields (such as credit records), the data collection agent can be triggered to re-collect the data. 3. Standardization, such as converting personal data from different data sources into the same date format and monetary unit (e.g., unifying it to "YYYY-MM-DD" date format and "yuan" monetary unit). 4. Remove outliers, such as identifying abnormal transactions in the transaction log using the standard deviation method, marking them and retaining the original data for subsequent verification; 5. De-identification processing, such as de-identifying sensitive data in personal data and displaying the complete data only in authorized scenarios.

[0109] Step 405, Multi-dimensional analysis and processing The qualification review intelligent agent can verify the validity of an individual's identity, compliance records, continuity of social security and housing provident fund payments, and completeness of tax declarations based on standardized due diligence data, and generate a qualification review result of "compliant" or "non-compliant". If it does not comply with the rules, it will mark the specific violation (such as "there is a credit card overdue record in May 2024"). The financial analysis AI can analyze a target individual's bank statements, income statements, and debt records for the past 6-12 months, calculate preset financial indicators such as income-to-debt ratio, average monthly savings rate, and income stability coefficient, fit income fluctuation trends through line charts, compare them with the income level of the same industry, and generate a financial status assessment (such as "average monthly income of 25,000 yuan, income-to-debt ratio of 30%, which is lower than the risk threshold of 45%, and the financial status is stable"). The risk assessment agent can integrate qualification review results, financial analysis results, and personal public opinion data, and use a combination of logistic regression models and expert rules to output risk ratings and risk point descriptions (referred to as risk point description information in this disclosure).

[0110] Step 406, Results Summary and Conflict Arbitration The collaborative scheduling agent receives the processing results from each functional agent. If there is a data conflict, it performs arbitration according to the preset conflict arbitration rules. After determining the final data, it summarizes the results to obtain the comprehensive due diligence result (referred to as the target task execution result in this disclosure).

[0111] Step 407, Report Generation The report generation agent calls a preset personal credit due diligence report template and automatically fills in the content of chapters such as identity verification, compliance records, financial status, and risk assessment to generate a structured, PDF due diligence report. Key data in the due diligence report can be marked with data source and compliance verification code.

[0112] Step 408, Results Display and Interaction The digital human interaction module can display structured due diligence reports on the terminal interface, allowing staff to obtain accurate answers through voice and text queries, and also to drill down into detailed data by clicking on report sections.

[0113] The digital human interaction module can include a virtual avatar rendering unit (such as supporting 2D / 3D virtual avatar customization), a speech recognition unit, a natural language processing unit, and a visualization display unit. It supports access from PC and mobile terminals, and all interaction processes are automatically logged to meet compliance audit requirements.

[0114] Step 408, Dynamic Update The collaborative scheduling agent monitors various heterogeneous data sources in real time. If changes are detected in the data of a target individual, the relevant functional agent is triggered to re-execute the corresponding task, and the report generation agent can update the report synchronously.

[0115] This disclosure proposes a due diligence processing method and platform based on multi-agent collaboration. The platform uses a digital human as the entry point for natural interaction and multi-agent collaboration as the core processing mechanism, comprehensively covering all stages of the due diligence process: "requirement initiation - data collection - data processing - multi-dimensional analysis - report generation - result feedback." Through collaborative work among agents and automated processing, the platform significantly reduces repetitive manual labor and effectively improves the efficiency of due diligence. Simultaneously, the multi-agent collaboration model not only reduces information loss and human resource consumption during transmission but also effectively avoids errors that may be caused by human factors. Furthermore, the natural interaction of the digital human lowers the user's difficulty of use, possesses the ability to resolve fuzzy requirements, and provides personalized query services. The platform also supports dynamic data updates and real-time report refreshes, ensuring that due diligence results accurately reflect the latest risk status of relevant entities, effectively mitigating decision-making risks caused by "outdated information," and improving the timeliness, accuracy, and effectiveness of generated reports. In terms of platform design, a plug-in and modular architecture is adopted, supporting the addition of new agents and expansion of data sources, exhibiting extremely high scalability.

[0116] To implement the above-described methods for generating due diligence reports, this disclosure also provides an apparatus for generating due diligence reports.

[0117] Figure 5 This is a schematic diagram of a due diligence report generation device provided in an embodiment of the present disclosure.

[0118] like Figure 5 As shown, the due diligence report generation device 500 includes: a processing module 501, a planning module 502, a first execution module 503, and a generation module 504.

[0119] The processing module 501 is used to obtain due diligence task instructions through the digital human interaction module and parse the due diligence task instructions to obtain the key due diligence parameters of the due diligence task.

[0120] The planning module 502 is used to plan tasks based on key due diligence parameters by scheduling an intelligent agent to obtain the task execution flow of the due diligence task; wherein, the task execution flow includes multiple sub-task nodes, and the sub-task nodes are used to indicate the sub-task type and sub-task execution strategy.

[0121] The first execution module 503 is used to schedule execution agents that match the subtask types indicated by each subtask node to execute corresponding subtask execution strategies according to the execution order among multiple subtask nodes, so as to obtain the target task execution result of the due diligence task.

[0122] The generation module 504 is used to generate a due diligence report for the due diligence task based on the execution results of the target task through a report generation agent.

[0123] In one possible implementation of this disclosure, the first execution module 503 is configured to: schedule execution agents that match the subtask type indicated by each subtask node to execute corresponding subtask execution strategies according to the execution order among multiple subtask nodes, so as to obtain the subtask execution results corresponding to each execution agent; and fuse the subtask execution results corresponding to at least some of the execution agents through the scheduling agent to obtain the target task execution result.

[0124] In one possible implementation of this disclosure, the key due diligence parameters include object attribute information of the object to be due diligence, and the subtask type includes a data acquisition subtask; the subtask execution strategy under the data acquisition subtask includes: based on the object attribute information of the object to be due diligence, obtaining multi-dimensional due diligence data as the result of subtask execution from multiple target data systems.

[0125] In one possible implementation of this disclosure, the subtask type further includes at least one of qualification review subtask, financial analysis subtask, and risk assessment subtask; The subtask execution strategies under the qualification review subtask include: Based on multi-dimensional due diligence data, the qualifications of the subject of due diligence are reviewed to determine the qualification review results as the execution results of the corresponding sub-tasks; The subtask execution strategies under the financial analysis subtask include: Based on multi-dimensional due diligence data, the indicator values ​​of the subject to due diligence are determined for multiple preset financial indicators, and based on the indicator values ​​of the subject to due diligence for multiple preset financial indicators, financial status analysis results are generated as the execution results of the corresponding sub-tasks. The subtask execution strategies under the risk assessment subtask include: Analyze at least one or more combinations of the qualification review results, financial status analysis results, and public opinion data of the target in the due diligence data to obtain risk assessment information as the result of the execution of the corresponding sub-task.

[0126] In one possible implementation of this disclosure, the due diligence report generation apparatus 500 may further include: The determination module is used to determine the target agent from multiple execution agents by scheduling agents in response to an update event of the due diligence data corresponding to the object to be due diligence.

[0127] The second execution module is used to execute the corresponding sub-task execution strategy based on the updated due diligence data by the target intelligent agent, and obtain the updated sub-task execution results; The update module is used to update the due diligence report of the object to be investigated by scheduling the intelligent agent based on the updated subtask execution results.

[0128] In one possible implementation of this disclosure, the due diligence report generation apparatus 500 may further include: The broadcast module is used to broadcast voice prompts through the digital human interaction module to indicate that the due diligence report of the subject to be investigated has been updated.

[0129] In one possible implementation of this disclosure, the subtask type includes a data preprocessing subtask; the subtask execution strategy under the data preprocessing subtask includes: preprocessing multi-dimensional due diligence data; wherein, the preprocessing includes at least one of data deduplication, filling missing values, standardization, removing outliers, and desensitization.

[0130] In one possible implementation of this disclosure, the key due diligence parameters include the due diligence type; the generation module 504 is configured to: determine a target report template that matches the due diligence type from multiple preset report templates based on the due diligence type using a report generation agent; and fill the target report template with the target task execution result using the report generation agent to obtain a due diligence report.

[0131] In one possible implementation of this disclosure, the due diligence report generation apparatus 500 may further include: The supplementary module is used to respond to missing values ​​in key due diligence parameters by interacting with the user through the digital human interaction module to supplement information on the missing values ​​in the key due diligence parameters.

[0132] In one possible implementation of this disclosure, the key due diligence parameters include the due diligence type; the planning module 502 is used to: query the mapping rule base based on the due diligence type by scheduling the intelligent agent to obtain the processing flow that matches the due diligence type; and determine the processing flow as the task execution flow of the due diligence task.

[0133] It should be noted that the due diligence report generation apparatus provided in this embodiment can achieve the above-mentioned... Figures 1 to 3 All method steps implemented in the method embodiment can achieve the same technical effect. Therefore, the parts that are the same as those in the method embodiment and their beneficial effects will not be described in detail here.

[0134] To implement the above embodiments, this disclosure also proposes an electronic device, wherein the electronic device can be any device with computing capabilities, the electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the due diligence report generation method proposed in the first aspect of the present disclosure.

[0135] As an example, Figure 6 This is a schematic diagram of the structure of an electronic device 600 as shown in an exemplary embodiment of this disclosure, as follows: Figure 6 As shown, the aforementioned electronic device 600 may further include: The memory 610 and processor 620 are connected by a bus 630, which connects different components (including the memory 610 and the processor 620). The memory 610 stores a computer program, which, when executed by the processor 620, implements the due diligence report generation method described in this embodiment of the present disclosure.

[0136] Bus 630 represents one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. For example, these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0137] Electronic device 600 typically includes a variety of electronic device readable media. These media can be any available media that can be accessed by electronic device 600, including volatile and non-volatile media, removable and non-removable media.

[0138] Memory 610 may also include computer system readable media in the form of volatile memory, such as random access memory (RAM) 640 and / or cache memory 650. Server 600 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 660 may be used to read and write non-removable, non-volatile magnetic media (… Figure 6 Not shown; usually referred to as a "hard drive"). Although Figure 6 As not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 630 via one or more data media interfaces. Memory 610 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of this disclosure.

[0139] A program / utility 680 having a set (at least one) of program modules 670 may be stored in, for example, memory 610. Such program modules 670 include—but are not limited to—an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 670 typically perform the functions and / or methods described in the embodiments of this disclosure.

[0140] Electronic device 600 can also communicate with one or more external devices 690 (e.g., keyboard, pointing device, display 691, etc.), and with one or more devices that enable a user to interact with electronic device 600, and / or with any device that enables electronic device 600 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed via input / output (I / O) interface 692. Furthermore, electronic device 600 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 693. As shown, network adapter 693 communicates with other modules of electronic device 600 via bus 630. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0141] The processor 620 executes various functional applications and data processing by running programs stored in the memory 610.

[0142] It should be noted that the implementation process and technical principles of the electronic device in this embodiment are explained in the foregoing description of the due diligence report generation method of this disclosure embodiment, and will not be repeated here.

[0143] To implement the above embodiments, this disclosure also proposes a non-transitory computer-readable storage medium storing a computer program that, when executed by a processor, implements the due diligence report generation method proposed in any of the foregoing embodiments of this disclosure.

[0144] To implement the above embodiments, this disclosure also proposes a computer program product that, when the instructions in the computer program product are executed by a processor, performs the due diligence report generation method proposed in any of the foregoing embodiments of this disclosure.

[0145] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0146] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this disclosure, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0147] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of preferred embodiments of this disclosure includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of this disclosure pertain.

[0148] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0149] It should be understood that various parts of this disclosure can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0150] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0151] Furthermore, the functional units in the various embodiments of this disclosure can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0152] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of the present disclosure have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present disclosure. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present disclosure.

Claims

1. A method for generating a due diligence report, characterized in that, The method includes: The due diligence task instructions are obtained through the digital human interaction module, and the due diligence task instructions are parsed to obtain the key due diligence parameters of the due diligence task. The task execution flow of the due diligence task is obtained by scheduling an intelligent agent to plan the task based on the key due diligence parameters; wherein, the task execution flow includes multiple sub-task nodes, and the sub-task nodes are used to indicate the sub-task type and sub-task execution strategy. The scheduling agent schedules execution agents that match the subtask type indicated by each subtask node to execute the corresponding subtask execution strategy according to the execution order among the multiple subtask nodes, so as to obtain the target task execution result of the due diligence task. The report generation agent generates a due diligence report for the due diligence task based on the execution results of the target task.

2. The method according to claim 1, characterized in that, The step of scheduling the execution agent according to the execution order among the multiple sub-task nodes, and scheduling the execution agent that matches the sub-task type indicated by each sub-task node to execute the corresponding sub-task execution strategy to obtain the target task execution result of the due diligence task, includes: The scheduling agent schedules execution agents that match the subtask type indicated by each subtask node to execute the corresponding subtask execution strategy according to the execution order among the multiple subtask nodes, so as to obtain the subtask execution result corresponding to each execution agent. The target task execution result is obtained by fusing the execution results of at least some of the subtasks corresponding to the execution agents through the scheduling agent.

3. The method according to claim 2, characterized in that, The key due diligence parameters include object attribute information of the object to be due diligenceed, and the subtask type includes data collection subtask; The subtask execution strategy under the data acquisition subtask includes: Based on the object attribute information of the object to be investigated, multi-dimensional due diligence data is obtained from multiple target data systems as the execution result of the subtask.

4. The method according to claim 3, characterized in that, The sub-task types also include at least one of the following: qualification review sub-task, financial analysis sub-task, and risk assessment sub-task; The subtask execution strategy under the qualification review subtask includes: Based on the multi-dimensional due diligence data, the qualifications of the subject to be due diligence are reviewed to determine the qualification review results as the execution results of the corresponding sub-tasks; The subtask execution strategies under the financial analysis subtask include: Based on the multi-dimensional due diligence data, the indicator values ​​of the subject to due diligence in multiple preset financial indicators are determined, and financial status analysis results are generated as the execution results of the corresponding sub-tasks based on the indicator values ​​of the subject to due diligence in multiple preset financial indicators. The subtask execution strategies under the risk assessment subtask include: Analyze at least one or more combinations of the qualification review results, the financial status analysis results, and the object public opinion data in the due diligence data to obtain risk assessment information as the result of the corresponding sub-task execution.

5. The method according to claim 3, characterized in that, After the report-generating agent generates a due diligence report for the due diligence task based on the execution result of the target task, the method further includes: In response to an update event of the due diligence data corresponding to the object to be due diligence, the target agent is determined from among the multiple execution agents by the scheduling agent; The target agent executes the corresponding subtask execution strategy based on the updated due diligence data to obtain the updated subtask execution results. The scheduling agent updates the due diligence report of the object to be investigated based on the updated subtask execution results.

6. The method according to claim 5, characterized in that, The method further includes: The digital human interaction module broadcasts a voice prompt to indicate that the due diligence report for the subject of due diligence has been updated.

7. The method according to claim 3, characterized in that, The subtask types include data preprocessing subtasks; The subtask execution strategy under the data preprocessing subtask includes: The multi-dimensional due diligence data is preprocessed; wherein the preprocessing includes at least one of the following: data deduplication, missing value filling, standardization, outlier removal, and desensitization.

8. The method according to any one of claims 1-7, characterized in that, The key due diligence parameters include the due diligence type; the process of generating a due diligence report for the due diligence task based on the execution result of the target task by the report generation agent includes: The report generation agent determines a target report template that matches the due diligence type from multiple preset report templates based on the due diligence type. The report generation agent fills the target report template with the execution results of the target task to obtain the due diligence report.

9. The method according to any one of claims 1-7, characterized in that, After parsing the due diligence task instructions to obtain key due diligence parameters, the method further includes: In response to the presence of missing values ​​in the key due diligence parameters, the system interacts with the user through the digital human interaction module to supplement the missing information in the key due diligence parameters.

10. The method according to any one of claims 1-7, characterized in that, The key due diligence parameters include the due diligence type; the task execution flow of the due diligence task is obtained by scheduling an intelligent agent based on the key due diligence parameters, including: The scheduling agent queries the mapping rule base based on the due diligence type to obtain a processing flow that matches the due diligence type. The processing flow is defined as the task execution flow for the due diligence task.

Images