Human-machine collaborative post dynamic allocation method and related device

Abstract

Embodiments of the present application relate to the technical field of artificial intelligence, and disclose a man-machine collaborative post dynamic allocation method and related device, which comprises: obtaining a replaceability evaluation result of a target task generated by an external evaluation process; sequentially attempting a plurality of analysis paths according to a preset priority to determine a target post that undertakes the target task; selecting a target man-machine collaboration strategy from a preset man-machine collaboration strategy pool according to a readiness level, and constraining the post type of the target post according to the readiness level; performing a graded readiness compliance check on the target post, marking the severity of each check item of the check, and gating the migration of the life cycle state machine of the target post from the verification state to the active state with the check result, and only when all key-level check items pass, the post is allowed to migrate to the active state and undertake the target task, otherwise the activation is blocked. The evaluation level of the embodiments of the present application not only drives collaboration but also constrains analysis, realizes adaptive automation and pre-execution compliance auditability.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, specifically to a method and apparatus for dynamic allocation of human-machine collaborative positions. Background Technology

[0002] Currently, as large language model-driven AI agents are introduced into enterprise business processes, the core obstacle to implementing AI in regulated industries (such as securities, finance, and healthcare) is how to translate decisions about "which tasks to assign to AI agents, which to retain to humans, and in what form of collaboration" into executable and manageable engineering implementations.

[0003] In the existing technology, there are two typical solutions and their respective shortcomings: The first type is the manual task routing mechanism of workflow engines. For example, a workflow engine based on business process modeling and annotation specifications uses a multi-level resolution chain of "assignee → candidate users → candidate groups" to determine the assignee of a manual task. If the previous level has no result, it falls back to the next level. The shortcomings of this mechanism are: its "candidates" are all homogeneous human users, not crossing the two heterogeneous subjects of humans and intelligent agents; and its routing is completely unaffected by the task-level substitutability evaluation results, that is, the two things "whether it should be handed over to automation" and "who should undertake it" are independent and do not affect each other.

[0004] The second category includes automated decision-making mechanisms based on risk scoring cards and human-in-the-loop mechanisms during operation. The former uses weighted summation to score multiple risk or capability dimensions of a task, and then determines whether to automate based on the total score. Its shortcoming is that the weighted summation can "average out" a single fatal weakness (for example, a task may perform well in most dimensions, but has irreversible, unverifiable, or legally prohibited fatal flaws, and may still get a total score that allows automation after weighting), which poses a structural risk of over-automation. The latter temporarily switches to human intervention during execution based on the confidence level output by the model. Its governance point is during execution, rather than before the executor is put on duty, making it difficult to meet the mandatory requirement of "auditable compliance checks before AI is put on duty" in regulated industries.

[0005] In summary, the existing technology has the following shortcomings: 1. Dualistic abstraction of executors. Mainstream workflow engines model executors as binary roles of "human task / system task," lacking a mechanism to incorporate human executors and AI agents into the same organizational governance abstraction. This results in AI agents being unable to bear organizational attributes such as job positions, readiness, and compliance constraints, making it difficult to integrate them into the enterprise's compliance and audit system.

[0006] 2. Decoupling of task assignment from substitutability. Task assignment to executors often relies on static rules, manual designation, or role-based access control, without dynamically selecting executors based on quantitative assessments of task-level substitutability; and assignment does not adapt accordingly when assessment results change.

[0007] 3. Lack of rollback robustness. A single dispatch path fails directly when the target executor is missing, lacking multi-level rollback.

[0008] 4. Lack of compliance gates before execution. Current assignments begin execution immediately after the executor is determined, lacking tiered readiness compliance checks before execution.

[0009] Therefore, a dynamic allocation method for executors is needed, which is driven by the results of task substitutability assessment, unified human-machine governance, multi-path fallback, and mandatory hierarchical compliance verification before executors take up their posts. Summary of the Invention

[0010] In view of the above problems, embodiments of the present invention provide a method and related apparatus for dynamic allocation of human-machine collaborative positions, which are used to solve the problems of abstract binary of the executor and lack of rollback robustness in the process of allocating human-machine collaborative positions by intelligent agents in the prior art.

[0011] According to one aspect of the present invention, a method for dynamic allocation of human-machine collaborative positions is provided, the method comprising: Obtain the substitutability assessment results for the target task generated by the external evaluation process; the results include a readiness level; Multiple parsing paths are tried sequentially according to a preset priority to determine the target position to undertake the target task; wherein, the multiple parsing paths include at least: instance-level explicit position assignment path and task-related candidate position matching path; Based on the readiness level, a target human-machine collaboration strategy is selected from a preset human-machine collaboration strategy pool, and the job type of the target job is constrained based on the readiness level; the human-machine collaboration strategy pool includes strategies that are fully executed by intelligent agents, strategies that are executed by intelligent agents and reviewed by human agents, strategies that are co-piloted by human agents, and strategies that are executed only by human agents. A graded readiness compliance check is performed on the target position. Each check item is marked with a severity level, which is either critical, important, or minor. The check result is used to gate the life cycle state machine of the target position from the verification state to the active state: the position is allowed to migrate to the active state and assume the target task if and only if all critical check items pass; otherwise, activation is blocked.

[0012] In one alternative approach, the method further includes obtaining an evaluation result of the substitutability of the target task generated by an external evaluation process; prior to the result including a readiness level, the method also includes: Human executors and AI agents are modeled as peer entities capable of performing the same role; wherein, the role is bound to a set of skills required for the role, carries a job type identifier indicating the human or AI agent, and has a lifecycle state machine subject to compliance verification.

[0013] In one alternative approach, the readiness level is taken from a set of discrete levels of increasing artificial preservation. The external evaluation process includes a three-gate sequential short-circuit evaluation of the target task, wherein the three-gate sequential short-circuit evaluation consists of an eligibility gate, a tolerance gate, and a feasibility gate, in that order. The qualification gate is a logical AND operation on a set of Boolean criteria; the tolerance gate is a maximum value on a set of ordinal risk dimensions; the feasibility gate is a minimum value on a set of ordinal readiness dimensions, subject to segmented capping constraints on the task clarity dimension; when any criterion of the qualification gate is negative, the system is short-circuited to the highest manual retention level.

[0014] In one alternative approach, the step of selecting a target human-machine collaboration strategy from a preset pool of human-machine collaboration strategies based on the readiness level, and constraining the job type of the target job based on the readiness level, includes: when the readiness level indicates that the autonomous execution strategy is most suitable, parsing only in the job set where the job type is an agent; when the readiness level indicates that the human job execution strategy must be retained, parsing only in the job set where the job type is a human.

[0015] In one alternative approach, the job lifecycle state machine includes a draft state, a configuration state, a verification state, an activation state, and a suspended state. Among them, when a job in the active state fails a key inspection item, it is forcibly moved to the suspended state. A job in the suspended state can only be moved to the active state after passing all critical level checks.

[0016] In one optional approach, the parsing path further includes a default job path and a fallback path; the default job path and the fallback path are sorted according to a preset priority and are placed after the candidate job matching path; The default job path is used to return the default job configured for the target role, and is only returned when its status is active or verified. The fallback path is used to map to a pre-configured default job type based on the executor type identifier carried on the target task or role, wherein the executor type identifier indicates one of human, intelligent agent, service, or external system.

[0017] According to another aspect of the present invention, a human-machine collaborative job dynamic allocation device is provided, comprising: An acquisition module is used to acquire the substitutability assessment results for the target task generated by an external evaluation process; the results include a readiness level. The parsing module is used to try multiple parsing paths sequentially according to a preset priority in order to determine the target position to undertake the target task; wherein, the multiple parsing paths include at least: instance-level explicit position assignment paths and task-related candidate position matching paths. The strategy selection module is used to select a target human-machine collaboration strategy from a preset human-machine collaboration strategy pool according to the readiness level, and to constrain the job type of the target job according to the readiness level; the human-machine collaboration strategy pool includes a strategy that is fully executed by the intelligent agent job, a strategy that is executed by the intelligent agent job and reviewed by the human job, a human-machine co-pilot collaboration strategy, and a strategy that is executed only by the human job. The graded readiness compliance verification module is used to perform graded readiness compliance verification on the target position. Each check item of the verification is marked with a severity level, which is one of the critical level, important level, or minor level. The verification result is used to gate the life cycle state machine of the target position from the verification state to the active state: the position is allowed to migrate to the active state and undertake the target task if and only if all critical level check items pass; otherwise, activation is blocked.

[0018] In an alternative embodiment, the device further includes: The modeling module is used to model human executors and AI agent executors as equivalent entities capable of performing the same job; wherein, the job is bound to a set of skills required for the role, carries a job type identifier indicating the human or AI agent, and has a lifecycle state machine subject to compliance verification.

[0019] According to another aspect of the present invention, a computer device is provided, including: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus; The memory is used to store at least one executable instruction, which causes the processor to perform the operation of the human-machine collaborative job dynamic allocation method.

[0020] According to another aspect of the present invention, a computer-readable storage medium is provided, the storage medium storing at least one executable instruction, which, when executed on a computer device, causes the computer device to perform the operation of the human-machine collaborative job dynamic allocation method.

[0021] This invention embodiment obtains the substitutability assessment results of a target task generated by an external evaluation process; the results include a readiness level; multiple parsing paths are tried sequentially according to a preset priority to determine the target position to undertake the target task; wherein, the multiple parsing paths include at least: instance-level explicit position assignment paths and task-related candidate position matching paths; a target human-machine collaboration strategy is selected from a preset human-machine collaboration strategy pool according to the readiness level, and the position type of the target position is constrained according to the readiness level; the human-machine collaboration strategy pool includes a strategy where the agent position has full authority to execute, a strategy where the agent position executes and is reviewed by a human position, a human-machine co-pilot collaboration strategy, and a strategy where only a human position executes. (Omitted); A graded readiness compliance check is performed on the target position. Each check item is marked with a severity level, which is either critical, important, or minor. The check results are used to gate the lifecycle state machine of the target position from the verification state to the active state. The position is allowed to migrate to the active state and undertake the target task only if all critical check items pass. Otherwise, activation is blocked. This model can model human executors and artificial intelligence agents as equivalent subjects that can perform the same position (human-machine isomorphism). The readiness level generated by the external evaluation process drives the selection of collaboration strategies and constrains the candidate executor types in the position analysis. The graded compliance gate before execution controls the migration of the position's lifecycle state machine to the active state.

[0022] The above description is merely an overview of the technical solutions of the embodiments of the present invention. In order to better understand the technical means of the embodiments of the present invention and to implement them in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the embodiments of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0023] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart illustrating the human-machine collaborative job dynamic allocation method provided in an embodiment of the present invention is shown. Figure 2 This diagram illustrates the parsing process in the human-machine collaborative job dynamic allocation method provided in an embodiment of the present invention. Figure 3A flowchart illustrating the lifecycle state machine in a human-machine collaborative job dynamic allocation method provided in another embodiment of the present invention is shown. Figure 4 This diagram illustrates the compliance verification process in a human-machine collaborative job dynamic allocation method provided by another embodiment of the present invention. Figure 5 This diagram illustrates the structure of the human-machine collaborative job dynamic allocation device provided in an embodiment of the present invention. Figure 6 A schematic diagram of the structure of a computer device provided in an embodiment of the present invention is shown. Detailed Implementation

[0024] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0025] To ensure consistency in terminology, the following definitions are made: The "substitutability assessment" referred to in this embodiment of the invention refers to the assessment of the extent to which a target task can be performed by an artificial intelligence agent in place of a human, and its quantitative output is characterized by a "readiness level".

[0026] The "readiness level" referred to in this embodiment of the invention is the quantitative output value of the evaluation, which is taken from a set of discrete levels with increasing human retention. In this embodiment of the invention, it is represented by the first level (L1), the second level (L2), the third level (L3), and the fourth level (L4), where the first level indicates that it is most suitable for autonomous execution by an intelligent agent, and the fourth level indicates that it must be retained for human execution.

[0027] From this point onward, when referring to evaluation methods or task attributes, "substitutability evaluation" will be used, and when referring to the output value of the evaluation, "readiness level" will be used. This readiness level is generated by an external evaluation process, and the embodiments of this invention directly consume its results.

[0028] To enable unified analysis and collaboration between humans and intelligent agents, this invention introduces the concept of "Position" as a unified abstraction for human-machine isomorphism: both human executors and AI executors are modeled as peer entities capable of holding the same position; each position carries a position type identifier, which can be either human or intelligent agent; each position is bound to its set of skills and has a lifecycle state. Specifically, in this embodiment, a unified position abstraction for human-machine isomorphism is performed beforehand: both human executors and AI executors are modeled as peer entities capable of holding the same position. Each position is bound to a set of skills required for the role and has a lifecycle state machine subject to compliance checks. This is a prerequisite for enabling subsequent position analysis to be performed uniformly between the two heterogeneous entities of humans and intelligent agents. Figure 3 As shown, the job lifecycle state machine includes a draft state, a configuration state, a verification state, an active state, and a suspended state. A job in the active state is forcibly migrated to the suspended state when a key check item for that job subsequently fails. A job in the suspended state can only be migrated to the active state after passing all key check items again.

[0029] Figure 1 A flowchart of a human-machine collaborative job dynamic allocation method provided in an embodiment of the present invention is shown. This method is executed by a computer device. The computer device can be a desktop computer, a distributed device, a smart terminal, etc., and the embodiments of the present invention do not impose specific limitations. Figure 1 As shown, the method includes the following steps: Step 110: Obtain the evaluation results of the alternatives to the target task generated by the external evaluation process.

[0030] The substitutability assessment result includes a readiness level, which characterizes the degree of automation of the target task. The readiness level is derived from a set of discrete, progressively increasing levels of human retention. The external assessment process includes a three-gate sequential short-circuit assessment of the target task, consisting of a qualification gate, a tolerance gate, and a feasibility gate. The qualification gate is a logical AND operation on a set of Boolean criteria; the tolerance gate is the maximum value among a set of ordinal risk dimensions; and the feasibility gate is the weakest link among a set of ordinal readiness dimensions, subject to segmented capping constraints based on task clarity dimensions. When any criterion of the qualification gate is negative, the process short-circuits to the highest human retention level.

[0031] Specifically, embodiments of the present invention obtain the substitutability assessment results of a target task, the results of which include at least a readiness level. This readiness level is generated by an external evaluation process. In one embodiment, the external evaluation process is a three-gate sequential short-circuit evaluation process, and the present invention directly consumes its output readiness level. In embodiments of the present invention, the readiness level may carry a technical baseline parameter, representing the readiness level relative to an explicit deployment technical baseline, denoted as L@TB-x (e.g., the second level is represented as L2@TB-2 relative to an activity-aware baseline). The generation and remapping of this technical baseline relativization are completed by the external evaluation process, and the present invention directly consumes the level representation it carries. To enable those skilled in the art to understand the specific form of this input parameter, a specific mechanism for implementing this relativization in the external evaluation process is disclosed: A maximum context support capability is configured for each technical baseline; during evaluation, the context requirement level of the task is compared with the maximum context support capability of the deployment baseline. When the context requirement level of the task exceeds the maximum context support capability of the deployment baseline, the context adaptation dimension F3 of the feasible gate is lowered or capped, thereby automatically shifting the level representation upwards as the baseline increases, without needing to re-evaluate the inherent elements of the task such as input, processing logic, output, and constraints. In one embodiment, the maximum context support capability of the pure prompt baseline and the task local baseline is the current task level (CDL-1), the activity-aware baseline is the activity level (CDL-2), and the enterprise context baseline is the enterprise level (CDL-4). A complete numerical trajectory from before to after the upgrade is given: A task, due to its context requirement being at the activity level (CDL-2), is deployed at the task local baseline (maximum context support is CDL-1). Its context adaptation dimension F3 is limited, resulting in a third-level rating relative to the task local baseline, denoted as L3@TB-1. When the deployment baseline is upgraded to the activity-aware baseline (maximum context support increases to CDL-2), F3 is no longer limited and is increased. Following external evaluation remapping, the task level is automatically updated to L2@TB-2, while the inherent elements of the task, such as input, processing logic, output, and constraints, remain unchanged. Embodiments of this invention can directly use the level representation carrying the technical baseline, or only its readiness level portion can be used.

[0032] The external evaluation process for assessing the substitutability of the target task is specifically conducted as follows: A structured description of the target task is obtained, including at least its input, processing logic, output, and constraint information. This structured description is then filled into a preset prompt template to obtain model input. The prompt template includes evaluation constraints, which at least indicate the preset evaluation dimensions that the language model needs to output and the value types for each dimension. The model input is provided to the language model, which, under the constraints of the evaluation constraints, generates an evaluation result corresponding to the target task. The evaluation result includes the judgment values ​​and corresponding self-assessment confidence levels for each preset evaluation dimension. The preset evaluation dimensions include at least a first-category dimension belonging to a preset mandatory review category and a second-category dimension quantified based on evaluation indicators. The evaluation result undergoes data validity verification; if the verification passes, the evaluation result is saved as an inactive draft. Based on the category of each evaluation dimension and the self-assessed confidence level corresponding to each evaluation dimension, a set of dimensions to be reviewed is determined. This set includes first-category dimensions that have not been marked as reviewed and second-category dimensions whose self-assessed confidence level is lower than a preset confidence threshold. If the set of dimensions to be reviewed is not empty, a manual review result is received for the judgment value of the target dimension in the set. In response to the received manual review result, the target dimension is removed from the set of dimensions to be reviewed. If the set of dimensions to be reviewed is empty, a manual confirmation instruction is received for the ineffective draft, and in response to the manual confirmation instruction, the ineffective draft is converted to an effective state, resulting in an effective substitutability evaluation result. This substitutability evaluation result includes the readiness level of the target task, which indicates the human-machine collaboration mode in which the target task is executed by an intelligent agent. The preset evaluation dimensions include 11 dimensions, belonging to three categories: Eligibility Category (Boolean, belonging to the first category), including E1 Regulatory Permissibility, E2 Ethical and Relational Essence, and E3 Accountability Attributability, each taking a Boolean value; Tolerance Category (Ordinal, belonging to the second category), including T1 Reversibility, T2 Scope of Impact, and T3 Verification Delay, each taking an integer from 1 to 5; and Feasibility Category (Ordinal, belonging to the second category), including F1 Task Clarity, F2 Model Adaptability, F3 Context Adaptability, F4 Tool / Data Adaptability, and F5 Governance Adaptability, each taking an integer from 1 to 5. The evaluation constraints included in the prompt template declare the above 11 dimensions and their value types and ranges, and include format requirements: each dimension must simultaneously output a judgment value, judgment basis text, and a self-assessment confidence value between 0 and 1, returned as a JSON object.

[0033] The sequential short-circuit evaluations of the three gates are as follows: Eligibility gate (conjunctive aggregation): Perform a logical AND operation on a set of Boolean criteria. In one implementation, the criteria include E1 regulatory permissibility, E2 ethical and relational nature, and E3 accountability attributability; if any one of these is negative, the evaluation is immediately short-circuited, directly determining it to Level 4 (the highest level of human retention), and it will not proceed to subsequent gates.

[0034] Tolerance gate (worst-case aggregation): Take the maximum value of a set of ordinal risk dimensions. In one implementation, the dimensions include T1 reversibility, T2 range of influence, and T3 verification delay, each taking an integer value from 1 to 5; take the maximum value of the three to obtain the risk value R, and map it to a risk level according to a preset threshold: R≤2 is mapped to "low", R=3 is mapped to "medium", R=4 is mapped to "high", and R=5 is mapped to "severe".

[0035] Feasibility gate (short-board aggregation and segmented capping constraint): For a set of ordinal readiness dimensions, the shortest value (minimum) is selected and constrained by segmented capping of the task clarity dimension. The task clarity dimension is capped according to a preset segmented threshold: when the task clarity dimension value is not greater than a first threshold, the feasibility readiness is capped at a first upper limit value; when the task clarity dimension value is equal to a second threshold, the feasibility readiness is capped at a second upper limit value; when the task clarity dimension value is not less than a third threshold, no capping is applied. Furthermore, after the decision matrix mapping, when any readiness dimension is lower than a preset threshold, the readiness level is raised to the corresponding lower limit according to preset rules. In one embodiment, the dimensions include F1 task clarity, F2 model adaptation, F3 context adaptation, F4 tool and data adaptation, and F5 governance adaptation, each taking an integer from 1 to 5 (5 being the most ready). The initial value of feasibility readiness is the minimum of F2, F3, F4, and F5. When task clarity F1 ≤ 2, feasibility readiness is capped at 2; when F1 = 3, it is capped at 3; and when F1 ≥ 4, it is not capped. Furthermore, the capability gap C = 6 − feasibility readiness.

[0036] Subsequently, the risk level and capability gap C are input into a decision matrix to obtain the readiness level. The complete mapping rules of the decision matrix are shown in Table 1 below: Table 1: Risk level Range of values ​​for ability gap C Readiness level any C>4.0 Level 4 (L4) Critical C ≤ 4.0 Level 3 (L3) High C ≤ 3.0 Level 2 (L2) High 3.0<C ≤ 4.0 Level 3 (L3) Low or Medium C ≤ 2.0 Level 1 (L1) Low or Medium 2.0<C ≤ 3.0 Level 2 (L2) Low or Medium 3.0<C ≤ 4.0 Level 3 (L3) After the above decision matrix mapping, a constraint floor correction is further applied. When any readiness dimension falls below a preset threshold, the readiness level is raised to the corresponding lower limit according to preset rules, as shown in Table 2. Specifically: Table 2: Triggering conditions Raise the target (lower level). Context adaptation F3 ≤ 2, and the current level is first level. Raised to the second level The risk level is F5 ≤ 2, and the risk level is high or severe, currently at level one or two. Raised to the third level The model fits F2 ≤ 2, and is currently at the first or second level. Raised to the third level Tools and data adaptation F4 ≤ 2, and currently at level 1 or 2. Raised to the third level In the embodiment of the present invention, short-board aggregation means that if there is a single fatal shortboard during aggregation, it cannot be averaged out by the strengths of other dimensions. Compared with weighted summation, the technical effect of short-board aggregation is that a single fatal shortboard (such as a low score in the dimension corresponding to irreversibility, non-verifiability, and regulatory prohibition) cannot be averaged out by the strengths of other dimensions, thereby structurally preventing excessive automation, which is a key safety attribute in regulated scenarios.

[0037] A complete numerical example is given below. Given the inputs of each evaluation dimension for a certain task: E1=True, E2=True, E3=True; T1=2, T2=4, T3=3; F1=4, F2=3, F3=2, F4=4, F5=4, the deployment technology baseline is the activity-aware baseline (its maximum context support capability is at the activity level, denoted as CDL-2), and the context requirement level of the task is at the activity level (CDL-2). Then: all qualification gates are true, and pass; the risk value R of the tolerance gate = max(2,4,3)=4, which is mapped to the risk level "High"; the initial feasible readiness of the feasibility gate = min(3,2,4,4)=2, since F1=4≥4 no capping is applied, so the feasible readiness = 2, and the capability gap C=6−2=4.0; according to the decision matrix, the risk level is "High" and C=4.0 (falling into the interval 3.0<C≤4.0) is mapped to the third level; in the constraint check, F3=2≤2 but the current level is already the third level, the context floor is not triggered, and all other conditions are not satisfied, so the final readiness level is the third level (L3).

[0038] Step 120: Sequentially try a plurality of parsing paths according to a preset priority to determine the target position that undertakes the target task.

[0039] Wherein, the plurality of parsing paths at least include: an instance-level explicit position assignment path and a task-associated candidate position matching path.

[0040] In the embodiment of the present invention, position parsing is attempted sequentially according to the following priority, returns once hit, and falls back to the next path if not hit. The parsing described in this embodiment is performed in the candidate position set constrained by the readiness level of step 110 (that is, the candidate positions are screened according to the target performer type determined in step 130), so that the parsing result is adaptive between human positions and agent positions along with the evaluation level. In a specific implementation of the present invention, as Figure 2 shown, the following several parsing paths are included, specifically: Path (a) is an instance-level explicit job assignment: If the workflow instance has an explicit job binding for the role to which the target task belongs, then the bound job is returned. Even if the bound job is in a suspended state, this path still returns the job, but marks it as degraded, and its activation is blocked by the critical-level check item in subsequent step S4 (job not suspended check, corresponding to the positionNotSuspended check item). That is, "parsing and returning" and "activation blocking" are two separate steps: this path is only responsible for returning the job; whether it is allowed to take up the job is determined by the S4 gate.

[0041] Path (b) is the candidate job associated with the task: If path (a) yields no results, the ordered list of candidate jobs associated with the target role is read, and each job is checked one by one according to a preset selection strategy (in one implementation, a priority strategy, i.e., in list order). Only when a candidate job is in an active or verified state and its skill set is a superset of the skill set required by the target role is the candidate job qualified, and the first qualified job is returned. The selection strategy can be selected from priority, adaptive, manual assignment, etc.

[0042] Path (c) is the default job position for the role: If there is no result for path (b), the default job position configured for the target role is read and returned only if its status is active or verified.

[0043] Path (d) is a fallback: If path (c) yields no results, the target task or role is mapped to the pre-configured default job type based on the executor type identifier (indicating one of four categories: human, intelligent agent, service, or external system). This path is a safety net, always returning a fallback job type; when paths (a) to (c) fail and only this path resolves, a degradation flag is set and an upper-level alarm is triggered to make the situation observable. It should be noted that the "executor type identifier" mentioned in this path refers to the executor information carried by the executor type rather than a specific job identifier in the task or role configuration, and is unrelated to development history; its value range includes the four categories: human, intelligent agent, service, and external system.

[0044] Step 130: Select a target human-machine collaboration strategy from the preset human-machine collaboration strategy pool according to the readiness level, and constrain the job type of the target job according to the readiness level.

[0045] The human-machine collaboration strategy pool includes strategies for full execution by intelligent agents, strategies for execution by intelligent agents with review by human agents, strategies for human-machine co-pilot collaboration, and strategies for execution solely by human agents.

[0046] In this embodiment of the invention, when the readiness level indicates that the autonomous execution strategy is most suitable, parsing is performed only in the set of positions with the job type of intelligent agent; when the readiness level indicates that the human job execution strategy must be retained, parsing is performed only in the set of positions with the job type of human. The parsing path also includes a default job path and a fallback path; the default job path and the fallback path are sorted according to a preset priority and are placed after the candidate job matching path; the default job path is used to return the default job configured for the target role, and is returned only when its state is active or verified; the fallback path is used to map to a pre-configured default job type based on the executor type identifier carried on the target task or role, whereby the executor type identifier indicates one of human, intelligent agent, service, or external system.

[0047] As previously described, the job positions in this embodiment of the invention are a unified abstraction of human-machine isomorphism. Both human executors and AI agent executors are modeled as equivalent entities capable of holding the same job position. Each job position carries a job type identifier (human or AI agent), and the job position is bound to its set of skills. A job position satisfies the skill superset matching condition and can be selected as a qualified candidate by path (b) in step 120 only if the skill set of a job position is a superset of the skill set required by the target role. The AI ​​agent job position is additionally bound to an AI agent specification identifier, which points to its model, hints, and tool configuration. It should be noted that this embodiment of the invention only references the AI ​​agent specification identifier and performs compliance determination on the readiness state of the execution context it points to; it does not involve the generation or assembly of the execution context content.

[0048] In this embodiment of the invention, based on the readiness level, a human-machine collaboration strategy is selected on one hand, and the executor type of the target position returned in step 120 is constrained on the other hand. These two aspects work together to achieve executor parsing where the degree of automation adapts to the evaluation level. Specifically, in one implementation of the invention, it includes: Level 1: The collaboration strategy is fully executed by the agent role; constraint step 120 is resolved only in the set of roles whose role type is agent.

[0049] Level 2: The collaboration strategy is executed by the agent job and reviewed by the human job; constraint step 120 only resolves the job to be executed in the set of jobs with the job type of agent.

[0050] Level 3: The collaboration strategy is human-machine co-pilot collaboration, that is, human-led and intelligent agent-assisted; constraint step 120 resolves a human-led position and an intelligent agent-assisted position to form a co-pilot pair.

[0051] Level 4: The collaboration strategy is to be performed only by human roles; constraint step 120 is resolved only in the set of roles of human type.

[0052] The aforementioned readiness levels determine the default collaboration mode of the collaboration strategy, with the following correspondence: Level 1 corresponds to Independent mode, Level 2 to Drafting and Review mode, Level 3 to Co-pilot mode, and Level 4 to Manual mode. One engineering implementation of this default correspondence maps each level to a fixed default collaboration mode. Based on this default, collaboration modes can be overridden and configured to other orthogonal collaboration modes under the same readiness level according to the task strategy. These collaboration modes are configured independently of the collaboration strategy and are used to define the initiator and escalation path of human-computer interaction under the same readiness level. The collaboration modes at least include Independent, Co-pilot, and Drafting and Review, and may further include Proposal and Approval, and Autonomous and Elevation. The Proposal and Approval mode and the Autonomous and Elevation mode are optional configuration items orthogonal to the readiness level, enabled by the task strategy according to its interaction requirements; for example, for a Level 2 task, in addition to its default Drafting and Review mode, its collaboration mode can be overridden and configured to be the Proposal-Approval mode according to the task strategy. The set of collaboration modes is an example and can be expanded as needed.

[0053] It is necessary to clarify the hierarchical relationship between collaboration strategy and collaboration mode: collaboration strategy is determined by the readiness level, which specifies who leads the execution and who reviews it; collaboration mode is another dimension of configuration orthogonal to collaboration strategy, which specifies the initiator of human-computer interaction and the upgrade path. The two are used in combination.

[0054] Step 140: Perform a graded readiness compliance check on the target position. Each check item is marked as critical, important, or minor, and the check result is used to gate the life cycle state machine of the target position from the verification state to the active state: the position is allowed to migrate to the active state and undertake the target task if and only if all critical check items pass; otherwise, activation is blocked.

[0055] Specifically, for positions classified as agents, a set of agent-specific checks is added, each categorized into critical, important, or minor levels. If any critical-level check fails, the position is deactivated and a blocking message containing the failed check identifier and its reason is returned. If an important or minor-level check fails, an alarm is issued. Figure 4 As shown, when performing critical level checks, if any critical check fails, activation of the target position is blocked, and the check item identifier and reason are returned. If all critical checks pass, important or minor level checks are performed. If an important or minor check fails, an alarm is issued directly, and the target position is activated according to the corresponding strategy. When both important and minor checks pass, the target position is activated directly.

[0056] In this embodiment of the invention, the parsing path identifier used, the selected collaboration strategy and mode, the identifier, status, severity level, and cause of each readiness check item, blocking information, and timestamp are recorded in the audit log using an append-only method. The method also includes persistent storage of job, role, and skill mappings, with the strategy and readiness gates shared within the same job state machine. The elements of this parsing and verification are recorded in the audit log using an append-only method. In one embodiment, the minimum set of fields recorded includes: task identifier, parsing path identifier (one of which is explicit instance assignment, candidate job, default role, or fallback executor type), selected collaboration strategy and mode, identifier, status, severity level, and cause of each readiness check item, blocking information, whether downgraded, and timestamp, for recording.

[0057] This invention embodiment obtains the substitutability assessment results of a target task generated by an external evaluation process; the results include a readiness level; multiple parsing paths are tried sequentially according to a preset priority to determine the target position to undertake the target task; wherein, the multiple parsing paths include at least: instance-level explicit position assignment paths and task-related candidate position matching paths; a target human-machine collaboration strategy is selected from a preset human-machine collaboration strategy pool according to the readiness level, and the position type of the target position is constrained according to the readiness level; the human-machine collaboration strategy pool includes a strategy where the agent position has full authority to execute, a strategy where the agent position executes and is reviewed by a human position, a human-machine co-pilot collaboration strategy, and a strategy where only a human position executes. (Omitted); A graded readiness compliance check is performed on the target position. Each check item is marked with a severity level, which is either critical, important, or minor. The check results are used to gate the lifecycle state machine of the target position from the verification state to the active state. The position is allowed to migrate to the active state and undertake the target task only if all critical check items pass. Otherwise, activation is blocked. This model can model human executors and artificial intelligence agents as equivalent subjects that can perform the same position (human-machine isomorphism). The readiness level generated by the external evaluation process drives the selection of collaboration strategies and constrains the candidate executor types in the position analysis. The graded compliance gate before execution controls the migration of the position's lifecycle state machine to the active state.

[0058] Figure 5 A schematic diagram of the structure of the human-machine collaborative job dynamic allocation device provided in an embodiment of the present invention is shown. Figure 5 As shown, the device 300 includes: The acquisition module 310 is used to acquire the substitutability assessment results of the target task generated by the external assessment process; the results include a readiness level. The parsing module 320 is used to try multiple parsing paths in sequence according to a preset priority in order to determine the target position to undertake the target task; wherein, the multiple parsing paths include at least: an instance-level explicit position assignment path and a task-related candidate position matching path; The strategy selection module 330 is used to select a target human-machine collaboration strategy from a preset human-machine collaboration strategy pool according to the readiness level, and to constrain the job type of the target job according to the readiness level; the human-machine collaboration strategy pool includes a strategy that is fully executed by the intelligent agent job, a strategy that is executed by the intelligent agent job and reviewed by the human job, a human-machine co-pilot collaboration strategy, and a strategy that is executed only by the human job. The graded readiness compliance verification module 340 is used to perform graded readiness compliance verification on the target position. Each check item of the verification is marked with a severity level, which is one of the critical level, important level, or minor level. The verification result is used to gate the life cycle state machine of the target position from the verification state to the active state: the position is allowed to migrate to the active state and undertake the target task if and only if all critical level check items pass; otherwise, activation is blocked.

[0059] In an alternative embodiment, the device further includes: The modeling module is used to model human executors and AI agent executors as equivalent entities capable of performing the same job; wherein, the job is bound to a set of skills required for the role, carries a job type identifier indicating the human or AI agent, and has a lifecycle state machine subject to compliance verification.

[0060] In one alternative approach, the readiness level is taken from a set of discrete levels of increasing artificial preservation. The external evaluation process includes a three-gate sequential short-circuit evaluation of the target task, wherein the three-gate sequential short-circuit evaluation consists of an eligibility gate, a tolerance gate, and a feasibility gate, in that order. The qualification gate is a logical AND operation on a set of Boolean criteria; the tolerance gate is a maximum value on a set of ordinal risk dimensions; the feasibility gate is a minimum value on a set of ordinal readiness dimensions, subject to segmented capping constraints on the task clarity dimension; when any criterion of the qualification gate is negative, the system is short-circuited to the highest manual retention level.

[0061] In one alternative approach, the step of selecting a target human-machine collaboration strategy from a preset pool of human-machine collaboration strategies based on the readiness level, and constraining the job type of the target job based on the readiness level, includes: when the readiness level indicates that the autonomous execution strategy is most suitable, parsing only in the job set where the job type is an agent; when the readiness level indicates that the human job execution strategy must be retained, parsing only in the job set where the job type is a human.

[0062] In one alternative approach, the job lifecycle state machine includes a draft state, a configuration state, a verification state, an activation state, and a suspended state. Among them, when a job in the active state fails a key inspection item, it is forcibly moved to the suspended state. A job in the suspended state can only be moved to the active state after passing all critical level checks.

[0063] In one optional approach, the parsing path further includes a default job path and a fallback path; the default job path and the fallback path are sorted according to a preset priority and are placed after the candidate job matching path; The default job path is used to return the default job configured for the target role, and is only returned when its status is active or verified. The fallback path is used to map to a pre-configured default job type based on the executor type identifier carried on the target task or role, wherein the executor type identifier indicates one of human, intelligent agent, service, or external system.

[0064] The working process of each module in the human-machine collaborative job dynamic allocation device of this invention is largely the same as the specific implementation steps of the aforementioned method embodiment, and will not be repeated here.

[0065] This invention embodiment obtains the substitutability assessment results of a target task generated by an external evaluation process; the results include a readiness level; multiple parsing paths are tried sequentially according to a preset priority to determine the target position to undertake the target task; wherein, the multiple parsing paths include at least: instance-level explicit position assignment paths and task-related candidate position matching paths; a target human-machine collaboration strategy is selected from a preset human-machine collaboration strategy pool according to the readiness level, and the position type of the target position is constrained according to the readiness level; the human-machine collaboration strategy pool includes a strategy where the agent position has full authority to execute, a strategy where the agent position executes and is reviewed by a human position, a human-machine co-pilot collaboration strategy, and a strategy where only a human position executes. (Omitted); A graded readiness compliance check is performed on the target position. Each check item is marked with a severity level, which is either critical, important, or minor. The check results are used to gate the lifecycle state machine of the target position from the verification state to the active state. The position is allowed to migrate to the active state and undertake the target task only if all critical check items pass. Otherwise, activation is blocked. This model can model human executors and artificial intelligence agents as equivalent subjects that can perform the same position (human-machine isomorphism). The readiness level generated by the external evaluation process drives the selection of collaboration strategies and constrains the candidate executor types in the position analysis. The graded compliance gate before execution controls the migration of the position's lifecycle state machine to the active state.

[0066] Figure 6 The diagram shows a structural schematic of a computer device provided in an embodiment of the present invention. The specific embodiments of the present invention do not limit the specific implementation of the computer device.

[0067] like Figure 6 As shown, the computer device may include: a processor 402, a communications interface 404, a memory 406, and a communications bus 408.

[0068] The processor 402, communication interface 404, and memory 406 communicate with each other via communication bus 408. Communication interface 404 is used to communicate with other network elements such as clients or other servers. The processor 402 executes program 410, specifically performing the relevant steps described in the embodiment of the human-machine collaborative job dynamic allocation method.

[0069] Specifically, program 410 may include program code, which includes computer-executable instructions.

[0070] Processor 402 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The computer device includes one or more processors, which may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0071] Memory 406 is used to store program 410. Memory 406 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0072] Specifically, program 410 can be called by processor 402 to cause the computer device to perform the following operations: Obtain the substitutability assessment results for the target task generated by the external evaluation process; the results include a readiness level; Multiple parsing paths are tried sequentially according to a preset priority to determine the target position to undertake the target task; wherein, the multiple parsing paths include at least: instance-level explicit position assignment path and task-related candidate position matching path; Based on the readiness level, a target human-machine collaboration strategy is selected from a preset human-machine collaboration strategy pool, and the job type of the target job is constrained based on the readiness level; the human-machine collaboration strategy pool includes strategies that are fully executed by intelligent agents, strategies that are executed by intelligent agents and reviewed by human agents, strategies that are co-piloted by human agents, and strategies that are executed only by human agents. A graded readiness compliance check is performed on the target position. Each check item is marked with a severity level, which is either critical, important, or minor. The check result is used to gate the life cycle state machine of the target position from the verification state to the active state: the position is allowed to migrate to the active state and assume the target task if and only if all critical check items pass; otherwise, activation is blocked.

[0073] In one alternative approach, the method further includes obtaining an evaluation result of the substitutability of the target task generated by an external evaluation process; prior to the result including a readiness level, the method also includes: Human executors and AI agents are modeled as peer entities capable of performing the same role; wherein, the role is bound to a set of skills required for the role, carries a job type identifier indicating the human or AI agent, and has a lifecycle state machine subject to compliance verification.

[0074] In one alternative approach, the readiness level is taken from a set of discrete levels of increasing artificial preservation. The external evaluation process includes a three-gate sequential short-circuit evaluation of the target task, wherein the three-gate sequential short-circuit evaluation consists of an eligibility gate, a tolerance gate, and a feasibility gate, in that order. The qualification gate is a logical AND operation on a set of Boolean criteria; the tolerance gate is a maximum value on a set of ordinal risk dimensions; the feasibility gate is a minimum value on a set of ordinal readiness dimensions, subject to segmented capping constraints on the task clarity dimension; when any criterion of the qualification gate is negative, the system is short-circuited to the highest manual retention level.

[0075] In one alternative approach, when the readiness level indicates that the autonomous execution strategy is most suitable, the algorithm is resolved only in the set of jobs with the job type of an agent; when the readiness level indicates that the human job execution strategy must be retained, the algorithm is resolved only in the set of jobs with the job type of a human.

[0076] In one alternative approach, the job lifecycle state machine includes a draft state, a configuration state, a verification state, an activation state, and a suspended state. Among them, when a job in the active state fails a key inspection item, it is forcibly moved to the suspended state. A job in the suspended state can only be moved to the active state after passing all critical level checks.

[0077] In one optional approach, the parsing path further includes a default job path and a fallback path; the default job path and the fallback path are sorted according to a preset priority and are placed after the candidate job matching path; The default job path is used to return the default job configured for the target role, and is only returned when its status is active or verified. The fallback path is used to map to a pre-configured default job type based on the executor type identifier carried on the target task or role, wherein the executor type identifier indicates one of human, intelligent agent, service, or external system.

[0078] This invention embodiment obtains the substitutability assessment results of a target task generated by an external evaluation process; the results include a readiness level; multiple parsing paths are tried sequentially according to a preset priority to determine the target position to undertake the target task; wherein, the multiple parsing paths include at least: instance-level explicit position assignment paths and task-related candidate position matching paths; a target human-machine collaboration strategy is selected from a preset human-machine collaboration strategy pool according to the readiness level, and the position type of the target position is constrained according to the readiness level; the human-machine collaboration strategy pool includes a strategy where the agent position has full authority to execute, a strategy where the agent position executes and is reviewed by a human position, a human-machine co-pilot collaboration strategy, and a strategy where only a human position executes. (Omitted); A graded readiness compliance check is performed on the target position. Each check item is marked with a severity level, which is either critical, important, or minor. The check results are used to gate the lifecycle state machine of the target position from the verification state to the active state. The position is allowed to migrate to the active state and undertake the target task only if all critical check items pass. Otherwise, activation is blocked. This model can model human executors and artificial intelligence agents as equivalent subjects that can perform the same position (human-machine isomorphism). The readiness level generated by the external evaluation process drives the selection of collaboration strategies and constrains the candidate executor types in the position analysis. The graded compliance gate before execution controls the migration of the position's lifecycle state machine to the active state.

[0079] This invention provides a computer-readable storage medium storing at least one executable instruction that, when executed on a computer device, causes the computer device to perform the human-machine collaborative job dynamic allocation method described in any of the above method embodiments.

[0080] Executable instructions can be used to cause computer devices to perform the following operations: Obtain the substitutability assessment results for the target task generated by the external evaluation process; the results include a readiness level; Multiple parsing paths are tried sequentially according to a preset priority to determine the target position to undertake the target task; wherein, the multiple parsing paths include at least: instance-level explicit position assignment path and task-related candidate position matching path; Based on the readiness level, a target human-machine collaboration strategy is selected from a preset human-machine collaboration strategy pool, and the job type of the target job is constrained based on the readiness level; the human-machine collaboration strategy pool includes strategies that are fully executed by intelligent agents, strategies that are executed by intelligent agents and reviewed by human agents, strategies that are co-piloted by human agents, and strategies that are executed only by human agents. A graded readiness compliance check is performed on the target position. Each check item is marked with a severity level, which is either critical, important, or minor. The check result is used to gate the life cycle state machine of the target position from the verification state to the active state: the position is allowed to migrate to the active state and assume the target task if and only if all critical check items pass; otherwise, activation is blocked.

[0081] In one alternative approach, the method further includes obtaining an evaluation result of the substitutability of the target task generated by an external evaluation process; prior to the result including a readiness level, the method also includes: Human executors and AI agents are modeled as peer entities capable of performing the same role; wherein, the role is bound to a set of skills required for the role, carries a job type identifier indicating the human or AI agent, and has a lifecycle state machine subject to compliance verification.

[0082] In one alternative approach, the readiness level is taken from a set of discrete levels of increasing artificial preservation. The external evaluation process includes a three-gate sequential short-circuit evaluation of the target task, wherein the three-gate sequential short-circuit evaluation consists of an eligibility gate, a tolerance gate, and a feasibility gate, in that order. The qualification gate is a logical AND operation on a set of Boolean criteria; the tolerance gate is a maximum value on a set of ordinal risk dimensions; the feasibility gate is a minimum value on a set of ordinal readiness dimensions, subject to segmented capping constraints on the task clarity dimension; when any criterion of the qualification gate is negative, the system is short-circuited to the highest manual retention level.

[0083] In one alternative approach, the step of selecting a target human-machine collaboration strategy from a preset pool of human-machine collaboration strategies based on the readiness level, and constraining the job type of the target job based on the readiness level, includes: when the readiness level indicates that the autonomous execution strategy is most suitable, parsing only in the job set where the job type is an agent; when the readiness level indicates that the human job execution strategy must be retained, parsing only in the job set where the job type is a human.

[0084] In one alternative approach, the job lifecycle state machine includes a draft state, a configuration state, a verification state, an activation state, and a suspended state. Among them, when a job in the active state fails a key inspection item, it is forcibly moved to the suspended state. A job in the suspended state can only be moved to the active state after passing all critical level checks.

[0085] In one optional approach, the parsing path further includes a default job path and a fallback path; the default job path and the fallback path are sorted according to a preset priority and are placed after the candidate job matching path; The default job path is used to return the default job configured for the target role, and is only returned when its status is active or verified. The fallback path is used to map to a pre-configured default job type based on the executor type identifier carried on the target task or role, wherein the executor type identifier indicates one of human, intelligent agent, service, or external system.

[0086] This invention embodiment obtains the substitutability assessment results of a target task generated by an external evaluation process; the results include a readiness level; multiple parsing paths are tried sequentially according to a preset priority to determine the target position to undertake the target task; wherein, the multiple parsing paths include at least: instance-level explicit position assignment paths and task-related candidate position matching paths; a target human-machine collaboration strategy is selected from a preset human-machine collaboration strategy pool according to the readiness level, and the position type of the target position is constrained according to the readiness level; the human-machine collaboration strategy pool includes a strategy where the agent position has full authority to execute, a strategy where the agent position executes and is reviewed by a human position, a human-machine co-pilot collaboration strategy, and a strategy where only a human position executes. (Omitted); A graded readiness compliance check is performed on the target position. Each check item is marked with a severity level, which is either critical, important, or minor. The check results are used to gate the lifecycle state machine of the target position from the verification state to the active state. The position is allowed to migrate to the active state and undertake the target task only if all critical check items pass. Otherwise, activation is blocked. This model can model human executors and artificial intelligence agents as equivalent subjects that can perform the same position (human-machine isomorphism). The readiness level generated by the external evaluation process drives the selection of collaboration strategies and constrains the candidate executor types in the position analysis. The graded compliance gate before execution controls the migration of the position's lifecycle state machine to the active state.

[0087] This invention provides a human-machine collaborative job dynamic allocation device for executing the above-described human-machine collaborative job dynamic allocation method.

[0088] This invention provides a computer program that can be invoked by a processor to cause a computer device to execute the human-machine collaborative job dynamic allocation method in any of the above method embodiments.

[0089] This invention provides a computer program product, which includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions, which, when executed on a computer, cause the computer to perform the human-machine collaborative job dynamic allocation method described in any of the above method embodiments.

[0090] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, the embodiments of the present invention are not directed to any particular programming language. It should be understood that the content of the invention described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of the invention.

[0091] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0092] Similarly, it should be understood that, in order to streamline the invention and aid in understanding one or more of the various aspects of the invention, features of the embodiments of the invention are sometimes grouped together in a single embodiment, figure, or description thereof in the above description of exemplary embodiments of the invention. However, this disclosure should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim.

[0093] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0094] It should be noted that the above embodiments are illustrative of the invention and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be construed as limiting the order of execution.

Claims

1. A method for dynamic allocation of human-machine collaborative positions, characterized in that, The method includes: Obtain the substitutability assessment results for the target task generated by the external evaluation process; the results include a readiness level; Multiple parsing paths are tried sequentially according to a preset priority to determine the target position to undertake the target task; wherein, the multiple parsing paths include at least: instance-level explicit position assignment path and task-related candidate position matching path; Based on the readiness level, a target human-machine collaboration strategy is selected from a preset human-machine collaboration strategy pool, and the job type of the target job is constrained according to the readiness level. The human-machine collaboration strategy pool includes strategies for full execution by an intelligent agent, execution by an intelligent agent with review by a human agent, human-machine co-pilot collaboration, and execution by a human agent only. The readiness level is taken from a set of discrete levels with increasing human retention. The readiness level represents the degree of automation of the target task and is generated by an external evaluation process, including: obtaining structured description information of the target task, which at least includes input information, processing logic information, output information, and constraint information of the target task; filling the structured description information into a preset prompt template to obtain model input; the prompt template includes evaluation constraints, which at least indicate the output required by the language model. The system defines preset evaluation dimensions and value types for each evaluation dimension; it provides the model input to the language model, which then generates evaluation results corresponding to the target task under the constraints of the evaluation constraints; the evaluation results include the judgment values ​​of each preset evaluation dimension and the corresponding self-assessment confidence level; the preset evaluation dimensions include at least a first type of dimension belonging to a preset mandatory review category, and a second type of dimension quantified based on evaluation indicators; the external evaluation process includes a three-gate sequential short-circuit evaluation of the target task, wherein the three-gate sequential short-circuit evaluation is, in order, a qualification gate, a tolerance gate, and a feasibility gate; wherein, the qualification gate is a logical AND operation on a set of Boolean criteria; the tolerance gate is the maximum value on a set of ordinal risk dimensions; the feasibility gate is the weakest link on a set of ordinal readiness dimensions, and is subject to the segmented capping constraint of the task clarity dimension; when any criterion of the qualification gate is negative, the system short-circuits to the highest manual retention level; A graded readiness compliance check is performed on the target position. Each check item is marked with a severity level, which is either critical, important, or minor. The check result is used to gate the life cycle state machine of the target position from the verification state to the active state: the position is allowed to migrate to the active state and assume the target task if and only if all critical check items pass; otherwise, activation is blocked.

2. The method according to claim 1, characterized in that, The method further includes obtaining an evaluation of the substitutability of the target task from an external evaluation process; prior to the results including a readiness level, the method also includes: Human executors and AI agents are modeled as peer entities capable of performing the same role; wherein, the role is bound to a set of skills required for the role, carries a job type identifier indicating the human or AI agent, and has a lifecycle state machine subject to compliance verification.

3. The method according to claim 1, characterized in that, The step of selecting a target human-machine collaboration strategy from a preset pool of human-machine collaboration strategies based on the readiness level, and constraining the job type of the target job based on the readiness level, includes: When the readiness level indicates that the autonomous execution strategy is most suitable, it is only resolved in the set of jobs where the job type is an agent. When the readiness level indicator requires the retention of human job execution strategies, it is only resolved in the set of jobs with human job type.

4. The method according to claim 1, characterized in that, The job lifecycle state machine includes a draft state, a configuration state, a verification state, an activation state, and a suspended state. Among them, positions that are in an active state will be forcibly moved to a suspended state when the key inspection items for the position subsequently fail. A job in the suspended state can only be moved to the activated state after passing all critical level checks.

5. The method according to claim 1, characterized in that, The parsing path also includes a default job path and a fallback path; the default job path and the fallback path are sorted according to a preset priority and are placed after the candidate job matching path. The default job path is used to return the default job configured for the target role, and is only returned when its status is active or verified. The fallback path is used to map to a pre-configured default job type based on the executor type identifier carried on the target task or role, wherein the executor type identifier indicates one of human, intelligent agent, service, or external system.

6. A human-machine collaborative job dynamic allocation device, characterized in that, The device includes: The acquisition module is used to acquire the substitutability assessment results for the target task generated by the external evaluation process; the results include a readiness level. The parsing module is used to try multiple parsing paths sequentially according to a preset priority in order to determine the target position to undertake the target task; wherein, the multiple parsing paths include at least: instance-level explicit position assignment paths and task-related candidate position matching paths. The strategy selection module is used to select a target human-machine collaboration strategy from a preset human-machine collaboration strategy pool based on the readiness level, and to constrain the job type of the target job based on the readiness level. The human-machine collaboration strategy pool includes strategies for full execution by the intelligent agent job, strategies for execution by the intelligent agent job and review by the human job, strategies for human-machine co-pilot collaboration, and strategies for execution only by the human job. The readiness level is taken from a set of discrete levels with increasing human retention. The readiness level represents the degree of automation of the target task, and the readiness level is generated by an external evaluation process, including: obtaining structured description information of the target task, the structured description information including at least the input information, processing logic information, output information, and constraint information of the target task; filling the structured description information into a preset prompt template to obtain model input; the prompt template includes evaluation constraints, the evaluation constraints being used to at least instruct the language model. The system requires outputting preset evaluation dimensions and the value types of each evaluation dimension; the model input is provided to the language model, which then generates evaluation results corresponding to the target task under the constraints of the evaluation constraints; the evaluation results include the judgment values ​​of each preset evaluation dimension and the corresponding self-assessment confidence level, wherein the preset evaluation dimensions include at least a first type of dimension belonging to the preset mandatory review category, and a second type of dimension quantified based on evaluation indicators; the external evaluation process includes a three-gate sequential short-circuit evaluation of the target task, wherein the three-gate sequential short-circuit evaluation is, in order, a qualification gate, a tolerance gate, and a feasibility gate; wherein, the qualification gate is a logical AND operation on a set of Boolean criteria; the tolerance gate is a maximum value on a set of ordinal risk dimensions; the feasibility gate is a weak link on a set of ordinal readiness dimensions, and is subject to the segmented capping constraint of the task clarity dimension; when any criterion of the qualification gate is negative, the system short-circuits to the highest manual retention level; The graded readiness compliance verification module is used to perform graded readiness compliance verification on the target position. Each check item of the verification is marked with a severity level, which is one of the critical level, important level, or minor level. The verification result is used to gate the life cycle state machine of the target position from the verification state to the active state: the position is allowed to migrate to the active state and undertake the target task if and only if all critical level check items pass; otherwise, activation is blocked.

7. The apparatus according to claim 6, characterized in that, The device further includes: The modeling module is used to model human executors and AI agent executors as equivalent entities capable of performing the same job; wherein, the job is bound to a set of skills required for the role, carries a job type identifier indicating the human or AI agent, and has a lifecycle state machine subject to compliance verification.

8. A computer device, characterized in that, include: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction, which causes the processor to perform the operation of the human-machine collaborative job dynamic allocation method as described in any one of claims 1-5.

9. A computer-readable storage medium, characterized in that, The storage medium stores at least one executable instruction, which, when executed on a computer device, causes the computer device to perform the operation of the human-machine collaborative job dynamic allocation method as described in any one of claims 1-5.

Images