Business execution intelligent agent behavior risk control method, device, equipment and medium

Abstract

The application discloses a behavior risk control method, device and equipment of an enterprise business execution agent, and a medium, relates to the technical field of agents, and comprises the following steps: acquiring execution information of an enterprise business execution agent in the process of executing an enterprise business task, and converting the execution information into structured information; if it is determined that the structured information meets a first preset risk level, the structured information is verified to obtain a verification result, the structured information is reviewed based on the enterprise business execution agent to obtain a review result, calibration information is generated based on the verification result and the review result, the calibration information is converted into a control parameter, and the enterprise business execution agent is adjusted by using the control parameter; and if it is determined that the structured information meets a second preset risk level, the enterprise business execution agent is adjusted based on the second preset risk level and a preset risk response strategy. The application can accurately control the behavior of an agent based on the execution result of the agent.

Description

Technical Field

[0001] This invention relates to the field of intelligent agent technology, and in particular to a method, apparatus, equipment and medium for controlling behavioral risks of intelligent agents in enterprise business execution. Background Technology

[0002] In enterprise-level applications, intelligent agents have gradually penetrated into the automated execution phase of complex business processes. Their execution results are directly related to the state of core business systems, having a critical impact on business stability and reliability. However, existing intelligent agent technology systems have significant shortcomings. They focus primarily on task planning before execution and state monitoring during execution, lacking a systematic design for handling results after task execution, making it difficult to form closed-loop control capabilities.

[0003] The core shortcomings of existing technologies are concentrated in several aspects, and are even more prominent in complex business scenarios. First, execution results are often treated as one-time outputs, lacking structured analysis, quantitative evaluation, and effective reuse of deviations between the results and expected goals, failing to transform deviations into effective basis for optimizing subsequent behaviors. Second, once result deviations or risks occur, they are difficult to transform into control signals that can be directly perceived by subsequent tasks, leading to repeated occurrences of similar deviations and restricting the stability of agent execution. Third, the processing methods rely excessively on anomaly alarms or manual review, and the execution results are not deeply integrated into the system control logic, failing to form rigid technical constraints on the agent's subsequent behavior, resulting in low efficiency and susceptibility to overlooking risks due to human negligence. Fourth, the risk levels of different results vary significantly in complex scenarios, but existing systems lack differentiated execution strategies based on risk levels, making it difficult to adapt to diverse risk management needs. Furthermore, agent execution results are mostly in unstructured form, making it difficult for the system to accurately analyze and trace deviations, and lacking a post-execution behavior reflection mechanism, only able to passively handle deviations and unable to form autonomous feedback and calibration capabilities.

[0004] Therefore, how to conduct precise risk control on the behavior of intelligent agents based on their execution results after they have completed business tasks is a technical problem that urgently needs to be solved. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a method, apparatus, device, and medium for controlling the behavioral risks of an enterprise business execution intelligent agent, enabling precise risk control of the agent's behavior based on its execution results after the agent completes its business tasks. The specific solution is as follows: Firstly, this application provides a method for controlling behavioral risks of an enterprise business execution intelligent agent, including: After the enterprise business execution agent completes the enterprise business task, the execution information of the enterprise business execution agent during the execution of the enterprise business task is obtained, and the execution information is converted into structured information. The structured information is subjected to risk assessment. If the structured information is determined to meet a first preset risk level, a structured verification operation is performed on the structured information to obtain a verification result. The structured information is then reviewed based on the enterprise business execution agent to obtain a review result. Calibration information is generated based on the verification result and the review result, and the calibration information is converted into control parameters. These control parameters are then used to adjust the behavior of the enterprise business execution agent accordingly. The calibration information includes behavioral deviation data, behavioral risk level, and behavioral adjustment strategies of the enterprise business execution agent. If the structured information is determined to meet the second preset risk level, the behavior of the enterprise business execution agent will be adjusted accordingly based on the second preset risk level and the preset risk response strategy; the level of the first preset risk level is lower than the level of the second preset risk level.

[0006] Optionally, obtaining the execution information of the enterprise business execution agent during the execution of the enterprise business task, and converting the execution information into structured information, includes: The execution steps, execution paths, target parameters, input and output data, execution results, status codes corresponding to the execution steps, status codes corresponding to the execution results, and context information with a preset association with the enterprise business task are obtained by the enterprise business execution intelligent agent in the process of executing the enterprise business task, so as to determine the corresponding execution information. The execution information is standardized and structured to obtain structured information.

[0007] Optionally, the risk assessment of the structured information includes: Based on the structured information, the task type of the enterprise's business task is determined, and based on the task type, the risk sensitivity and risk control strategy corresponding to the enterprise's business task are determined. Based on the structured information, the operation result information, data change information, and status change information of the business system corresponding to the enterprise business task are determined, so as to determine the target information corresponding to the enterprise business task; Based on the structured information, identify a first similar task in the historical execution tasks that has preset similar conditions to the enterprise business task, and determine the task failure rate of the first similar task. Based on the structured information, the corresponding business rules, permission constraints, and strategy requirements are determined to identify the risk rules corresponding to the enterprise's business tasks. The risk assessment model uses the risk sensitivity, risk control strategy, target information, task failure rate, and risk rules to assess the risk of the structured information.

[0008] Optionally, the step of performing a structured verification operation on the structured information to obtain the verification result includes: The structured information is subjected to consistency verification, integrity verification, and compliance verification to complete the first verification operation; The execution steps, tool call information, and state change information in the structured information are verified to complete the second verification operation; Determine the deviation information between the structured information and the preset expected target corresponding to the enterprise's business task, and generate the corresponding deviation type and deviation degree based on the deviation information; The verification result of the structured information is obtained based on the first verification operation, the second verification operation, the deviation type, and the deviation degree.

[0009] Optionally, adjusting the behavior of the enterprise business execution agent using the control parameters includes: The control parameters are loaded into the behavior control component of the enterprise business execution agent to obtain the adjusted enterprise business execution agent. A second similar task with preset similar conditions to the enterprise business task is identified, and the execution strategy of the second similar task is adjusted by the adjusted enterprise business execution agent; the second similar task is the task to be executed by the enterprise business execution agent. The adjusted enterprise business execution agent adjusts the strategy selection weights, tool invocation order, and task execution parameters of the second similar task.

[0010] Optionally, after adjusting the behavior of the enterprise business execution agent using the control parameters, the method further includes: The adjusted enterprise business execution agent executes the current enterprise business task to obtain the current execution result. Determine whether the current execution result meets the preset result deviation condition, and obtain the determination result; If the judgment result does not meet the requirements, the behavior of the adjusted enterprise business execution agent will be adjusted accordingly based on the preset risk response strategy.

[0011] Optionally, adjusting the behavior of the enterprise business execution agent based on the second preset risk level and preset risk response strategy includes: When the risk level of the structured information is the third preset risk level in the second preset risk level, the enterprise business task is re-executed by the enterprise business execution agent based on a preset number of retries. When the risk level of the structured information is the fourth preset risk level in the second preset risk level, then a corresponding alternative path that meets the preset security is determined based on the execution path in the execution information, and the enterprise business execution agent re-executes the enterprise business task based on the alternative path. The third preset risk level is lower than the fourth preset risk level.

[0012] Secondly, this application provides a behavioral risk control device for an enterprise business execution intelligent agent, comprising: The information conversion module is used to obtain the execution information of the enterprise business execution intelligent agent during the execution of the enterprise business task after the enterprise business execution intelligent agent completes the enterprise business task, and convert the execution information into structured information. The first intelligent agent adjustment module is used to perform risk assessment on the structured information. If the structured information is determined to meet a first preset risk level, a structured verification operation is performed on the structured information to obtain a verification result. The module then performs a review operation on the structured information based on the enterprise business execution intelligent agent to obtain a review result. Calibration information is generated based on the verification result and the review result, and the calibration information is converted into control parameters. These control parameters are then used to adjust the behavior of the enterprise business execution intelligent agent accordingly. The calibration information includes behavioral deviation data, behavioral risk level, and behavioral adjustment strategies of the enterprise business execution intelligent agent. The second intelligent agent adjustment module is used to adjust the behavior of the enterprise business execution intelligent agent based on the second preset risk level and the preset risk response strategy if the structured information is determined to meet the second preset risk level; the level of the first preset risk level is lower than the level of the second preset risk level.

[0013] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor is used to execute the computer program to implement the aforementioned behavioral risk control method for enterprise business execution intelligent agents.

[0014] Fourthly, this application provides a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned behavioral risk control method for an enterprise business execution intelligent agent.

[0015] In this application, after the enterprise business execution agent completes an enterprise business task, execution information of the enterprise business execution agent during the execution of the enterprise business task is obtained and converted into structured information. Risk assessment is performed on the structured information. If the structured information is determined to meet a first preset risk level, a structured verification operation is performed on the structured information to obtain a verification result. The structured information is then reviewed based on the enterprise business execution agent to obtain a review result. Calibration information is generated based on the verification result and the review result, and the calibration information is converted into control parameters to adjust the behavior of the enterprise business execution agent accordingly. The calibration information includes behavioral deviation data, behavioral risk level, and behavioral adjustment strategy information of the enterprise business execution agent. If the structured information is determined to meet a second preset risk level, the behavior of the enterprise business execution agent is adjusted accordingly based on the second preset risk level and a preset risk response strategy. The level of the first preset risk level is lower than the level of the second preset risk level. As can be seen from the above, in this application, after the enterprise business execution intelligent agent completes the enterprise business task, it first collects the execution information during the execution of the task, and then converts the execution information into structured information. Risk assessment is performed on the structured information. If it is determined to meet the first preset risk level, a structured verification operation is performed on the structured information to obtain the verification result. Simultaneously, the enterprise business execution intelligent agent performs a review operation on the structured information to obtain the review result. Then, calibration information is generated by combining the verification result and the review result, and the calibration information is converted into control parameters. These control parameters are used to adjust the behavior of the enterprise business execution intelligent agent accordingly. The calibration information includes the behavioral deviation data, behavioral risk level, and behavioral adjustment strategy of the enterprise business execution intelligent agent. If the structured information is determined to meet the second preset risk level, the behavior of the enterprise business execution intelligent agent is adjusted accordingly by combining the second preset risk level and the preset risk response strategy. Furthermore, the level of the first preset risk level is lower than the level of the second preset risk level. In this way, this application can conduct precise risk control on the behavior of the intelligent agent based on its execution result after the intelligent agent completes the business task. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0017] Figure 1This application discloses a flowchart of a behavioral risk control method for an enterprise business execution intelligent agent. Figure 2 This is a schematic diagram of the behavioral risk control device for an enterprise business execution intelligent agent disclosed in this application; Figure 3 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Currently, existing intelligent agent technology systems have significant shortcomings. Their focus is largely on task planning before execution and state monitoring during execution, lacking a systematic design for handling results after task completion, making it difficult to form closed-loop control capabilities. The core deficiencies of existing technologies are concentrated in several aspects, and are more pronounced in complex business scenarios. First, execution results are often treated as one-time outputs, lacking structured analysis, quantitative evaluation, and effective reuse of deviations between results and expected goals, failing to transform deviations into effective basis for optimizing subsequent behaviors. Second, once result deviations or risks occur, they are difficult to transform into control signals that can be directly perceived by subsequent tasks, leading to repeated occurrences of similar deviations and hindering the stability of intelligent agent execution. Third, processing methods rely excessively on anomaly alarms or manual review; execution results are not deeply integrated into the system control logic, failing to provide rigid technical constraints on subsequent agent behavior, resulting in low efficiency and susceptibility to overlooking risks due to human error. Fourth, the risk levels of different results vary significantly in complex scenarios, but existing systems lack differentiated execution strategies based on risk levels, making it difficult to adapt to diverse risk management needs. Meanwhile, the execution results of intelligent agents are mostly in unstructured form, making it difficult for the system to accurately analyze and trace deviations. Furthermore, the system lacks a post-execution behavior reflection mechanism, only able to passively handle deviations and unable to develop autonomous feedback and calibration capabilities. Therefore, this application provides a method, apparatus, equipment, and medium for behavioral risk control of enterprise business execution intelligent agents, enabling precise risk control of the agent's behavior based on its execution results after the agent completes its business tasks.

[0020] See Figure 1 As shown in the figure, this invention discloses a method for controlling behavioral risks of an enterprise business execution intelligent agent, including: Step S11: After the enterprise business execution agent completes the enterprise business task, obtain the execution information of the enterprise business execution agent in the process of executing the enterprise business task, and convert the execution information into structured information.

[0021] In this embodiment, after the enterprise business execution agent completes a specific enterprise business task according to the received business instructions, it needs to retrospectively analyze the entire process of executing that task. The enterprise business task could be the automatic response and work order generation of an intelligent customer service representative to a customer's product inquiry, or the automatic approval and status update of an employee leave application process by a human resources intelligent agent. Such tasks involve clearly defined rule judgments and system operations. Therefore, it is first necessary to collect and obtain all execution information generated by the intelligent agent during the execution of the enterprise business task.

[0022] Specifically, the execution steps, execution paths, target parameters, input and output data, execution results, status codes corresponding to the execution steps, status codes corresponding to the execution results, and context information that has a preset association with the enterprise business task are obtained by the enterprise business execution intelligent agent in the process of executing the enterprise business task, so as to determine the corresponding execution information.

[0023] To facilitate efficient risk assessment, the acquired execution information needs to be converted into structured information with a standardized format and clear semantics. Therefore, the execution information is standardized and structured to obtain structured information.

[0024] Step S12: Perform risk assessment on the structured information. If the structured information is determined to meet the first preset risk level, perform a structured verification operation on the structured information to obtain a verification result. Then, perform a review operation on the structured information based on the enterprise business execution intelligence to obtain a review result. Generate calibration information based on the verification result and the review result, and convert the calibration information into control parameters to adjust the behavior of the enterprise business execution intelligence accordingly. The calibration information includes behavioral deviation data, behavioral risk level, and behavioral adjustment strategy of the enterprise business execution intelligence.

[0025] In this embodiment, after acquiring and forming structured information about the behavior of the enterprise business execution agent, a comprehensive risk assessment needs to be performed on this structured information to evaluate whether there are potential risks in the task execution process and the degree of those risks. The purpose of the risk assessment is to identify links in the agent's behavior that may deviate from expectations, violate rules, or lead to adverse consequences, and to provide a basis for decision-making on whether to initiate in-depth risk control procedures.

[0026] In one specific implementation, the task type of the enterprise business task is determined based on the structured information, and the risk sensitivity and risk control strategy corresponding to the enterprise business task are determined based on the task type; the operation result information, data change information, and state change information of the business system corresponding to the enterprise business task are determined based on the structured information to determine the target information corresponding to the enterprise business task; a first similar task with preset similar conditions to the enterprise business task is determined from historical executed tasks based on the structured information, and the task failure rate of the first similar task is determined; corresponding business rules, permission constraints, and strategy requirements are determined based on the structured information to determine the risk rules corresponding to the enterprise business task. A risk assessment model is used to assess the risk of the structured information based on the risk sensitivity, the risk control strategy, the target information, the task failure rate, and the risk rules.

[0027] When the risk assessment result determines that the current structured information meets the first preset risk level, it means that a medium or higher level of risk requiring intervention and control has been detected, which then triggers more detailed structured verification operations and agent self-inspection operations.

[0028] Specifically, the verification operation involves: performing consistency verification, integrity verification, and compliance verification on the structured information to complete the first verification operation; verifying the execution steps, tool call information, and state change information in the structured information to complete the second verification operation; determining the deviation information between the structured information and the preset expected target corresponding to the enterprise business task, and generating corresponding deviation types and deviation degrees based on the deviation information; and obtaining the verification result of the structured information based on the first verification operation, the second verification operation, the deviation type, and the deviation degree.

[0029] Among these, the self-censorship operation of the agent focuses on understanding behavioral intent and evaluating decision quality. The agent attempts to reconstruct the decision chain during task execution, analyzing the reasons for choosing a particular execution path in a specific context. The censorship operation may refer to the agent's richer internal state history or learned experience to obtain a censorship result regarding the rationality of the behavior and the quality of the decision.

[0030] After completing the aforementioned verification and review operations, the final calibration information includes detailed behavioral deviation data, a clear behavioral risk level, and preliminary behavioral adjustment strategy recommendations. However, this abstract calibration information cannot be directly applied to the agent; therefore, it needs to be converted into control parameters.

[0031] Furthermore, the control parameters are loaded into the behavior control component of the enterprise business execution agent to obtain the adjusted enterprise business execution agent; a second similar task with preset similar conditions is determined, and the execution strategy of the second similar task is adjusted through the adjusted enterprise business execution agent; the second similar task is the task to be executed by the enterprise business execution agent; the strategy selection weight, tool invocation order and task execution parameters of the second similar task are adjusted through the adjusted enterprise business execution agent.

[0032] To verify the effectiveness of the adjustments and form a control loop, after initial adjustments using control parameters, the adjusted enterprise business execution agent executes the current enterprise business task to obtain the current execution result. It is then determined whether the current execution result meets preset result deviation conditions, yielding a judgment result. If the judgment result indicates that the condition is not met, the behavior of the adjusted enterprise business execution agent is adjusted accordingly based on the preset risk response strategy. This allows for deeper or different-dimensional adjustments to the behavior of the adjusted enterprise business execution agent, ensuring that its behavioral risks are effectively controlled within an acceptable range.

[0033] Step S13: If the structured information is determined to meet the second preset risk level, the behavior of the enterprise business execution agent is adjusted accordingly based on the second preset risk level and the preset risk response strategy; the level of the first preset risk level is lower than the level of the second preset risk level.

[0034] In this embodiment, the second preset risk level represents a higher level of risk identified in the risk assessment. This typically means that the behavior of the enterprise's business execution agent has clearly deviated from security standards, triggered critical alarms, or is highly likely to cause substantial business damage, requiring immediate and more direct and effective intervention measures than those in step S12. In other words, when the output of the risk assessment model indicates that the current structured information meets the second preset risk level, the system will no longer rely solely on generating calibration parameters for gradual adjustments. Instead, it will execute clearly targeted behavioral adjustment instructions based on pre-defined risk response strategies for different high-risk scenarios to quickly contain the risk or correct the error.

[0035] Specifically, the second preset risk level can be further subdivided. When the risk level of the structured information is the third preset risk level within the second preset risk level, the enterprise business execution agent re-executes the enterprise business task based on a preset number of retries. When the risk level of the structured information is the fourth preset risk level within the second preset risk level, an alternative path that meets preset security is determined based on the execution path in the execution information, and the enterprise business execution agent re-executes the enterprise business task based on the alternative path. Therefore, this application effectively improves the security of agent behavior and the reliability of business task execution through a tiered risk response mechanism. The third preset risk level is lower than the fourth preset risk level.

[0036] As can be seen from the above, in this application, after the enterprise business execution intelligent agent completes the enterprise business task, it first collects the execution information during the execution of the task, and then converts the execution information into structured information. Risk assessment is performed on the structured information. If it is determined to meet the first preset risk level, a structured verification operation is performed on the structured information to obtain the verification result. Simultaneously, the enterprise business execution intelligent agent performs a review operation on the structured information to obtain the review result. Then, calibration information is generated by combining the verification result and the review result, and the calibration information is converted into control parameters. These control parameters are used to adjust the behavior of the enterprise business execution intelligent agent accordingly. The calibration information includes the behavioral deviation data, behavioral risk level, and behavioral adjustment strategy of the enterprise business execution intelligent agent. If the structured information is determined to meet the second preset risk level, the behavior of the enterprise business execution intelligent agent is adjusted accordingly by combining the second preset risk level and the preset risk response strategy. Furthermore, the level of the first preset risk level is lower than the level of the second preset risk level. In this way, this application can conduct precise risk control on the behavior of the intelligent agent based on its execution result after the intelligent agent completes the business task.

[0037] The technical solutions of the embodiments of this application will be described in detail below.

[0038] This invention proposes a technical solution for controlling the behavioral risks of intelligent agents based on post-execution reflection and behavior review mechanisms. By introducing a result-level control mechanism after the intelligent agent completes its task, the execution result is transformed from a one-time output into a control input that the system can absorb and utilize, thereby forming a closed-loop calibration system for subsequent tasks.

[0039] Specifically, the first step is to collect and structure the execution results.

[0040] Regarding execution path and parameter acquisition, after the agent completes task execution, the system automatically collects execution steps and paths, key parameters and input / output data, execution results and status codes, and task-related context information such as task type, affiliated business system, and operation time. In terms of structured representation, the system standardizes and structures the collected raw execution information, generating a unified data structure including task identifiers, step sequences, parameter values, execution result status, and log information, providing a parsable data foundation for subsequent reflection and review.

[0041] Secondly, a risk assessment mechanism is implemented. Based on the structured execution results, the system assesses risks and generates corresponding risk scores and risk levels as behavioral control signals.

[0042] The risk assessment model comprehensively analyzes factors such as task type, output format, historical failure rate, and predefined risk rules. Task type refers to different risk sensitivities and control strategies corresponding to different task types; output format includes operation success / failure, data changes, and changes in business system status; historical failure rate is the frequency of failures or deviations of similar tasks in historical executions; and predefined risk rules are enterprise business rules, permission constraints, and policy consistency requirements. Correspondingly, the system sets different risk assessment thresholds based on task type and output format: a low-risk threshold allows for slight deviations, with automatic system calibration; a medium-risk threshold triggers post-execution reflection and some behavioral restrictions; and a high-risk threshold triggers risk response strategies, such as manual confirmation or path degradation. Through this design, risk scoring is not only used to display results but also directly for subsequent behavioral control.

[0043] Then, behavioral review and post-implementation reflection are conducted. Based on risk assessment, the system performs the following actions, including structured verification, AI (Artificial Intelligence) behavioral self-review, and a review and reflection process for the generated reflection results.

[0044] Structured verification includes: verifying the consistency, completeness, and compliance of execution results; automatically verifying execution steps, tool calls, and business status changes; checking deviations from expected goals, and generating deviation type identifiers and deviation degrees. AI behavior self-review includes: the intelligent agent performing self-behavior analysis based on the review results; identifying possible causes of deviations and optimizable points in the execution path. Reflection result generation includes: the system generating post-execution reflection conclusions based on the verification and self-review results. These conclusions include: approval (execution result meets expectations and requires no further processing) and initiation of risk response procedures (execution result has deviations or risks, requiring the activation of control strategies).

[0045] Next, regarding risk response strategies, when the execution result is determined to be risky, the system automatically selects the appropriate risk response strategy based on the risk level and strategy requirements, including but not limited to: retry strategy, degradation strategy, and manual confirmation of the result.

[0046] The retry strategy includes: within a safe range, allowing agents to automatically retry the same task a limited number of times; the system limits the number of retries to prevent infinite loops and ensure business security. The degradation strategy includes: when a task is difficult to execute safely under current capabilities or risk conditions, selecting a "safer, more controllable, and more deterministic" alternative execution path; degradation paths may include simplifying task operations, using low-risk tools, or reducing cross-system call steps. Manual confirmation results include: for high-risk tasks or execution results with significant deviations, the system submits the task to manual review as a fallback; the manual confirmation results are fed back to the system for subsequent risk assessment and behavior calibration.

[0047] Finally, calibration strategy selection and behavior control can be performed, which may include loading calibration information and applying control parameters.

[0048] The calibration information loading process involves loading the calibration information generated from post-execution reflection and review, such as deviation type, risk level, and behavior adjustment suggestions, into the agent's behavior control module. Control parameter application involves converting the calibration information into control parameters to adjust the agent's subsequent behavior. These parameters include: strategy selection weights (prioritizing similar task execution strategies); tool invocation priority (adjusting the tool invocation order based on risk control); and decision thresholds or execution parameter ranges (dynamically adjusting task execution parameters based on risk level).

[0049] In addition, this application also includes closed-loop calibration and continuous optimization.

[0050] Closed-loop calibration includes: the system automatically loading control parameters during subsequent executions of similar tasks, implementing technical constraints on future task behavior based on execution result deviations; and forming a closed-loop feedback mechanism through a continuous execution-reflection-calibration process, improving the stability of the agent in complex tasks. Continuous optimization includes: the system dynamically optimizing the risk assessment model and calibration strategy based on historical execution deviations and risk handling records; enabling differentiated control for different task types, avoiding a "one-size-fits-all" approach, and improving the overall reliability of business execution.

[0051] Therefore, this application, by introducing a post-execution reflection and behavior review mechanism, extends the governance of intelligent agent behavior from "pre-execution planning and in-execution monitoring" to "post-execution result-driven control." It overcomes the technical bottlenecks of traditional intelligent agent systems, such as one-time output of execution results, difficulty in feedback and calibration, and achieves systematic absorption and continuous optimization of complex task execution results. Specifically, this application can improve the stability and controllability of intelligent agents, meaning that without modifying the internal structure of the original model, closed-loop control of behavior can be achieved through system-level mechanisms, ensuring the reliable execution of complex tasks in multi-system and multi-tool environments. This application can reduce the probability of repeated deviations in execution results, meaning that through result-level risk assessment and calibration strategies, the system can quantitatively manage and continuously optimize deviations, significantly reducing the recurrence of errors or abnormal operations in similar tasks. This application can enhance execution consistency and credibility, meaning that reflection, review, and risk response are integrated into the task execution closed loop, ensuring that the intelligent agent outputs more consistent and predictable results in enterprise-level business scenarios. This application can improve the interpretability and governability of the system, meaning that execution results and deviation information are structured, quantified, and usable for decision-making, facilitating risk tracking, strategy optimization, and compliance review, meeting the regulatory and governance needs of financial, government, and high-reliability enterprise management scenarios.

[0052] Accordingly, see Figure 2 As shown in the figure, this application embodiment provides a behavioral risk control device for an enterprise business execution intelligent agent, including: The information conversion module 11 is used to obtain the execution information of the enterprise business execution intelligent agent during the execution of the enterprise business task after the enterprise business execution intelligent agent completes the enterprise business task, and convert the execution information into structured information. The first intelligent agent adjustment module 12 is used to perform risk assessment on the structured information. If the structured information is determined to meet a first preset risk level, a structured verification operation is performed on the structured information to obtain a verification result. The structured information is then reviewed based on the enterprise business execution intelligent agent to obtain a review result. Calibration information is generated based on the verification result and the review result, and the calibration information is converted into control parameters to adjust the behavior of the enterprise business execution intelligent agent accordingly. The calibration information includes behavioral deviation data, behavioral risk level, and behavioral adjustment strategy information of the enterprise business execution intelligent agent. The second intelligent agent adjustment module 13 is used to adjust the behavior of the enterprise business execution intelligent agent according to the second preset risk level and the preset risk response strategy if the structured information is determined to meet the second preset risk level; the level of the first preset risk level is lower than the level of the second preset risk level.

[0053] In some specific embodiments, the information conversion module 11 specifically includes: The first information determination unit is used to acquire the execution steps, execution paths, target parameters, input and output data, execution results, status codes corresponding to the execution steps, status codes corresponding to the execution results, and context information that has a preset association with the enterprise business task in the process of the enterprise business execution intelligent agent executing the enterprise business task, so as to determine the corresponding execution information. The information processing unit is used to standardize and structure the execution information to obtain structured information.

[0054] In some specific embodiments, the first intelligent agent adjustment module 12 specifically includes: The risk determination unit is used to determine the task type of the enterprise business task based on the structured information, and to determine the risk sensitivity and risk control strategy corresponding to the enterprise business task based on the task type. The second information determination unit is used to determine the operation result information, data change information, and status change information of the business system corresponding to the enterprise business task based on the structured information, so as to determine the target information corresponding to the enterprise business task. The failure rate determination unit is used to determine, based on the structured information, a first similar task that has preset similar conditions to the enterprise business task in the historical execution tasks, and to determine the task failure rate of the first similar task. The rule determination unit is used to determine the corresponding business rules, permission constraints, and policy requirements based on the structured information, so as to determine the risk rules corresponding to the enterprise business tasks; The information judgment unit is used to perform risk judgment on the structured information based on the risk sensitivity, the risk control strategy, the target information, the task failure rate, and the risk rules through a risk judgment model.

[0055] In some specific embodiments, the first intelligent agent adjustment module 12 specifically includes: The first information verification unit is used to perform consistency verification, integrity verification and compliance verification on the structured information to complete the first verification operation; The second information verification unit is used to verify the execution steps, tool call information and state change information in the structured information to complete the second verification operation. The data generation unit is used to determine the deviation information between the structured information and the preset expected target corresponding to the enterprise business task, and to generate the corresponding deviation type and deviation degree based on the deviation information; The first result determination unit is used to obtain the verification result of the structured information based on the first verification operation, the second verification operation, the deviation type, and the deviation degree.

[0056] In some specific embodiments, the first intelligent agent adjustment module 12 specifically includes: The model adjustment unit is used to load the control parameters into the behavior control component of the enterprise business execution agent to obtain the adjusted enterprise business execution agent; The strategy adjustment unit is used to determine a second similar task that has preset similar conditions to the enterprise business task, and to adjust the execution strategy of the second similar task through the adjusted enterprise business execution agent; the second similar task is the task to be executed by the enterprise business execution agent. The parameter adjustment unit is used to adjust the strategy selection weight, tool invocation order, and task execution parameters of the second similar task of the execution strategy through the adjusted enterprise business execution agent.

[0057] In some specific embodiments, the first intelligent agent adjustment module 12 further includes: The second result determination unit is used to execute the current enterprise business task through the adjusted enterprise business execution intelligent agent to obtain the current execution result; The result judgment unit is used to determine whether the current execution result meets the preset result deviation condition and obtain the judgment result. The behavior adjustment unit is used to adjust the behavior of the adjusted enterprise business execution agent based on the preset risk response strategy if the judgment result does not meet the requirements.

[0058] In some specific embodiments, the second intelligent agent adjustment module 13 specifically includes: The first task execution unit is used to re-execute the enterprise business task through the enterprise business execution agent based on a preset number of retries when the risk level of the structured information is the third preset risk level in the second preset risk level. The second task execution unit is configured to determine a corresponding alternative path that meets the preset security requirements based on the execution path in the execution information when the risk level of the structured information is the fourth preset risk level in the second preset risk level, and then re-execute the enterprise business task based on the alternative path through the enterprise business execution agent; wherein the level of the third preset risk level is lower than the level of the fourth preset risk level.

[0059] Furthermore, embodiments of this application also disclose an electronic device, Figure 3This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the behavioral risk control method for enterprise business execution intelligent agents disclosed in any of the foregoing embodiments. Furthermore, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0060] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0061] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0062] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the behavioral risk control method for an enterprise business execution intelligent agent executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include computer programs capable of performing other specific tasks.

[0063] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned behavioral risk control method for an enterprise business execution intelligent agent. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.

[0064] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0065] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0066] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0067] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0068] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for controlling behavioral risks of an enterprise business execution intelligent agent, characterized in that, include: After the enterprise business execution agent completes the enterprise business task, the execution information of the enterprise business execution agent during the execution of the enterprise business task is obtained, and the execution information is converted into structured information. The structured information is subjected to risk assessment. If the structured information is determined to meet a first preset risk level, a structured verification operation is performed on the structured information to obtain a verification result. The structured information is then reviewed based on the enterprise business execution agent to obtain a review result. Calibration information is generated based on the verification result and the review result, and the calibration information is converted into control parameters. These control parameters are then used to adjust the behavior of the enterprise business execution agent accordingly. The calibration information includes behavioral deviation data, behavioral risk level, and behavioral adjustment strategies of the enterprise business execution agent. If the structured information is determined to meet the second preset risk level, the behavior of the enterprise business execution agent will be adjusted accordingly based on the second preset risk level and the preset risk response strategy. The first preset risk level is lower than the second preset risk level.

2. The behavioral risk control method for enterprise business execution intelligent agents according to claim 1, characterized in that, The step of obtaining the execution information of the enterprise business execution agent during the execution of the enterprise business task, and converting the execution information into structured information, includes: The execution steps, execution paths, target parameters, input and output data, execution results, status codes corresponding to the execution steps, status codes corresponding to the execution results, and context information with a preset association with the enterprise business task are obtained by the enterprise business execution intelligent agent in the process of executing the enterprise business task, so as to determine the corresponding execution information. The execution information is standardized and structured to obtain structured information.

3. The behavioral risk control method for enterprise business execution intelligent agents according to claim 1, characterized in that, The risk assessment of the structured information includes: Based on the structured information, the task type of the enterprise's business task is determined, and based on the task type, the risk sensitivity and risk control strategy corresponding to the enterprise's business task are determined. Based on the structured information, the operation result information, data change information, and status change information of the business system corresponding to the enterprise business task are determined, so as to determine the target information corresponding to the enterprise business task; Based on the structured information, identify a first similar task in the historical execution tasks that has preset similar conditions to the enterprise business task, and determine the task failure rate of the first similar task. Based on the structured information, the corresponding business rules, permission constraints, and strategy requirements are determined to identify the risk rules corresponding to the enterprise's business tasks. The risk assessment model uses the risk sensitivity, risk control strategy, target information, task failure rate, and risk rules to assess the risk of the structured information.

4. The behavioral risk control method for enterprise business execution intelligent agents according to claim 3, characterized in that, The step of performing a structured verification operation on the structured information to obtain the verification result includes: The structured information is subjected to consistency verification, integrity verification, and compliance verification to complete the first verification operation; The execution steps, tool call information, and state change information in the structured information are verified to complete the second verification operation; Determine the deviation information between the structured information and the preset expected target corresponding to the enterprise's business task, and generate the corresponding deviation type and deviation degree based on the deviation information; The verification result of the structured information is obtained based on the first verification operation, the second verification operation, the deviation type, and the deviation degree.

5. The behavioral risk control method for enterprise business execution intelligent agents according to claim 1, characterized in that, The step of adjusting the behavior of the enterprise business execution agent using the control parameters includes: The control parameters are loaded into the behavior control component of the enterprise business execution agent to obtain the adjusted enterprise business execution agent. A second similar task with preset similar conditions to the enterprise business task is identified, and the execution strategy of the second similar task is adjusted by the adjusted enterprise business execution agent; the second similar task is the task to be executed by the enterprise business execution agent. The adjusted enterprise business execution agent adjusts the strategy selection weights, tool invocation order, and task execution parameters of the second similar task.

6. The behavioral risk control method for enterprise business execution intelligent agents according to claim 5, characterized in that, After adjusting the behavior of the enterprise business execution agent using the control parameters, the method further includes: The adjusted enterprise business execution agent executes the current enterprise business task to obtain the current execution result. Determine whether the current execution result meets the preset result deviation condition, and obtain the determination result; If the judgment result does not meet the requirements, the behavior of the adjusted enterprise business execution agent will be adjusted accordingly based on the preset risk response strategy.

7. The behavioral risk control method for an enterprise business execution intelligent agent according to any one of claims 1 to 6, characterized in that, The adjustment of the behavior of the enterprise business execution agent based on the second preset risk level and preset risk response strategy includes: When the risk level of the structured information is the third preset risk level in the second preset risk level, the enterprise business task is re-executed by the enterprise business execution agent based on a preset number of retries. When the risk level of the structured information is the fourth preset risk level in the second preset risk level, then a corresponding alternative path that meets the preset security is determined based on the execution path in the execution information, and the enterprise business execution agent re-executes the enterprise business task based on the alternative path. The third preset risk level is lower than the fourth preset risk level.

8. A behavioral risk control device for an enterprise business execution intelligent agent, characterized in that, include: The information conversion module is used to obtain the execution information of the enterprise business execution intelligent agent during the execution of the enterprise business task after the enterprise business execution intelligent agent completes the enterprise business task, and convert the execution information into structured information. The first intelligent agent adjustment module is used to perform risk assessment on the structured information. If the structured information is determined to meet the first preset risk level, a structured verification operation is performed on the structured information to obtain a verification result. The enterprise business execution intelligent agent is then used to review the structured information to obtain a review result. Calibration information is generated based on the verification result and the review result, and the calibration information is converted into control parameters to adjust the behavior of the enterprise business execution intelligent agent accordingly. The calibration information includes behavioral deviation data, behavioral risk level, and behavioral adjustment strategies of the enterprise business execution intelligent agent. The second intelligent agent adjustment module is used to adjust the behavior of the enterprise business execution intelligent agent according to the second preset risk level and the preset risk response strategy if the structured information is determined to meet the second preset risk level. The first preset risk level is lower than the second preset risk level.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the behavioral risk control method for an enterprise business execution agent as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store computer programs; wherein, when the computer programs are executed by a processor, they implement the behavioral risk control method for an enterprise business execution intelligent agent as described in any one of claims 1 to 7.

Images