Account receivable processing method and device, electronic device and storage medium

Abstract

The invention relates to an account receivable processing method and device, an electronic device and a storage medium, and the method comprises the steps: recognizing contract data of account receivable, carrying out the correlation query based on the contract data, and obtaining the heterogeneous business data of the account receivable; constructing an intelligent agent driven by the large language model, and configuring a plurality of functional tools for the intelligent agent; the function tool is used for executing query, analysis and result sending operations on the heterogeneous business data; receiving a target task for account receivable input by a user, performing task planning on the target task by using the intelligent agent, and judging and calling a function tool to execute the target task based on a task planning result; wherein the intelligent agent is iteratively called according to the execution result of the target task until it is judged that the current task execution result meets the preset task requirement of the target task, and the final task execution result is output. Task planning, tool calling and feedback are achieved through the intelligent agent, and the problem that the account receivable information state is lagged is solved.

Description

Technical Field

[0001] This application relates to the field of language large model intelligent agent technology, and in particular to accounts receivable processing methods, apparatus, electronic devices and storage media. Background Technology

[0002] As businesses continue to expand and credit sales models become more widespread, accounts receivable are increasingly becoming a significant portion of current assets. Their management efficiency and risk control capabilities have become crucial factors affecting cash flow security and stable operations. However, the widely used traditional manual ledgers or basic financial software still struggle to achieve real-time tracking and systematic management of accounts receivable status, leading to slow collection responses, longer collection cycles, and consequently, increased pressure on working capital.

[0003] Currently, various business systems in related technologies often operate independently, lacking effective data integration and forming "data silos." This not only makes it difficult for non-financial personnel to quickly obtain customized accounts receivable analysis reports across systems, but also prevents companies from monitoring accounts receivable collection progress in real time. Due to the lack of coordination and delayed information acquisition in existing technology systems, it is difficult to track the status of accounts receivable in real time, resulting in a problem of lagging accounts receivable information status.

[0004] There is currently no effective solution to the problem of delayed accounts receivable information status in related technologies. Summary of the Invention

[0005] This embodiment provides an accounts receivable processing method, apparatus, electronic device, and storage medium to solve the problem of delayed accounts receivable information status in related technologies.

[0006] Firstly, this embodiment provides an accounts receivable processing method, including:

[0007] Identify the contract data of accounts receivable, perform correlation queries based on the contract data, and obtain the heterogeneous business data of the accounts receivable;

[0008] Construct an intelligent agent driven by a large language model, and configure multiple functional tools for the intelligent agent; the functional tools are used to perform query, analysis, and result sending operations on the heterogeneous business data;

[0009] The system receives user input regarding a target task for the accounts receivable, utilizes the intelligent agent to plan the task, and, based on the result of the task planning, determines and invokes the functional tools to execute the target task. The system iteratively invokes the intelligent agent according to the execution result of the target task until it is determined that the current task execution result meets the preset task requirements of the target task, and then outputs the final task execution result. The target task includes proactive early warning and on-demand interactive querying of the accounts receivable.

[0010] In some embodiments, the accounts receivable contract data includes contract number and accounts receivable period; the heterogeneous business data includes accounts receivable data, invoicing data, and work order data.

[0011] In some embodiments, identifying the contract data for accounts receivable, querying based on the association of the contract data, and obtaining heterogeneous business data of the accounts receivable include:

[0012] The contract data is identified by optical character recognition to obtain the contract number and payment period of the accounts receivable;

[0013] Based on pre-set database connection credentials, a connection is established with the heterogeneous business data to integrate the business detail data of all the heterogeneous business data; the database connection credentials include the database Internet Protocol address, port, account, and password.

[0014] In some embodiments, the functional tools include a TEXT2SQL tool, a data query tool, an accounts receivable analysis tool, and a targeted delivery tool. The TEXT2SQL tool uses natural language to analyze database tables, retrieves relevant database schema information and relationships through retrieval enhancement generation technology, and generates corresponding structured query statements. The data query tool executes the structured query statements and returns query results from multiple heterogeneous business data sets. The accounts receivable analysis tool uses a large language model to analyze the query results and generate analysis conclusions containing targeted suggestions. The targeted delivery tool sends the analysis conclusions or warning information to designated users.

[0015] In some embodiments, receiving user input regarding a target task for the accounts receivable, utilizing the intelligent agent to plan the target task, determining and invoking the functional tool to execute the target task based on the result of the task planning; wherein, based on the execution result of the target task, the intelligent agent is iteratively invoked until it is determined that the current task execution result meets the preset task requirements of the target task, and the final task execution result is output; wherein, the target task includes proactive early warning and on-demand interactive query for the accounts receivable, including:

[0016] The task planning is performed by the intelligent agent. For each step of the task planning, the intelligent agent determines whether it is necessary to call the functional tool. If so, the target functional tool corresponding to the target task is determined from all the functional tools and called.

[0017] In some embodiments, the method further includes:

[0018] By setting a preset interval, the intelligent agent is automatically invoked periodically to push information about accounts receivable that are about to become overdue or have already become overdue to the business processing party.

[0019] In some embodiments, the target task input by the user is in the form of a query statement in natural language; the result of the target task output by the agent is in the form of a query result in natural language format.

[0020] Secondly, this embodiment provides an accounts receivable processing device, including: a preparation module, a construction module, and a result module; wherein:

[0021] The preparation module is used to identify the contract data of accounts receivable, perform related queries based on the contract data, and obtain the heterogeneous business data of the accounts receivable.

[0022] The construction module is used to build an intelligent agent driven by a large language model and configure multiple functional tools for the intelligent agent; the functional tools are used to perform query, analysis and result sending operations on the heterogeneous business data;

[0023] The result module receives user input regarding a target task for the accounts receivable, utilizes the intelligent agent to plan the target task, and, based on the result of the task planning, determines and invokes the functional tools to execute the target task. Specifically, based on the execution result of the target task, the intelligent agent is iteratively invoked until it is determined that the current task execution result meets the preset task requirements of the target task, and the final task execution result is output. The target task includes proactive early warning and on-demand interactive querying of the accounts receivable.

[0024] Thirdly, this embodiment provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the accounts receivable processing method described in the first aspect above.

[0025] Fourthly, this embodiment provides a storage medium storing a computer program that, when executed by a processor, implements the accounts receivable processing method described in the first aspect above.

[0026] Compared with related technologies, this embodiment provides an accounts receivable processing method, apparatus, electronic device, and storage medium. In this method, firstly, contract data for accounts receivable is identified, and related queries are performed based on the contract data to obtain heterogeneous business data of accounts receivable. Secondly, an intelligent agent driven by a large language model is constructed, and multiple functional tools are configured for the intelligent agent. These functional tools are used to perform queries, analysis, and result transmission operations on the heterogeneous business data. Finally, user input of a target task for accounts receivable is received, and the intelligent agent performs task planning for the target task. Based on the result of the task planning, the functional tools are called to execute the target task. The intelligent agent is iteratively called according to the execution result of the target task until it is determined that the current task execution result meets the preset task requirements of the target task, and the final task execution result is output. The target task includes proactive early warning of accounts receivable and on-demand interactive queries. It can achieve periodic proactive early warning of accounts receivable through a cyclical process of task planning, intelligent agent functional tool calls, and result feedback, combined with task feedback to continuously optimize the large model's prompts, thereby solving the problem of delayed accounts receivable information status.

[0027] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects and advantages of this application more readily apparent. Attached Figure Description

[0028] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0029] Figure 1 This is a hardware structure block diagram of a terminal for an accounts receivable processing method according to an embodiment of this application;

[0030] Figure 2 This is a flowchart of an accounts receivable processing method according to an embodiment of this application;

[0031] Figure 3 This is a flowchart of an accounts receivable processing method according to another embodiment of this application;

[0032] Figure 4 This is a flowchart of an accounts receivable processing method according to another embodiment of this application;

[0033] Figure 5 This is a structural block diagram of the accounts receivable processing device in this embodiment. Detailed Implementation

[0034] To better understand the purpose, technical solution, and advantages of this application, the application is described and illustrated below in conjunction with the accompanying drawings and embodiments.

[0035] Unless otherwise defined, the technical or scientific terms used in this application shall have the general meaning understood by one of ordinary skill in the art to which this application pertains. Words such as “a,” “an,” “an,” “the,” “the,” and “these” used in this application do not indicate quantitative limitation and may be singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or modules (units) is not limited to the listed steps or modules (units) but may include steps or modules (units) not listed, or may include other steps or modules (units) inherent to these processes, methods, products, or devices. Words such as “connected,” “linked,” and “coupled” used in this application are not limited to physical or mechanical connections but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. Normally, the character " / " indicates that the objects before and after it are in an "or" relationship. The terms "first," "second," "third," etc., used in this application are merely to distinguish similar objects and do not represent a specific order of objects.

[0036] The method embodiments provided in this example can be executed in a terminal, computer, or similar electronic device with a certain computing power. For example, it can run on a terminal. Figure 1 This is a hardware structure block diagram of a terminal for an accounts receivable processing method according to an embodiment of this application. For example... Figure 1 As shown, a terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 and a memory 104 for storing data are also included. The processor 102 may be, but is not limited to, a microprocessor (MCU) or a programmable logic device (FPGA). The terminal may also include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that… Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the terminal described above. For example, the terminal may also include components that are larger than... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown are illustrated.

[0037] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the accounts receivable processing method in this embodiment. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0038] The transmission device 106 is used to receive or send data via a network. This network includes a wireless network provided by the terminal's communication provider. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 can be a Radio Frequency (RF) module used for wireless communication with the Internet.

[0039] This embodiment provides a method for processing accounts receivable. Figure 2 This is a flowchart of an accounts receivable processing method according to an embodiment of this application, such as... Figure 2 As shown, the process includes the following steps:

[0040] Step S210: Identify the contract data of accounts receivable, perform correlation queries based on the contract data, and obtain heterogeneous business data of accounts receivable.

[0041] Intelligent accounts receivable processing begins with deep digitization of contracts. Besides the amount and customer name, a contract may contain information such as the legal names and unique social credit codes of both parties, the total contract price, installment payment arrangements, agreed payment terms, and the conditions for the commencement of those terms. Among all this data, the contract number, customer identification document (ID), accurate amount receivable, and the conditions and number of days for the commencement of payment terms constitute key contract data. Quickly and accurately obtaining this key contract data improves the accuracy and efficiency of subsequent accounts receivable processing.

[0042] However, transforming this information from static contract text into structured data that can be processed by computers requires effective recognition technology. Traditional manual data entry methods are inefficient and error-prone, making them unsuitable for managing massive amounts of contracts. Currently, the technological means of recognizing contract data have evolved from traditional manual entry to multi-layered, automated intelligent solutions. Structured parsing technology can directly parse digitally generated and standardized contracts. For example, by processing the underlying code of the document or using regular expressions to accurately match text with fixed patterns of amounts and dates, efficient and low-error information extraction can be achieved. Image recognition and conversion technology is key to processing unstructured contracts such as scanned documents and images. Optical Character Recognition (OCR) technology is a core technology in the field of document recognition, capable of recognizing text images in images and converting them into computer-editable and processable text.

[0043] After successfully identifying contract data, it is correlated with heterogeneous business data. This correlation is established using key identifiers such as the contract number and customer ID from the contract data, based on corresponding database connection credentials. After correlating the heterogeneous business data, multiple other business systems, including customer relationship management systems, enterprise resource planning systems, logistics systems, and external credit reporting systems, can be connected via Application Programming Interface (API) technology. This transforms a single accounts receivable record into a multi-dimensional risk and value view. The correlation between contract data and heterogeneous business data enables early identification and assessment of credit risk, providing end-to-end information support for reconciliation and collection, and constructing a complete and reliable chain of business evidence from contract to accounts receivable information.

[0044] Step S220: Construct an intelligent agent driven by a large language model and configure multiple functional tools for the intelligent agent; the functional tools are used to perform query, analysis and result sending operations on heterogeneous business data.

[0045] Large language models are a type of artificial intelligence with powerful language understanding and generation capabilities. They can understand complex questions posed by users in natural language, much like a human. A large language model-driven agent is an autonomous task execution system with the large language model as its "brain." Its core capabilities lie in planning task steps, deciding which functional tools to use, and understanding the results returned by those tools. In this step, to enable the agent to actually operate business systems, a variety of functional tools are configured for it. These functional tools essentially encapsulate standardized interfaces for specific business capabilities. Currently, general-purpose large models with strong reasoning capabilities, such as Generative Pretrained Transformer 4 (GPT-4) or Claude 3, can be used. These models perform excellently in understanding complex business logic and handling multi-step tasks. In addition, industry-specific large models trained with financial expertise can be utilized. These models have a better understanding of financial terminology and business processes, and can meet the query needs of accounts receivable agents.

[0046] For example, when a user submits a request, the intelligent agent autonomously plans the task steps. First, it calls a query tool to obtain data, then uses an analysis and calculation tool to perform data analysis, and finally sends the analysis results to the designated customer via a push notification tool. It is by configuring functional tools for intelligent agents driven by large language models that an intelligent hub that understands intent, allocates resources, and drives business decisions can be built.

[0047] Step S230: Receive user input regarding the target task for accounts receivable, utilize an intelligent agent to plan the target task, and based on the result of the task planning, determine and invoke functional tools to execute the target task; wherein, based on the execution result of the target task, iteratively invoke the intelligent agent until it is determined that the current task execution result meets the preset task requirements of the target task, and output the final task execution result; wherein, the target task includes proactive early warning and on-demand interactive query for accounts receivable.

[0048] Upon receiving a user's target task expressed in natural language, the agent does not simply treat it as a keyword query. Instead, it leverages its powerful reasoning capabilities to break down and analyze this ambiguous task instruction, transforming it into a structured, executable task plan. Subsequently, based on this plan, the agent autonomously determines the functional tools to be invoked at each step and issues precise instructions for execution.

[0049] Crucially, the agent doesn't passively accept the initial execution result, but rather engages in an iterative optimization cycle of evaluation. Pre-defined task requirements serve as objective standards for judging the quality of the execution result. These requirements can include: completeness requirements—whether the result covers all explicit and implicit dimensions in the task instructions. For example, querying account status should include not only the amount and payment period, but also core status (such as whether it's overdue) and key risk level attributes. Depth requirements—whether the result transcends a simple data listing to provide interpretive or relational insights. For example, does the result merely list numbers, or does it further relate to the key business events that led to those numbers? Actionability requirements—whether the result provides direct evidence for decision-making. For example, the result of an early warning task should not only be a risk label, but should include specific warning reasons and preliminary handling suggestions to make it actionable. Based on these requirements, the agent automatically judges the result of each task execution. If the result fails to meet the user's task requirements, an iterative process is triggered until the output fully meets the pre-defined user task requirements.

[0050] The intelligent agent automatically determines the outcome of each task execution based on preset requirements for completeness, depth, and actionability. This determination process is not a simple rule matching, but rather involves multi-dimensional semantic analysis of the results using a well-trained large language model to identify knowledge gaps, logical loopholes, or missing information. When the model determines that the current result does not fully meet the task requirements, the agent automatically triggers an iterative optimization process. This process relies on continuous prompting and tuning of the large language model based on massive amounts of real-world business scenario examples and corresponding return results.

[0051] In the prompt optimization phase of the planning stage, we injected the thinking framework of domain experts and typical task decomposition patterns, enabling the model to generate more systematic and comprehensive task decomposition schemes. For example, for complex tasks such as accounts receivable analysis, the optimized prompt can guide the model to automatically identify and analyze multiple dimensions, including aging analysis, bad debt assessment, and customer credit association, rather than simply summarizing data. Regarding the optimization of functional tool calling logic, we built a functional tool description library and a calling case library to train the agent to understand the applicable scenarios and output characteristics of each functional tool. This training allows the agent to accurately select the most appropriate combination of functional tools when faced with specific sub-tasks, much like an experienced business professional. Particularly important is the enhanced training of the reflection mechanism. We designed structured reflection templates to transform abstract task requirements into specific, machine-executable check items. These check items include information element integrity verification, data correlation analysis, and conclusion support sufficiency assessment. After extensive training with real-world cases, the model can autonomously identify various defects in the results.

[0052] This continuous optimization mechanism based on practical feedback ensures that the agent's performance improves with increased usage frequency. Furthermore, the large model records successes and failures in each iteration, forming a self-evolving training data pool. Through regular model retraining and prompt-based fine-tuning, the agent gradually evolves from a simple task executor into an intelligent assistant with domain insight capabilities. Ultimately, the output, validated through multiple iterations, not only ensures the completeness and accuracy of information but also provides in-depth insights with business value. For example, in accounts receivable management scenarios, the agent outputs no longer a simple list of accounts receivable but a comprehensive solution including aging structure analysis, recovery probability prediction, risk level assessment, and targeted collection suggestions. This achieves a qualitative leap from data presentation to decision support, helping business personnel quickly grasp core issues and take effective action.

[0053] Compared to related technologies, the independent operation of various business systems such as contracts, sales, finance, and logistics creates data silos, preventing the real-time flow and synchronization of core accounts receivable information. This information fragmentation leaves enterprises in a passive, reactive state of monitoring accounts receivable, rather than proactively managing them. Specifically, there is a delay in the entire process, from the fulfillment of contractual payment conditions to the confirmation of performance by the logistics system, and then to the generation and status update of accounts receivable in the financial system. Therefore, related technologies struggle to track the status of accounts receivable in real time, resulting in a lag in accounts receivable information.

[0054] Steps S210 to S230 above involve: First, identifying contract data for accounts receivable, performing correlation queries based on the contract data, and obtaining heterogeneous business data for accounts receivable; second, constructing an intelligent agent driven by a large language model and configuring multiple functional tools for the agent; these tools are used to perform queries, analysis, and result sending operations on the heterogeneous business data; subsequently, receiving user input for a target task related to accounts receivable, utilizing the intelligent agent to plan the target task, judging based on the task planning results, and calling the functional tools to execute the target task; iteratively calling the intelligent agent based on the execution results of the target task until it is determined that the current task execution result meets the preset task requirements of the target task, and outputting the final task execution result; the target tasks include proactive early warning of accounts receivable and on-demand interactive queries. This system, through a cyclical process of task planning, intelligent agent functional tool calls, and result feedback, combined with task feedback to continuously optimize the large model's prompts, enables periodic proactive early warnings for accounts receivable, thus solving the problem of delayed accounts receivable information status.

[0055] Optionally, in one embodiment, the contract data for accounts receivable includes the contract number and the accounts receivable period; the heterogeneous business data includes accounts receivable data, invoicing data, and work order data.

[0056] The accounts receivable contract data includes contract numbers and accounts receivable periods. By using key identifiers such as contract numbers and customer IDs, a cross-system data network can be built. This network allows data fragments originally scattered across various business systems to be integrated into a complete business view. The accounts receivable period clarifies the time dimension of accounts receivable, including key time elements such as the start date, period length, and due date, providing a time benchmark for accounts receivable status analysis. Based on this fundamental contract data, multi-level data connections can be established to obtain heterogeneous business data from different accounts receivable transactions.

[0057] Heterogeneous business data includes accounts receivable data, invoicing data, and work order data. Accounts receivable data records the amount and details of accounts receivable, invoicing data records invoice information including invoice issuance status, invoice amount, and invoicing time, and work order data records the business details of contract performance, including specific progress, resource input, and acceptance status.

[0058] Furthermore, in one embodiment, identifying contract data for accounts receivable, querying based on contract data associations, and obtaining heterogeneous business data for accounts receivable includes:

[0059] By recognizing contract data using optical character recognition, the contract number and payment period of accounts receivable are obtained; based on pre-set database connection credentials, a connection is established with heterogeneous business data to integrate the business details of all heterogeneous business data; the database connection credentials include the database Internet Protocol address, port, account, and password.

[0060] OCR is a technology that converts text information in an image into computer-editable and searchable coded text. First, the acquired contract image of accounts receivable is preprocessed, including grayscale conversion, binarization, noise reduction, and layout analysis to optimize image quality and locate text regions. Then, character segmentation and feature extraction are performed, splitting text lines into individual characters and extracting key features such as shape and contour. Finally, a pre-trained recognition model is used to compare and classify the features, outputting the final recognized text. Post-processing rules (such as dictionary-based correction) can further improve recognition accuracy. OCR technology can extract key contract data, such as contract numbers and accounts receivable terms, from complex and varied contract documents.

[0061] Based on pre-set database connection credentials, connections are established with various heterogeneous business data to form a contract business database. Database connection credentials are a set of addressing parameters used for connecting contract data and heterogeneous business data. The core elements of a database connection credential include: a database Internet Protocol address (IPA address) used to locate the target database server on the network; a port used to specify the specific communication channel for the database service; and an authentication account and corresponding password used to verify access permissions and ensure data security. By establishing connections between contract data and heterogeneous business data, unified access and integration of accounts receivable-related business details can be achieved.

[0062] In one embodiment, the functional tools include a TEXT2SQL tool, a data query tool, an accounts receivable analysis tool, and a targeted delivery tool. The TEXT2SQL tool uses natural language to analyze database tables, retrieves relevant database schema information and relationships through retrieval enhancement generation technology, and generates corresponding structured query statements. The data query tool executes the structured query statements and returns query results from multiple heterogeneous business data. The accounts receivable analysis tool uses a large language model to analyze the query results and generate analysis conclusions containing targeted suggestions. The targeted delivery tool sends the analysis conclusions or warning information to designated users.

[0063] Specifically, through tools such as TEXT2SQL, data query tools, accounts receivable analysis tools, and targeted delivery tools, a closed loop can be formed, from understanding task requirements to performing task analysis and then delivering task results.

[0064] The TEXT2SQL tool enables intelligent translation between the agent and the large model language database. It accurately converts the user's query intent, expressed in natural language, into a structured query language statement that the large model language database can understand and execute. To achieve this, the TEXT2SQL tool employs a retrieval-enhanced generation technology framework. For example, when receiving a natural language command such as "Query details of accounts receivable for customer A that are over 90 days old," the TEXT2SQL tool first performs a vectorized retrieval from the enterprise's data dictionary and metadata database, recalling contract business database schema information related to concepts such as "customer," "accounts receivable," and "age," including specific table names, field names, inter-table relationships, and business constraints. Subsequently, in the generation phase, the large language model constructs a syntactically correct and semantically accurate structured query language (SQL) statement based on this contextual information.

[0065] After receiving the SQL generated by the TEXT2SQL tool, the data query tool performs data queries in the contract business database through database connection credentials. The core capability of the data query tool lies in its ability to distribute query requests to multiple business system data sources in parallel, efficiently collecting and merging the returned result sets. For example, in a single query, the data query tool can simultaneously retrieve account amounts from the financial database, customer credit ratings from the customer relationship management system, and performance status from the project management system, ultimately integrating these scattered data into a unified view.

[0066] Accounts receivable analysis tools are based on data obtained from TEXT2SQL tools and data query tools. Leveraging the powerful reasoning and inductive capabilities of large language models, they transform fragmented data points into meaningful analytical conclusions. These tools can not only identify abnormal patterns (such as an abnormal increase in the concentration of amounts within a specific aging period), but also perform attribution analysis based on business knowledge and generate actionable recommendations.

[0067] Targeted delivery tools can automatically send specific analytical conclusions or early warning information to designated users or user groups through internal enterprise email systems, instant messaging tools, or business workbenches, based on preset rules (such as risk level, amount threshold, or responsible personnel). For example, when the system identifies a significant risk, the targeted delivery tool will immediately send an early warning report containing detailed analysis and handling recommendations to designated users, including relevant business leaders and financial risk control specialists. All functional tools are uniformly scheduled through an intelligent agent, forming a collaborative whole.

[0068] In one embodiment, user input is received regarding a target task related to accounts receivable. An intelligent agent performs task planning for the target task, and based on the task planning results, functional tools are invoked to execute the target task. The intelligent agent is iteratively invoked according to the execution result of the target task until it is determined that the current task execution result meets the preset task requirements, and the final task execution result is output. The target task includes proactive early warning and on-demand interactive querying of accounts receivable, including:

[0069] Task planning is performed by an intelligent agent. For each step of the task planning, the intelligent agent determines whether it is necessary to call a functional tool. If so, it selects the target functional tool corresponding to the target task from all functional tools and calls it.

[0070] Specifically, in the task planning phase, the agent does not simply forward instructions. Instead, upon receiving a target task posed by the user in natural language, the agent leverages its reasoning abilities to deeply deconstruct the task. The agent comprehensively considers the task's complexity, data dependencies, and best practices to generate a step-by-step, executable task blueprint.

[0071] Following this, the crucial functional tool identification and invocation phase begins. For each specific step in the task planning, the agent autonomously determines its execution method. By identifying which steps require interaction with external systems or specialized calculations, the agent invokes the appropriate functional tools. This identification process is not based on simple keyword matching, but rather on the agent's deep understanding of the tool's functionality. The agent maintains a functional tool knowledge base, defining the capability boundaries, input parameters, and output format of each tool. By matching the semantic description of the specific steps of the current task with the functional tool knowledge base, the agent can select the most suitable target functional tool from all available tools and generate invocation instructions that conform to its interface specifications.

[0072] Ultimately, the functional tools undergo iterative execution and optimization. This entire process is a dynamic, iterative optimization process. The agent is not satisfied with a one-time execution result. After receiving the results returned by the functional tools, the agent automatically evaluates the current results based on preset task requirements. If the results are found to be unsatisfactory, the agent automatically initiates a new cycle from planning to judgment to invocation. This iterative mechanism ensures that the output from the agent is no longer a crude intermediate product, but a truly expected task result.

[0073] For example, for the accounts receivable overdue warning task, the intelligent agent first plans the following steps: (1) query all accounts receivable details due next month; (2) obtain the historical payment behavior and credit status of the customers corresponding to these accounts; (3) perform risk rating according to preset rules; (4) generate a warning report and send it to the relevant person in charge. The intelligent agent performs intelligent scheduling for each step based on the functional tool knowledge base. First, the TEXT2SQL tool converts natural language instructions into structured query statements; then, the data query tool is called to query the customer's business database through customer ID to query historical transaction records, payment habits and other data, and calls the external credit scoring interface to obtain the latest credit rating information as needed; further, the accounts receivable analysis tool is called to use the large language model to comprehensively analyze the multi-dimensional data obtained in the first two steps, to estimate the expected risk of customer accounts receivable, and automatically generate a warning report containing a list of high-risk accounts, risk cause analysis and handling suggestions; finally, the targeted sending tool is called to send the report to the relevant business person in charge via email or communication tools according to the preset recipient rules.

[0074] In addition, in one embodiment, the above accounts receivable processing method may further include: periodically and automatically calling the intelligent agent at preset intervals to push information about accounts receivable that are about to be overdue or have already been overdue to the business processing party.

[0075] Specifically, the system proactively initiates accounts receivable data analysis tasks at a preset frequency. Once invoked, the intelligent agent autonomously executes a risk assessment process. First, it retrieves the latest status of all outstanding accounts using data query tools. Second, utilizing accounts receivable analysis tools, based on preset overdue thresholds, aging distribution, and customer credit changes, it accurately identifies accounts that are about to become overdue (e.g., due within 3 days) or have already become overdue. Finally, it generates a structured early warning report containing a list of high-risk accounts, risk level assessments, and suggested countermeasures. Based on this, a targeted delivery tool is activated, sending the early warning information to the corresponding business processing parties according to preset push rules. This periodic proactive push mechanism transforms traditional post-overdue collection into pre-overdue early warning and monitoring, effectively reducing risk response time and ensuring that business teams can obtain key risk information immediately, enabling abnormal accounts to be promptly identified, allocated, and followed up.

[0076] In one embodiment, the target task input by the user is in the form of a query statement in natural language; the result of the target task output by the agent is in the form of query results in natural language format.

[0077] Users input their target tasks using natural language query statements. This approach allows business personnel without technical expertise to interact with the system using everyday language, without needing to understand the underlying data structures and query syntax. Correspondingly, the agent outputs task results in natural language query format, ensuring intuitiveness and readability. Specifically, the agent doesn't simply return raw data tables; instead, it performs in-depth analysis and structured organization of the query results, generating logically structured natural language reports.

[0078] For example, a user could enter the query: "Please find accounts receivable that are overdue for more than 90 days and amount to more than 500,000 yuan, and analyze which customers they are mainly concentrated in." In response to this query, the agent might return: "Upon query, there are 15 accounts receivable that meet the criteria, totaling 28.5 million yuan. These accounts are mainly concentrated in three customers: Company A (3 accounts, 8.2 million yuan), Group B (5 accounts, 12.6 million yuan), and C Shares (2 accounts, 4.3 million yuan). Among them, Group B has the largest proportion of overdue amounts, and its historical payment records show 3 overdue instances in the past six months; it is recommended to prioritize its handling."

[0079] Figure 3 This is a flowchart of an accounts receivable processing method according to another embodiment of this application, as shown below. Figure 3 As shown, the process includes the following steps:

[0080] Step S310, Data Preparation and Environment Configuration. Integrate and standardize all relevant business data, including core business data and connection credential configuration. Core business data includes contract data, accounts receivable data, invoicing data, and work order data. Database connection credentials include IP address, port, account, and password.

[0081] Step S320: Build and encapsulate the functional tools required by the intelligent agent. These tools include a data query tool, a TEXT2SQL tool, an accounts receivable analysis tool, and a targeted sending tool, providing flexible push capabilities and supporting the sending of analysis results or early warning information to designated users or user groups via email, system notifications, etc.

[0082] Step S330: Construct and optimize the accounts receivable agent. The agent, built around a large language model, will be optimized using expected question-and-answer scenarios related to accounts receivable. This optimization may include adjustments to planning capabilities, functional tool invocation capabilities, and reflective capabilities.

[0083] Step S340: The trained agent is deployed in practical applications. The trained agent is capable of proactively alerting to overdue accounts receivable and providing on-demand interactive queries.

[0084] Figure 4 This is a flowchart of an accounts receivable processing method according to another embodiment of this application. Figure 4 As shown, the workflow of this accounts receivable processing method begins with the top-level intelligent agent, which acts as the core controller, driving the entire task from start to finish. The agent first plans the task and then schedules the corresponding functional tools at the lower level to execute specific operations. The functional tool layer contains four core modules: when information needs to be queried from the database, the agent calls the TEXT2SQL tool to parse the instructions and then passes them to the data query tool for execution; after acquiring the data, the accounts receivable analysis tool is invoked for in-depth calculations and risk analysis; finally, the targeted sending tool is responsible for outputting the analysis results in the form of alarms. All these tools operate on the underlying data foundation, which uniformly manages accounts receivable data, invoicing data, work order data, and contract data. Contract data is identified and entered using OCR technology, ensuring the accuracy and usability of the original data. The agent determines whether to call a functional tool and which tool to call based on the target task of sending alarms or querying accounts receivable, and outputs the final result in natural language.

[0085] Figure 5 This is a structural block diagram of the accounts receivable processing device 50 in this embodiment, as shown below. Figure 5 As shown, the accounts receivable processing device 50 includes a preparation module 52, a construction module 54, and a result module 56. The preparation module 52 identifies the contract data of accounts receivable, performs correlation queries based on the contract data, and obtains heterogeneous business data of accounts receivable. The construction module 54 constructs an intelligent agent driven by a large language model and configures multiple functional tools for the intelligent agent. These functional tools are used to perform queries, analysis, and result sending operations on the heterogeneous business data. The result module 56 receives user input regarding the target task for accounts receivable, uses the intelligent agent to plan the target task, judges and calls the functional tools to execute the target task based on the result of the task planning, and iteratively calls the functional tools according to the execution result of the target task until it is determined that the current task execution result meets the preset task requirements of the target task, and outputs the final task execution result. The target task includes proactive early warning and on-demand interactive querying of accounts receivable.

[0086] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0087] This embodiment also provides an electronic device including a memory and a processor, the memory storing a computer program and the processor being configured to run the computer program to perform the steps in any of the above method embodiments.

[0088] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0089] Optionally, in this embodiment, the processor can be configured to perform the following steps via a computer program:

[0090] S1 identifies contract data for accounts receivable, performs correlation queries based on the contract data, and obtains heterogeneous business data for accounts receivable.

[0091] S2 constructs an intelligent agent driven by a large language model and configures multiple functional tools for the intelligent agent; the functional tools are used to perform query, analysis and result sending operations on heterogeneous business data;

[0092] S3 receives user input regarding the target task for accounts receivable, utilizes an intelligent agent to plan the target task, and judges and calls functional tools to execute the target task based on the result of the task planning. Among them, the functional tools are iteratively called according to the execution result of the target task until it is determined that the current task execution result meets the preset task requirements of the target task, and the final task execution result is output. Among them, the target task includes proactive early warning of accounts receivable and on-demand interactive query.

[0093] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated in this embodiment.

[0094] Furthermore, in conjunction with the accounts receivable processing methods provided in the above embodiments, this embodiment can also provide a storage medium for implementation. The storage medium stores a computer program; when executed by a processor, the computer program implements any of the accounts receivable processing methods in the above embodiments.

[0095] It should be understood that the specific embodiments described herein are merely illustrative of the application and not intended to limit it. All other embodiments derived by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0096] Obviously, the accompanying drawings are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar situations based on these drawings without any creative effort. Furthermore, it is understood that although the work done in this development process may be complex and lengthy, for those skilled in the art, certain design, manufacturing, or production modifications made based on the technical content disclosed in this application are merely conventional technical means and should not be considered as insufficient disclosure of this application.

[0097] The term "embodiment" in this application refers to a specific feature, structure, or characteristic described in connection with an embodiment that may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily imply the same embodiment, nor does it imply that it is mutually exclusive with or independent of other embodiments. It will be clearly or implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0098] The above embodiments merely illustrate several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of patent protection. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the appended claims.

Claims

1. A method for processing accounts receivable, characterized in that, include: Identify the contract data of accounts receivable, perform correlation queries based on the contract data, and obtain the heterogeneous business data of the accounts receivable; Construct an intelligent agent driven by a large language model, and configure multiple functional tools for the intelligent agent; the functional tools are used to perform query, analysis, and result sending operations on the heterogeneous business data; The system receives user input regarding a target task for the accounts receivable, utilizes the intelligent agent to plan the task, and, based on the result of the task planning, determines and invokes the functional tools to execute the target task. The system iteratively invokes the intelligent agent according to the execution result of the target task until it is determined that the current task execution result meets the preset task requirements of the target task, and then outputs the final task execution result. The target task includes proactive early warning and on-demand interactive querying of the accounts receivable.

2. The accounts receivable processing method according to claim 1, characterized in that, The accounts receivable contract data includes contract number and accounts receivable period; the heterogeneous business data includes accounts receivable data, invoicing data, and work order data.

3. The accounts receivable processing method according to claim 2, characterized in that, The process of identifying accounts receivable contract data, querying based on the association of the contract data, and obtaining heterogeneous business data of the accounts receivable includes: The contract data is identified by optical character recognition to obtain the contract number and payment period of the accounts receivable; Based on pre-set database connection credentials, a connection is established with the heterogeneous business data to integrate the business detail data of all the heterogeneous business data; the database connection credentials include the database Internet Protocol address, port, account, and password.

4. The accounts receivable processing method according to claim 1, characterized in that, The functional tools include a TEXT2SQL tool, a data query tool, an accounts receivable analysis tool, and a targeted delivery tool; the TEXT2SQL tool uses natural language to analyze database tables, retrieves relevant database schema information and relationships through retrieval enhancement generation technology, and generates corresponding structured query statements; the data query tool executes the structured query statements and returns query results from multiple heterogeneous business data; The accounts receivable analysis tool uses a large language model to analyze the query results and generate analytical conclusions that include targeted suggestions; The targeted sending tool sends the analysis conclusions or early warning information to designated users.

5. The accounts receivable processing method according to claim 1, characterized in that, The process involves receiving user input regarding a target task related to the accounts receivable, utilizing the intelligent agent to plan the task, and determining and invoking the functional tools to execute the target task based on the task planning results. Specifically, the intelligent agent is iteratively invoked based on the execution result of the target task until it is determined that the current task execution result meets the preset task requirements of the target task, and the final task execution result is output. The target task includes proactive early warning and on-demand interactive querying of the accounts receivable, including: The task planning is performed by the intelligent agent. For each step of the task planning, the intelligent agent determines whether it is necessary to call the functional tool. If so, the target functional tool corresponding to the target task is determined from all the functional tools and called.

6. The accounts receivable processing method according to claim 1, characterized in that, The method further includes: By setting a preset interval, the intelligent agent is automatically invoked periodically to push information about accounts receivable that are about to become overdue or have already become overdue to the business processing party.

7. The accounts receivable processing method according to claim 1, characterized in that, The target task input by the user is in the form of a query statement in natural language; the result of the target task output by the agent is in the form of a query result in natural language format.

8. An accounts receivable processing device, characterized in that, include: The module consists of a preparation module, a construction module, and a results module; among which: The preparation module is used to identify the contract data of accounts receivable, perform correlation queries based on the contract data, and obtain the heterogeneous business data of the accounts receivable. The construction module is used to build an intelligent agent driven by a large language model and configure multiple functional tools for the intelligent agent; the functional tools are used to perform query, analysis and result sending operations on the heterogeneous business data; The result module receives user input regarding a target task for the accounts receivable, utilizes the intelligent agent to plan the target task, and, based on the result of the task planning, determines and invokes the functional tools to execute the target task. Specifically, based on the execution result of the target task, the intelligent agent is iteratively invoked until it is determined that the current task execution result meets the preset task requirements of the target task, and the final task execution result is output. The target task includes proactive early warning and on-demand interactive querying of the accounts receivable.

9. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the accounts receivable processing method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the accounts receivable processing method according to any one of claims 1 to 7.

Images