An e-commerce customer service cross-domain context enhancement action closed-loop processing method and system

Abstract

This invention discloses a method and system for cross-domain context-enhanced action closed-loop processing in e-commerce customer service, relating to the field of intelligent customer service technology in e-commerce. The method includes: determining a unified customer identity identifier based on the mapping between external contact identifiers in instant messaging and user identifiers in the business system; based on the unified identity identifier, acquiring multi-source business data through read-only bypass connections and read-only interfaces of the business system to construct a structured business context; inputting customer dialogue history, business context, and customer service operation intent into an artificial intelligence model for analysis to generate standard action objects; issuing a unified short-term token after customer service confirmation, invoking the corresponding business system to execute the operation, and implementing single-path control for operations with external side effects; and writing the execution results and follow-up matters back to the customer service task system to form a closed-loop processing flow. This invention achieves one-stop intelligent response and automatic processing across business systems in e-commerce customer service, improving customer service processing efficiency and business execution accuracy.

Description

Technical Field

[0001] This invention relates to the field of intelligent customer service technology in e-commerce, and in particular to a method and system for closed-loop processing of cross-domain context-enhanced actions in e-commerce customer service. Background Technology

[0002] With the rapid development of the e-commerce industry, online customer service has become a crucial service entry point for e-commerce platforms to connect customers with their business systems. During a single customer interaction, customer service personnel typically need to handle various tasks, including order inquiries, address changes, logistics tracking, refunds and returns, coupon reissues, payment processing, and order arrival tracking. These tasks rely on multiple independent business systems, such as order systems, logistics systems, procurement systems, marketing systems, refund systems, and Enterprise Resource Planning (ERP) systems, to complete a full customer service process. Therefore, customer service representatives need to frequently switch between these systems to complete a single customer service interaction.

[0003] Currently, some e-commerce platforms have introduced AI-powered customer service, generating responses through knowledge-based question-and-answer systems or large language models to improve customer service efficiency. However, existing AI customer service typically relies solely on customer dialogue text for reasoning, lacking the ability to acquire real-time business data. It cannot unify the association between instant messaging identities and user identities in business systems such as orders, logistics, refunds, and marketing, nor can it obtain real-time business context such as order status, logistics milestones, refund amounts, and purchase arrival information. This results in generated responses lacking specificity and failing to directly complete business processing.

[0004] On the other hand, even if the AI ​​model can generate corresponding processing suggestions, customer service still needs to manually log in to the corresponding business system to complete business operations such as refunds, address changes, coupon issuance, and payment. The AI ​​output and business execution are independent of each other, making it impossible to form a closed-loop processing flow from intelligent judgment to business execution. At the same time, each business system uses different identity authentication and permission management methods, making it difficult to achieve unified authorization, unified auditing, and unified risk control for cross-system business calls. If business write interfaces are directly opened, it is easy to bypass the business verification mechanism, increasing business risks. In addition, when multiple AI models participate in business judgment at the same time, without a unified execution control mechanism, duplicate execution problems with external side effects, such as duplicate refunds, duplicate order creation, and duplicate payment, may occur.

[0005] Therefore, existing technologies are insufficient to achieve integrated processing of unified customer identity association, real-time business context acquisition across business systems, AI-generated executable action suggestions, unified authentication and execution across systems, and closed-loop write-back of business results. This makes it difficult to meet the application requirements of e-commerce customer service to complete intelligent response and automatic handling across multiple business systems within a single workbench. Summary of the Invention

[0006] In view of the above-mentioned problems, this invention proposes a cross-domain context-enhanced action closed-loop processing method and system for e-commerce customer service to solve the problems of fragmented cross-system business data, lack of real-time business context in AI responses, and inability to automatically execute intelligent suggestions in existing e-commerce customer service. It realizes one-stop automated customer service processing based on unified identity recognition, business context enhancement, intelligent action generation, and secure cross-system execution.

[0007] To address the aforementioned technical problems, the present invention provides the following technical solution: Firstly, embodiments of the present invention provide a method for closed-loop processing of cross-domain context-enhanced actions in e-commerce customer service, comprising: receiving a customer message; determining the customer's identity based on the mapping relationship between an external contact identifier in instant messaging and a user identifier in a business system; obtaining a corresponding unified customer identity identifier; based on the unified identity identifier, concurrently acquiring business data corresponding to the customer from multiple heterogeneous business systems through a combination of read-only bypass connections and read-only interfaces in the business system, and constructing a structured business context; inputting the customer's dialogue history, the structured business context, and the customer service operation intent into multiple artificial intelligence models for concurrent analysis, and generating candidate responses. The system includes the content and corresponding standard action objects, whereby the standard action objects at least include action type, target business system, target document identifier, action parameters, and risk level; responds to customer service confirmation of the standard action objects, issues a unified short-term token for cross-business system calls based on the confirmed standard action objects, and routes the confirmed standard action objects to the corresponding business systems according to the action type; the corresponding business systems perform corresponding business operations after authentication based on the unified short-term token, and only allow the single path corresponding to the confirmed standard action objects to execute business operations with external side effects; obtains the execution results of the business operations, and writes the execution results and follow-up matters back to the customer service task system.

[0008] As a preferred embodiment of the cross-domain context enhancement action closed-loop processing method for e-commerce customer service described in this invention, the following steps are taken: receiving a customer message, determining the customer's identity based on the mapping relationship between the instant messaging external contact identifier and the business system user identifier, and obtaining the corresponding unified customer identity identifier, including: obtaining the instant messaging external contact identifier corresponding to the customer message, and querying the corresponding business system user identifier according to a pre-established identity mapping relationship table; when a unique corresponding business system user identifier is found, the business system user identifier is determined as the customer's unified identity identifier; when there is no corresponding mapping relationship or multiple candidate mapping relationships exist, mapping verification is performed based on the customer's historical session records, order association information, and contact information association information, and the verification result is written into the identity mapping relationship table to update the mapping relationship between the customer's unified identity identifier and the instant messaging external contact identifier.

[0009] As a preferred embodiment of the cross-domain context enhancement action closed-loop processing method for e-commerce customer service described in this invention, the method involves: concurrently acquiring business data corresponding to the customer from multiple heterogeneous business systems through a combination of read-only bypass connections and read-only interfaces of business systems, and constructing a structured business context. This includes: establishing read-only bypass connections corresponding to the order system, logistics system, procurement system, marketing system, refund system, ERP system, and exception handling system; calling the read-only interfaces of each business system to concurrently acquire business data corresponding to the unified identity identifier of the customer; establishing a unified field mapping relationship based on the data types returned by each business system; standardizing the fields of data from different business systems; and associating and assembling the data according to a preset context data structure to generate a structured business context. The structured business context includes at least a customer identity field, an order status field, a logistics node field, a refund executable field, a coupon available field, a purchase arrival field, and an exception status field.

[0010] For the data structure consistency detection obtained by the read-only bypass connection, when a change in the business data structure is detected, the corresponding data parsing rules are updated according to the field mapping relationship, and the structured business context is reconstructed.

[0011] As a preferred embodiment of the cross-domain context-enhanced action closed-loop processing method for e-commerce customer service described in this invention, the method involves: inputting customer dialogue history, the structured business context, and customer service operation intent into multiple artificial intelligence models for concurrent analysis to generate candidate response content and corresponding standard action objects. This includes: encoding customer dialogue history, the structured business context, and customer service operation intent into model input data, and sending them in parallel to multiple artificial intelligence models for inference to obtain multiple candidate response results; using a lightweight intent parsing model to perform semantic parsing on the customer dialogue content, extracting at least one business parameter from order number, recipient, delivery address, refund amount, and product information, and fusing the business parameter with customer dialogue history, structured business context, and customer service operation intent to form model input data for concurrent analysis by the multiple artificial intelligence models; and outputting data to the multiple artificial intelligence models. The candidate response results are used to extract the action type, target business system, and target document identifier, and business consistency is calculated in conjunction with the structured business context. If the consistency of multiple candidate response results meets a preset threshold, the corresponding candidate response result is determined as the target candidate response result. If the consistency of multiple candidate response results does not meet the preset threshold, the model input data is reconstructed based on the order status field, logistics node field, refund executable field, and abnormal status field in the business context, and sent to the artificial intelligence model for further analysis. The candidate response results output by each artificial intelligence model are uniformly parsed, and corresponding standard action objects are generated according to the preset action object description specifications. The standard action object includes at least the action type, target business system, target document identifier, action parameters, risk level, and candidate response text. Action routing information is generated based on the action type and target business system of the standard action object.

[0012] As a preferred embodiment of the cross-domain context-enhanced action closed-loop processing method for e-commerce customer service described in this invention, the method includes: responding to the customer service's confirmation operation on the standard action object; issuing a unified short-term token for cross-business system calls based on the confirmed standard action object; and routing the confirmed standard action object to the corresponding business system according to the action type. This includes: responding to the customer service's confirmation instruction on the standard action object; performing action parameter integrity verification and risk level verification on the standard action object; determining the standard action object as an action to be executed when the verification passes; generating a unified short-term token based on the action to be executed, the unified short-term token including at least an operator identifier, authorized action type, target business system, target document identifier, expiration time, and unique execution identifier; determining the target business system interface to be called based on the pre-established mapping relationship between action types and business systems; binding and encapsulating the action to be executed object with the unified short-term token; generating a cross-business system call request; and sending it to the corresponding business system.

[0013] As a preferred embodiment of the cross-domain context-enhanced action closed-loop processing method for e-commerce customer service described in this invention, the method includes: the corresponding business system performing corresponding business operations after authentication based on the unified short-term token, and allowing only the single path corresponding to the confirmed standard action object to execute business operations with external side effects. This includes: the corresponding business system receiving a business call request carrying a unified short-term token, verifying the unified short-term token, and parsing the operator identifier, authorized action type, target business system, target document identifier, and expiration time in the unified short-term token; when the verification is successful and the unified short-term token is within its validity period, performing local business rule verification; when the local business rule verification is successful, executing the corresponding business operation based on the target document identifier, wherein the local business rule verification includes at least one of business permission verification, business status verification, and business risk control verification; for business operations with external side effects, establishing a unique execution queue based on the unique execution identifier in the unified short-term token, and allowing only the standard action object confirmed by customer service to enter the unique execution queue to execute the corresponding business operation.

[0014] As a preferred embodiment of the cross-domain context-enhanced action closed-loop handling method for e-commerce customer service described in this invention, the following steps are included: obtaining the execution result of a business operation and writing the execution result and follow-up items back to the customer service task system, comprising: receiving the business operation execution result returned by the corresponding business system, and associating the business operation execution result with the corresponding standard action object according to the target document identifier; writing the business operation execution result and the follow-up items generated by the artificial intelligence model into the customer service task system, and updating the closed-loop status of the corresponding task, wherein the closed-loop status includes at least the following: pending confirmation, token issued, accepted by the business side, successful execution, failed execution, pending manual follow-up, and notified customer; continuously obtaining business status change information returned by the corresponding business system, and when a change in business status is detected, updating the closed-loop status in the customer service task system, and marking the corresponding task as notified customer after the customer service confirms the completion of the customer notification, thus completing the customer service closed-loop handling process.

[0015] Secondly, embodiments of the present invention provide an e-commerce customer service cross-domain context-enhanced action closed-loop processing system, comprising: an identity locking module, used to receive customer messages, determine the customer identity based on the mapping relationship between instant messaging external contact identifiers and business system user identifiers, and obtain a corresponding unified customer identity identifier; a business context aggregation module, used to, based on the unified customer identity identifier, concurrently obtain business data corresponding to the customer from multiple heterogeneous business systems through a combination of read-only bypass connections and read-only interfaces of business systems, and construct a structured business context; and an intelligent judgment module, used to input customer dialogue history, the structured business context, and customer service operation intentions into multiple artificial intelligence models for concurrent judgment, and generate candidate response content and corresponding standard action objects. The standard action object includes at least an action type, a target business system, a target document identifier, action parameters, and a risk level; the action confirmation and routing module is used to respond to the customer service's confirmation operation of the standard action object, issue a unified short-term token for cross-business system calls based on the confirmed standard action object, and route the confirmed standard action object to the corresponding business system according to the action type; the cross-system execution control module is used to allow the corresponding business system to perform the corresponding business operation after authentication based on the unified short-term token, and to allow only the single path corresponding to the confirmed standard action object to be executed for business operations with external side effects; the closed-loop write-back module is used to obtain the execution result of the business operation and write the execution result and follow-up items back to the customer service task system.

[0016] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, they implement the steps of the cross-domain context-enhanced action closed-loop processing method for e-commerce customer service as described in the first aspect of the present invention.

[0017] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of the cross-domain context-enhanced action closed-loop processing method for e-commerce customer service as described in the first aspect of the present invention.

[0018] The beneficial effects of this invention are as follows: By unifying customer identity locking, enhancing cross-system business context, intelligently analyzing multiple models, unifying short-term token authentication, and executing single-path closed loops, this invention achieves one-stop intelligent response and automatic handling of e-commerce customer service across business systems, thereby improving customer service processing efficiency and business execution accuracy. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of the method of the present invention. Figure 2 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0023] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0024] Example 1: Refer to Figures 1-2 This is the first embodiment of the present invention. This embodiment provides a closed-loop processing method for cross-domain context enhancement actions in e-commerce customer service, including: S1: receiving customer messages, determining the customer identity based on the mapping relationship between the instant messaging external contact identifier and the business system user identifier, and obtaining the corresponding unified customer identity identifier.

[0025] Furthermore, the system obtains the instant messaging external contact identifier corresponding to the customer message and queries the corresponding business system user identifier based on the pre-established identity mapping relationship table. When a unique corresponding business system user identifier is found, the business system user identifier is determined as the customer's unified identity identifier. When there is no corresponding mapping relationship or multiple candidate mapping relationships exist, the system performs mapping verification based on the customer's historical session records, order association information, and contact information association information, and writes the verification result into the identity mapping relationship table to update the mapping relationship between the customer's unified identity identifier and the instant messaging external contact identifier.

[0026] For example, a pre-established identity mapping table is queried based on the instant messaging external contact identifier. The identity mapping table records the correspondence between instant messaging external contact identifiers and business system user identifiers, as shown in Table 1 for example: Table 1 Identity Mapping Relationship Table Instant Messaging External Contact Identifier Business System User Identifier Mapping state IM01 UID01 Confirmed IM02 UID02 Confirmed

[0027] When the system finds that the external contact identifier IM01 in instant messaging uniquely corresponds to the user identifier UID01 in the business system, it directly identifies the user identifier UID01 as the unified identity identifier for that customer and uses it as the data query identifier for subsequent order system, logistics system, marketing system, refund system and ERP system.

[0028] Furthermore, when no corresponding relationship exists in the identity mapping table or multiple candidate business system user identifiers are retrieved, the system initiates the identity mapping verification process. For example, the instant messaging external contact identifier IM03 corresponds to two candidate business system user identifiers, as shown in Table 2: Table 2 Candidate Identity Mapping Table Instant Messaging External Contact Identifier Candidate business system user identifier IM03 UID03 IM03 UID04

[0029] The system retrieves customer historical conversation records, order association information, and contact information association information for consistency verification. Historical conversation records include recent customer inquiries; order association information includes order number, recipient, and contact information; and contact information association information includes instant messaging platform registration information or contact information provided by the customer.

[0030] For example, if a customer's historical conversation contains an order number ORD01, the order system query will find that the business system user identifier corresponding to order ORD01 is UID03, and the contact information associated with the order is consistent with the contact information registered on the instant messaging platform. Therefore, the system determines that the business system user identifier UID03 is the customer's unified identity identifier.

[0031] After identity verification, the system writes the correspondence between the instant messaging external contact identifier IM03 and the business system user identifier UID03 into the identity mapping table and updates the identity mapping relationship. When a customer message corresponding to the instant messaging external contact identifier is received again, the system directly determines the customer's unified identity identifier based on the updated identity mapping relationship, without needing to perform mapping verification again.

[0032] S2: Based on the customer's unified identity identifier, business data corresponding to the customer is obtained concurrently from multiple heterogeneous business systems through a combination of read-only bypass connection and read-only interface of business system, and a structured business context is constructed.

[0033] Furthermore, read-only bypass connections are established with the order system, logistics system, procurement system, marketing system, refund system, ERP system, and exception handling system. The read-only interfaces of each business system are invoked to concurrently acquire business data corresponding to the unified customer identity identifier. A unified field mapping relationship is established based on the data types returned by each business system. Field standardization processing is performed on the data from different business systems, and the data is associated and assembled according to a preset context data structure to generate a structured business context. The structured business context includes at least customer identity fields, order status fields, logistics node fields, refund execution fields, coupon availability fields, procurement arrival fields, and exception status fields. The consistency of the data structure acquired by the read-only bypass connection is checked. When a change in the business data structure is detected, the corresponding data parsing rules are updated according to the field mapping relationship, and the structured business context is reconstructed.

[0034] For example, after determining that the customer's unified identity is UID01, the system uses the customer's unified identity as the data association identifier across business systems. It establishes read-only bypass connections for the order system, logistics system, procurement system, marketing system, refund system, ERP system, and exception handling system, and simultaneously calls the read-only interfaces of each business system to concurrently obtain the business data corresponding to the customer's unified identity.

[0035] Table 3 shows examples of data returned by each business system: Table 3 Business Data Returned by Each Business System Business System Return business data Order System Order Number: OD01; Order Status: Shipped Logistics system Logistics milestones: Shipped from distribution center; Logistics status: In transit Procurement System Product Restock Status: No restock required Marketing System Available coupons: Coupon A Refund System Refundable status: Partial refunds are allowed. ERP system Business processing status: Normal Exception handling system Abnormal status: No abnormality

[0036] After acquiring the aforementioned business data, the system standardizes the data fields of different business systems based on a pre-established unified field mapping relationship. For example, the order status field in the order system, the logistics status field in the logistics system, the refund status field in the refund system, and the business processing status field in the ERP system are mapped to unified field names, and then associated and assembled according to the data structure of the pre-defined structured business context to generate a structured business context, as shown in Table 4. Table 4 Structured Business Context Field Name Field content Customer identity field UID01 Order status field Shipped Logistics node field The distribution center has already shipped the goods. Refund executable field Partial refunds allowed Coupon available fields Coupon A Purchase arrival field No need to restock Abnormal status field No abnormalities After generating the structured business context, the system caches the structured business context in the current customer service session and uses it as data input for subsequent business analysis by the artificial intelligence model.

[0037] Furthermore, when the system detects a change in the business data structure obtained from the read-only bypass connection, such as when the order system adds a new field called "delivery method," the system first determines the data type of the new field based on the pre-established field mapping relationship. If a corresponding mapping exists in the field mapping relationship, the system automatically updates the data parsing rules and re-parses the data returned by the order system. If no corresponding mapping exists, the new field is marked as a field to be mapped. After completing the field mapping configuration, the structured business context is reconstructed to ensure that the data from each business system can be associated and assembled according to a unified data structure and used for subsequent business processing.

[0038] S3: Input the customer's dialogue history, the structured business context, and the customer service operation intent into multiple artificial intelligence models for concurrent analysis, and generate candidate response content and corresponding standard action objects. The standard action objects include at least the action type, target business system, target document identifier, action parameters, and risk level.

[0039] Furthermore, the customer dialogue history, the structured business context, and the customer service operation intent are encoded into model input data and sent in parallel to multiple artificial intelligence models for reasoning to obtain multiple candidate response results;

[0040] A lightweight intent parsing model is used to perform semantic parsing on customer dialogue content, extracting at least one business parameter from order number, recipient, delivery address, refund amount and product information, and integrating the business parameters with customer dialogue history, structured business context and customer service operation intent to form model input data for concurrent analysis by multiple artificial intelligence models;

[0041] The action type, target business system, and target document identifier are extracted from the candidate response results output by multiple artificial intelligence models, and the business consistency is calculated in combination with the structured business context. If the consistency of multiple candidate response results meets a preset threshold, the corresponding candidate response result is determined as the target candidate response result. If the consistency of multiple candidate response results does not meet the preset threshold, the model input data is reconstructed based on the order status field, logistics node field, refund executable field, and abnormal status field in the business context, and sent to the artificial intelligence model for further analysis.

[0042] The candidate response results output by each artificial intelligence model are uniformly analyzed, and corresponding standard action objects are generated according to the preset action object description specifications. The standard action object includes at least the action type, target business system, target document identifier, action parameters, risk level, and candidate response text. Action routing information is generated based on the action type and target business system of the standard action object.

[0043] It should be noted that the multiple artificial intelligence models described in this embodiment adopt a complementary collaborative reasoning approach for concurrent analysis. Each artificial intelligence model performs different tasks for customer intent understanding, business context reasoning, and business action generation, in order to reduce the probability of a single model producing erroneous reasoning results.

[0044] Specifically, the first AI model is used for customer semantic understanding. It uses a large language model to perform contextual semantic encoding on the customer's dialogue history to obtain the customer's current business needs. The second AI model is used for business context reasoning. Based on the order status field, logistics node field, refund executable field, coupon available field, purchase arrival field, and abnormal status field in the structured business context, it analyzes whether the current business meets the execution conditions and obtains the business status reasoning result. The third AI model is used for business action generation. Based on the customer's business needs and the business status reasoning result, it generates candidate response content and corresponding standard action objects.

[0045] Furthermore, the lightweight intent parsing model uses a combination of sequence labeling and entity recognition to perform semantic parsing on customer input, identify business entities such as order number, recipient, delivery address, refund amount, product information and contact information, and generate structured business parameters based on the relationships between entities, which serve as business input for multiple artificial intelligence models.

[0046] Furthermore, after each AI model completes its inference, it outputs candidate response results along with the corresponding action type, target business system, target document identifier, action parameters, and risk level. The system then establishes a business consistency evaluation model based on the structured business context and performs interactive consistency analysis on multiple candidate response results.

[0047] The business consistency evaluation model calculates the degree of matching between each candidate response result and the order status field, logistics node field, refund execution field, coupon availability field, purchase arrival field, and abnormal status field, and generates a business consistency score based on preset weights. The score can be expressed as: ,in, Assign a business consistency score to the candidate response results. The number of business fields participating in the business consistency calculation. For business field serial number, The weight coefficient is the one corresponding to the i-th business field. When the maximum business consistency score among multiple candidate response results is greater than the preset threshold, the corresponding candidate response result is determined as the target candidate response result; otherwise, the model input data is reorganized according to the business field with the low score, and the reorganized model input data is sent to the multiple artificial intelligence models again for collaborative reasoning to obtain the updated target candidate response result.

[0048] Furthermore, during the re-inference process, only business fields with consistency conflicts are incrementally updated, while the remaining structured business context remains unchanged, thereby reducing the amount of data for repeated model inference and improving the efficiency of cross-business system context enhancement analysis.

[0049] For example, after generating the structured business context, the customer service system obtains the customer's current conversation content as: "The order has not been delivered yet, please help me check the logistics. If it cannot be delivered today, please change the delivery address." The system encodes the customer's historical dialogues, the structured business context, and the customer service operation intent to form the model input data. The customer's historical dialogues reflect recent customer inquiries, the structured business context reflects business information such as order status, logistics status, refund status, and coupon status, and the customer service operation intent represents the current business processing entry point selected by the customer service workbench, such as checking logistics or changing the delivery address.

[0050] Subsequently, the system uses a lightweight intent parsing model to semantically parse the current dialogue content of the customer and extract business parameters from the customer dialogue. For example, the extraction results are shown in Table 5: Table 5. Extraction Results of the Lightweight Intent Parsing Model Business parameters Extraction results Order Number ORD01 recipient USER_A Shipping address ADDRESS_A Product Information Product A Customer demands Track logistics and change delivery address

[0051] The system integrates the aforementioned business parameters with customer history, structured business context, and customer service operation intentions to form model input data for concurrent analysis by multiple artificial intelligence models. This integrated model input data includes not only customer natural language information but also structured business information such as order status fields, logistics node fields, refund actionable fields, and coupon availability fields, thus constructing a complete business context.

[0052] Subsequently, the system sends the model input data in parallel to multiple different artificial intelligence models for inference, and each artificial intelligence model generates corresponding candidate response results. For example, the inference results of each artificial intelligence model are shown in Table 6: Table 6. Candidate response results from multiple artificial intelligence models Artificial intelligence model Candidate response results Model A The current order is being shipped; please change the customer's shipping address. Model B The current order is expected to arrive; customers are advised to continue waiting. Model C The current order supports changing the shipping address, and it is recommended to check the logistics information at the same time.

[0053] After receiving the candidate response results returned by each artificial intelligence model, the system performs unified parsing on each candidate response result and converts them into standard action objects according to preset action object description specifications. For example, the natural language processing results generated by different artificial intelligence models are uniformly converted into a data structure containing action type, target business system, target document identifier, action parameters, risk level, and candidate response text, as shown in Table 7.

[0054] Table 7 Standard Action Objects Field Name Field content Action type Change shipping address Target business system Order System Target document identifier ORD01 Action parameters ADDRESS_A Risk level Medium risk Candidate response text The current order supports changing the shipping address. Please confirm whether to proceed.

[0055] Furthermore, the system generates action routing information based on the action type in the standard action object and the target business system. For example, when the action type is to modify the delivery address and the target business system is the order system, action routing information for the corresponding order system interface is generated; when the action type is to query logistics and the target business system is the logistics system, action routing information for the corresponding logistics system interface is generated.

[0056] S4: Respond to the customer service's confirmation operation for the standard action object, issue a unified short-term token for cross-business system calls based on the confirmed standard action object, and route the confirmed standard action object to the corresponding business system according to the action type.

[0057] Furthermore, in response to the customer service's confirmation instruction for the standard action object, the standard action object is subjected to action parameter integrity verification and risk level verification. When the verification passes, the standard action object is determined to be the action object to be executed.

[0058] A unified short-term token is generated based on the action object to be executed. The unified short-term token includes at least the operator identifier, authorized action type, target business system, target document identifier, expiration time, and unique execution identifier.

[0059] Based on the pre-established mapping relationship between action types and business systems, the target business system interface to be called is determined, and the action object to be executed is bound and encapsulated with the unified short-term token to generate a cross-business system call request, which is then sent to the corresponding business system.

[0060] For example, after generating standard action objects, the customer service workbench displays the candidate standard action objects generated by the artificial intelligence model to customer service personnel for confirmation. For instance, for customer inquiry information with a unified customer identity identifier of UID01, the standard action object generated by the system includes: action type is modifying the delivery address, target business system is the order system, target document identifier is ORD01, action parameter is the new delivery address information, risk level is medium risk, and corresponding candidate response text.

[0061] After reviewing the standard action object, customer service personnel perform a confirmation operation through the customer service workbench. Upon receiving the confirmation instruction from the customer service representative, the system first verifies the integrity of the action parameters and the risk level of the standard action object.

[0062] The action parameter integrity check is used to determine whether the current standard action object contains the necessary parameters required to perform the corresponding business operation. For example, when the action type is to modify the delivery address, the system checks whether the standard action object contains the target order identifier, the customer's unified identity identifier, and the delivery address information to be modified; if all of the above parameters exist, it is determined that the action parameter integrity meets the execution conditions.

[0063] Furthermore, the system performs risk level verification on the standard action objects according to preset risk rules. For example, for operations that may affect business status, such as modifying the delivery address, processing refunds, and making payments to customer service, the system determines whether the execution requirements are met based on the action type, the status of the target document, and the operation permission information. When the action parameters are complete and the risk level meets the execution conditions, the system identifies the standard action object as the action object to be executed.

[0064] Subsequently, the system generates a unified short-term token based on the action to be executed. The unified short-term token contains information such as operator identifier, authorized action type, target business system, target document identifier, expiration time, and unique execution identifier.

[0065] For example, for the current operation of modifying the delivery address, the system generates a corresponding unified short-term token. The operator identifier is used to record the identity of the customer service personnel who performed the confirmation operation. The authorized action type is used to limit the current token to only perform the operation of modifying the delivery address. The target business system is used to limit the call to the order system interface. The target document identifier is used to limit the operation object to order ORD01. The expiration time is used to limit the validity period of the token. The unique execution identifier is used to identify the current business operation task.

[0066] Furthermore, based on the pre-established mapping relationship between action types and business systems, the system determines the target business system interface corresponding to the action object to be executed. For example, when the action type is to modify the delivery address, the system determines the corresponding target business system as the order system and calls the address modification interface in the order system; when the action type is to process a refund, the system determines the corresponding target business system as the refund system and calls the refund processing interface; when the action type is to reissue a coupon, the system determines the corresponding target business system as the marketing system and calls the coupon issuance interface.

[0067] After identifying the target business system interface, the system binds and encapsulates the action object to be executed with a unified short-term token, generating a cross-business system call request. The call request includes the action type corresponding to the action object to be executed, the target document identifier, action parameters, and the unified short-term token, and is sent to the corresponding business system.

[0068] S5: After the corresponding business system completes the authentication based on the unified short-term token, it executes the corresponding business operation, and for business operations with external side effects, only the single path corresponding to the confirmed standard action object is allowed to be executed.

[0069] Furthermore, the corresponding business system receives a business call request carrying a unified short-term token, verifies the unified short-term token, and parses the operator identifier, authorized action type, target business system, target document identifier, and expiration time in the unified short-term token. When the verification is successful and the unified short-term token is within its validity period, local business rule verification is performed. When the local business rule verification is successful, the corresponding business operation is executed according to the target document identifier. The local business rule verification includes at least one of business permission verification, business status verification, and business risk control verification. For business operations with external side effects, a unique execution queue is established based on the unique execution identifier in the unified short-term token. Only standard action objects confirmed by customer service are allowed to enter the unique execution queue to execute the corresponding business operation.

[0070] For example, after generating a cross-business system call request, the target business system receives the business call request, wherein the business call request carries a unified short-term token and a corresponding action object to be executed.

[0071] For example, the current action to be executed is to modify the order delivery address, the target business system is the order system, and the business call request contains the target document identifier ORD01 and the corresponding unified short-term token TOKEN01.

[0072] After receiving the business call request, the order system first verifies the signature of the unified short-term token and parses the information contained in the unified short-term token, such as the operator identifier, authorized action type, target business system, target document identifier, expiration time, and unique execution identifier.

[0073] Among them, the operator identifier is used to confirm the identity of the customer service personnel corresponding to the current business operation; the authorized action type is used to limit the business actions that the current request is allowed to execute; the target business system is used to confirm whether the current request is sent to the corresponding business system; the target document identifier is used to determine the specific business processing object; the expiration time is used to determine whether the current token is within the validity period; and the unique execution identifier is used to identify the current business operation task.

[0074] When the unified short-term token verification is successful and it is detected that the unified short-term token has not exceeded the preset validity period, the order system further executes local business rule verification.

[0075] Specifically, the order system performs business permission verification based on the parsed operator identifier to determine whether the current customer service personnel have the authority to modify order information; it queries the current status of the corresponding order based on the target document identifier ORD01 and performs business status verification to determine whether the current order meets the conditions for modifying the delivery address; at the same time, it performs business risk control verification based on preset risk rules to determine whether there is any abnormal modification risk in the current operation.

[0076] For example, when a customer service representative's role is order customer service, their authorized operation scope includes order inquiry, modification of delivery address, and logistics inquiry; if the authorized action type corresponding to the business call request falls within the authorized operation scope, the business permission verification passes; otherwise, the current business operation is terminated, and a permission verification failure message is returned.

[0077] After completing the business permission verification, the order system queries the corresponding order information based on the target document identifier ORD01 and performs order status verification.

[0078] Specifically, order status verification includes determining whether the current business status of the order meets the execution conditions corresponding to the action to be performed. For example, when the action type is "modify delivery address," the system obtains the order's payment status, shipping status, and delivery status, and makes judgments based on preset business rules. When the order is in a paid but not yet delivered status, the system determines that the current order meets the delivery address modification conditions. When the order has been signed for or cancelled, the system determines that the current order does not meet the delivery address modification conditions and returns a business status verification failure message.

[0079] Furthermore, after the business status verification is passed, the order system performs business risk control verification.

[0080] Specifically, the system obtains the historical operation records corresponding to the current operator and the target order, the number of business modifications within the preset time window of the current order, and the parameter changes between the current business request and historical business requests. It then calculates a business risk score based on preset risk rules. The formula for calculating the business risk score is as follows: ,in, Assign a risk score to the current business operation. This represents the number of times a target order can be modified within a preset time window. This represents the number of historical abnormal operations corresponding to the current operator and the target order. This represents the degree of parameter change between the current business request and historical business requests. , , The numbers represent the number of modifications, respectively. When the risk score is less than the preset risk threshold, the current business request is determined to meet the business risk control requirements. When the risk score is greater than or equal to the preset risk threshold, the current business request is determined to have an abnormal operation risk, and the current business operation is terminated. At the same time, the corresponding standard action object is marked as pending manual review.

[0081] Once the business permission verification, business status verification, and business risk control verification are all passed, the order system retrieves the corresponding order information based on the target document identifier ORD01, and performs a delivery address modification operation, updating the new delivery address parameters to the corresponding order data.

[0082] Furthermore, for business operations with external side effects, such as refunds, customer service payments, and creation of business work orders, the system adopts a unique execution control mechanism to prevent multiple artificial intelligence models from triggering the same business operation at the same time.

[0083] For example, multiple AI models generated three candidate action objects: refund processing, continue waiting, and logistics inquiry. Among them, the refund processing action object and the customer service payment action object may both result in changes to business data.

[0084] The system establishes a unique execution queue based on the unique execution identifier in the unified short-term token, and only standard action objects that have been confirmed by customer service are allowed to enter the unique execution queue.

[0085] For example, after customer service confirms the execution of the refund processing action, the refund processing action object enters the execution queue with a unique execution identifier EXEC01, and the refund system executes the corresponding refund operation; other refund suggestion objects generated by artificial intelligence models are only displayed as candidate suggestions on the customer service workbench because they have not been confirmed by customer service, and do not generate corresponding business write requests, nor do they enter the business execution queue.

[0086] Furthermore, when multiple AI models simultaneously output the same type of business action, such as multiple models suggesting a refund, the system still determines a unique action object to be executed based on the customer service confirmation result, and only allows the execution path corresponding to that action object to call the refund system interface, thereby avoiding duplicate refunds, duplicate order creation, or duplicate modification of business data.

[0087] S6: Obtain the execution result of the business operation and write the execution result and follow-up matters back to the customer service task system.

[0088] Furthermore, the system receives the business operation execution results returned by the corresponding business system, and associates the business operation execution results with the corresponding standard action objects based on the target document identifier; it writes the business operation execution results and the follow-up items generated by the artificial intelligence model into the customer service task system, and updates the closed-loop status of the corresponding task, wherein the closed-loop status includes at least the following: pending confirmation, token issued, accepted by the business side, successful execution, failed execution, pending manual follow-up, and notified customer;

[0089] Continuously acquire business status change information returned by the corresponding business system. When a change in business status is detected, update the closed-loop status in the customer service task system. After the customer service confirms the completion of the customer notification, mark the corresponding task as notified to the customer, thus completing the customer service closed-loop handling process.

[0090] For example, after completing the business operation, the target business system returns the business operation execution result to the customer service task system.

[0091] For example, after customer service confirms the execution of the operation to modify the delivery address, the order system updates the corresponding order information based on the target document identifier ORD01 and returns the business operation execution result to the customer service task system. The business operation execution result includes the execution request identifier, the target document identifier, the execution status, and business processing information.

[0092] After receiving the business operation execution result returned by the order system, the system queries the corresponding standard action object based on the target document identifier ORD01, and associates the business operation execution result with the standard action object.

[0093] For example, the system determines the standard action object corresponding to the target document identifier ORD01 as follows: Action type: Modify delivery address; Target business system: Order system; Target document identifier: ORD01; Action parameter: ADDRESS_A. Subsequently, the system binds the execution status returned by the order system to this standard action object, forming a corresponding business operation record. When the order system returns a success message, the system determines that the business operation corresponding to the current standard action object has been completed; when the order system returns a failure message, the system retains the corresponding failure reason and generates a subsequent manual processing task.

[0094] Furthermore, the system writes the results of business operations and follow-up items generated by the artificial intelligence model into the customer service task system, and updates the closed-loop status of the corresponding tasks according to the business processing flow.

[0095] For example, in a customer request to change their shipping address, the customer service task initially has a status of "pending confirmation." Once the customer service representative confirms the standard action object, the system updates the task status to "token issued." When the order system receives the call request and passes the business verification, the task status is updated to "accepted by the business side." Once the order modification is completed, the task status is updated to "execution successful."

[0096] Meanwhile, the AI ​​model generates follow-up items based on the current business context. For example, after the order address is modified, a confirmation notification needs to be sent to the customer.

[0097] The system associates the above follow-up items with the corresponding customer service tasks and displays them on the customer service workbench.

[0098] Furthermore, the system continuously acquires business status change information returned by the corresponding business systems. For example, in a refund scenario, the refund system returns status change information such as "refund application completed" and "refund amount received"; in a logistics scenario, the logistics system returns information on changes in transportation nodes.

[0099] When a change in business status is detected, the system updates the closed-loop status in the customer service task system based on the target document identifier.

[0100] For example, when the refund system returns a refund completed status, the system updates the corresponding task status from successful execution to pending customer notification; after customer service personnel confirm that a refund completed notification has been sent to the customer, the system receives the customer service confirmation and updates the corresponding task status to customer notified.

[0101] Furthermore, when a business operation fails, the system updates the customer service task status to "execution failed" or "awaiting manual follow-up" based on the returned failure information, and synchronizes the reason for the failure and handling suggestions to the customer service workbench.

[0102] This embodiment also provides an e-commerce customer service cross-domain context-enhanced action closed-loop processing system, including: an identity locking module, used to receive customer messages, determine the customer's identity based on the mapping relationship between instant messaging external contact identifiers and business system user identifiers, and obtain the corresponding unified customer identity identifier; a business context aggregation module, used to, based on the unified customer identity identifier, concurrently obtain business data corresponding to the customer from multiple heterogeneous business systems through a combination of read-only bypass connections and read-only interfaces of business systems, and construct a structured business context; and an intelligent judgment module, used to input customer dialogue history, the structured business context, and customer service operation intentions into multiple artificial intelligence models for concurrent judgment, generate candidate response content and corresponding standard action objects, and the... The standard action object includes at least the action type, target business system, target document identifier, action parameters, and risk level; the action confirmation and routing module is used to respond to the customer service's confirmation operation of the standard action object, issue a unified short-term token for cross-business system calls based on the confirmed standard action object, and route the confirmed standard action object to the corresponding business system according to the action type; the cross-system execution control module is used to allow the corresponding business system to perform the corresponding business operation after authentication based on the unified short-term token, and to allow only the single path corresponding to the confirmed standard action object to be executed for business operations with external side effects; the closed-loop write-back module is used to obtain the execution result of the business operation and write the execution result and follow-up items back to the customer service task system.

[0103] This embodiment also provides a computer device applicable to a cross-domain context-enhanced action closed-loop handling method for e-commerce customer service, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the cross-domain context-enhanced action closed-loop handling method for e-commerce customer service as proposed in the above embodiment.

[0104] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0105] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements a closed-loop processing method for cross-domain context enhancement actions in e-commerce customer service as proposed in the above embodiment.

[0106] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for closed-loop processing of cross-domain context-enhanced actions in e-commerce customer service, characterized in that, include: Upon receiving customer messages, determine the customer's identity based on the mapping relationship between the instant messaging external contact identifier and the business system user identifier, and obtain the corresponding unified customer identity identifier; Based on the customer's unified identity identifier, business data corresponding to the customer is obtained concurrently from multiple heterogeneous business systems through a combination of read-only bypass connection and read-only interface of business system, and a structured business context is constructed. The system inputs customer dialogue history, the structured business context, and customer service operation intent into multiple AI models for concurrent analysis, generating candidate response content and corresponding standard action objects. Each standard action object includes at least an action type, target business system, target document identifier, action parameters, and risk level. The system responds to customer service confirmation of the standard action object, issues a unified short-term token for cross-business system calls based on the confirmed standard action object, and routes the confirmed standard action object to the corresponding business system according to the action type. The corresponding business system performs the corresponding business operation after authentication based on the unified short-term token, and for business operations with external side effects, only the single path corresponding to the confirmed standard action object is allowed to execute. The system obtains the execution result of the business operation and writes the execution result and follow-up information back to the customer service task system.

2. The method for closed-loop processing of cross-domain context enhancement actions in e-commerce customer service as described in claim 1, characterized in that, The process of receiving customer messages and determining customer identity based on the mapping relationship between instant messaging external contact identifiers and business system user identifiers, and obtaining the corresponding unified customer identity identifier, includes: obtaining the instant messaging external contact identifier corresponding to the customer message, and querying the corresponding business system user identifier according to a pre-established identity mapping relationship table; when a unique corresponding business system user identifier is found, the business system user identifier is determined as the unified customer identity identifier; when there is no corresponding mapping relationship or there are multiple candidate mapping relationships, a mapping verification is performed based on the customer's historical session records, order association information, and contact information association information, and the verification result is written into the identity mapping relationship table to update the mapping relationship between the unified customer identity identifier and the instant messaging external contact identifier.

3. The method for closed-loop processing of cross-domain context enhancement actions in e-commerce customer service as described in claim 1, characterized in that, The method combines read-only bypass connections and read-only interfaces of business systems to concurrently acquire business data corresponding to the customer from multiple heterogeneous business systems and construct a structured business context. This includes: establishing read-only bypass connections corresponding to the order system, logistics system, procurement system, marketing system, refund system, ERP system, and exception handling system; calling the read-only interfaces of each business system to concurrently acquire business data corresponding to the customer's unified identity identifier; establishing a unified field mapping relationship based on the data types returned by each business system; standardizing the fields of data from different business systems; and associating and assembling the data according to a preset context data structure to generate a structured business context. The structured business context includes at least customer identity fields, order status fields, logistics node fields, refund execution fields, coupon availability fields, procurement arrival fields, and exception status fields. The method also includes consistency checks on the data structure acquired through the read-only bypass connections. When a change in the business data structure is detected, the corresponding data parsing rules are updated according to the field mapping relationship, and the structured business context is reconstructed.

4. The method for closed-loop processing of cross-domain context enhancement actions in e-commerce customer service as described in claim 1, characterized in that, The process of inputting customer dialogue history, structured business context, and customer service operation intent into multiple artificial intelligence models for concurrent analysis to generate candidate response content and corresponding standard action objects includes: encoding customer dialogue history, structured business context, and customer service operation intent into model input data, and sending them in parallel to multiple artificial intelligence models for inference to obtain multiple candidate response results; using a lightweight intent parsing model to perform semantic parsing on the customer dialogue content, extracting at least one business parameter from order number, recipient, delivery address, refund amount, and product information, and fusing the business parameters with customer dialogue history, structured business context, and customer service operation intent to form model input data for concurrent analysis by the multiple artificial intelligence models; and extracting action type and target business from the candidate response results output by the multiple artificial intelligence models. The system identifies the business system and target document, and calculates business consistency based on the structured business context. If the consistency of multiple candidate response results meets a preset threshold, the corresponding candidate response result is determined as the target candidate response result. If the consistency of multiple candidate response results does not meet the preset threshold, the model input data is reconstructed based on the order status field, logistics node field, refund executable field, and abnormal status field in the business context, and sent to the artificial intelligence model for further analysis. The candidate response results output by each artificial intelligence model are uniformly parsed, and corresponding standard action objects are generated according to the preset action object description specifications. The standard action object includes at least the action type, target business system, target document identifier, action parameters, risk level, and candidate response text. Action routing information is generated based on the action type and target business system of the standard action object.

5. The method for closed-loop processing of cross-domain context enhancement actions in e-commerce customer service as described in claim 1, characterized in that, The process of responding to customer service's confirmation of the standard action object, issuing a unified short-term token for cross-business system calls based on the confirmed standard action object, and routing the confirmed standard action object to the corresponding business system according to the action type includes: responding to customer service's confirmation instruction for the standard action object, performing action parameter integrity verification and risk level verification on the standard action object, and determining the standard action object as an action to be executed when the verification passes; generating a unified short-term token based on the action to be executed, the unified short-term token including at least an operator identifier, authorized action type, target business system, target document identifier, expiration time, and unique execution identifier; determining the target business system interface to be called based on the pre-established mapping relationship between action types and business systems, binding and encapsulating the action to be executed with the unified short-term token, generating a cross-business system call request, and sending it to the corresponding business system.

6. The method for closed-loop processing of cross-domain context enhancement actions in e-commerce customer service as described in claim 1, characterized in that, The process of executing corresponding business operations after authentication by the corresponding business system based on the unified short-term token, and allowing only the single path corresponding to the confirmed standard action object to execute business operations with external side effects, includes: the corresponding business system receiving a business call request carrying a unified short-term token, verifying the unified short-term token, and parsing the operator identifier, authorized action type, target business system, target document identifier, and expiration time in the unified short-term token; when the verification is successful and the unified short-term token is within its validity period, performing local business rule verification; when the local business rule verification is successful, executing the corresponding business operation based on the target document identifier, wherein the local business rule verification includes at least one of business permission verification, business status verification, and business risk control verification; for business operations with external side effects, establishing a unique execution queue based on the unique execution identifier in the unified short-term token, and only allowing standard action objects confirmed by customer service to enter the unique execution queue to execute the corresponding business operation.

7. The method for closed-loop processing of cross-domain context enhancement actions in e-commerce customer service as described in claim 1, characterized in that, The step of obtaining the execution result of the business operation and writing the execution result and follow-up items back to the customer service task system includes: receiving the business operation execution result returned by the corresponding business system, and associating the business operation execution result with the corresponding standard action object according to the target document identifier; writing the business operation execution result and the follow-up items generated by the artificial intelligence model into the customer service task system, and updating the closed-loop status of the corresponding task, wherein the closed-loop status includes at least pending confirmation, token issued, business side accepted, execution successful, execution failed, pending manual follow-up, and customer notified; continuously obtaining business status change information returned by the corresponding business system, and when a change in business status is detected, updating the closed-loop status in the customer service task system, and marking the corresponding task as notified to the customer after the customer service confirms the completion of the customer notification, thus completing the customer service closed-loop handling process.

8. A cross-domain context-enhanced action closed-loop processing system for e-commerce customer service, characterized in that, include: The identity locking module is used to receive customer messages, determine the customer's identity based on the mapping relationship between the instant messaging external contact identifier and the business system user identifier, and obtain the corresponding unified customer identity identifier; The business context aggregation module is used to concurrently obtain business data corresponding to the customer from multiple heterogeneous business systems based on the customer's unified identity identifier, through a combination of read-only bypass connection and read-only interface of business system, and construct a structured business context. The intelligent analysis module is used to input customer dialogue history, the structured business context and customer service operation intent into multiple artificial intelligence models for concurrent analysis, and generate candidate response content and corresponding standard action objects. The standard action objects include at least action type, target business system, target document identifier, action parameters and risk level. The action confirmation and routing module is used to respond to the customer service's confirmation operation on the standard action object, issue a unified short-term token for cross-business system calls based on the confirmed standard action object, and route the confirmed standard action object to the corresponding business system according to the action type. The cross-system execution control module is used to execute corresponding business operations after the corresponding business system completes authentication based on the unified short-term token, and only allows the execution of business operations with external side effects to be performed through a single path corresponding to the confirmed standard action object; the closed-loop write-back module is used to obtain the execution result of the business operation and write the execution result and follow-up matters back to the customer service task system.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the cross-domain context enhancement action closed-loop processing method for e-commerce customer service as described in 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 the processor, it implements the steps of the cross-domain context enhancement action closed-loop processing method for e-commerce customer service as described in any one of claims 1 to 7.

Images