An agent organization automatic construction method and system based on panoramic HR atlas

Abstract

This invention discloses a method and system for automated construction of intelligent agent organizations based on a panoramic HR map, belonging to the technical field of enterprise-level AI intelligent agent platform architecture construction. The method analyzes a panoramic HR map reflecting human resource information, constructs multiple workstation-based office space nodes with hierarchical relationships based on reporting relationships, creates decision modules and workstations within each node (the decision module serves as a collaboration interface, and the workstation is used to bind the executor intelligent agent), generates task contexts for workstations based on AI models, and constructs a self-context for the executor intelligent agent to form its identity recognition and decision-making basis, establishes data uplink channels and file downlink distribution channels based on hierarchical relationships, and establishes an authorization link where higher-level nodes have command-issuing permissions to lower-level nodes, lower-level nodes have data insight permissions to higher-level nodes, and data is isolated between nodes at the same level by default. This invention achieves automated and accurate mapping of organizational structure and "create-and-use" intelligent agents.

Description

Technical Field

[0001] This invention relates to the field of automated construction technology of organizational structure for enterprise-level AI intelligent agent platforms, and in particular to an automated construction method and system for intelligent agent organizations based on a panoramic HR map. Background Technology

[0002] In the construction and application of enterprise-level artificial intelligence (AI) agent platforms, a core aspect is accurately mapping the real-world organizational structure of an enterprise to a digital agent organization. Existing agent platforms typically employ a relatively flat "agent team" model for this mapping, simulating organizational structure by assigning different role labels (such as manager, employee, etc.) to agents. However, this approach struggles to fully replicate the complex and strictly hierarchical organizational entity model found in real enterprises.

[0003] Specifically, the existing technology has the following main drawbacks: First, the organizational model is disconnected from the real-world architecture. Existing technologies lack a structured distinction between organizational units, decision-making cores, and execution units. The intelligent agent organizations created differ significantly from the actual organizational structures of enterprises, which have clear hierarchies and reporting relationships, leading to low efficiency in subsequent management and collaboration.

[0004] Secondly, newly created agents lack immediate usability. After creation, agents typically only possess basic role identities and permissions, lacking structured definitions of their specific job responsibilities and the ability to perceive the context of the organizational environment. This makes the agent essentially an "empty shell," unable to understand and execute its tasks from the outset, requiring extensive manual configuration and debugging.

[0005] Secondly, the permission configuration and organization building process is disconnected from the company's existing human resources (HR) system. Administrators typically need to manually create and configure the agent organizational structure and permission relationships. This process is not only tedious and error-prone, but also fails to utilize the comprehensive HR map data that the company has already accumulated, which includes a complete organizational tree and reporting relationships, resulting in data silos and duplication of effort.

[0006] Therefore, existing technologies have problems such as organizational model mismatch, insufficient intelligent agent initialization, reliance on manual construction process, and inability to automate the use of HR data when mapping enterprise organizational structure to digital intelligent agent organization. It is difficult to efficiently and accurately build a complex intelligent agent organization that is "ready to use upon creation". Summary of the Invention

[0007] The main objective of this invention is to provide an automated construction method and system for intelligent agent organizations based on a panoramic HR map. This aims to solve the technical problems in the prior art when mapping enterprise organizational structure to digital intelligent agent organizations, such as the disconnect between the organizational model and the real structure, the lack of task definition and context awareness after the intelligent agent is created, and the disconnect between the construction process and the enterprise HR system and the reliance on manual configuration.

[0008] To achieve the above objectives, this invention provides an automated construction method for intelligent agent organizations based on a panoramic HR graph, comprising the following steps: acquiring and parsing a panoramic HR graph reflecting human resource information; constructing multiple workstation-based office space nodes with hierarchical relationships based on the reporting relationships therein; creating decision modules and workstations within each workstation-based office space node; wherein the decision module is used for decision analysis and serves as the external collaboration interface for the workstation-based office space node, and the workstation is an organizational placeholder node for binding corresponding executor intelligent agents as needed; generating a corresponding task context for each workstation based on a large language model (e.g., models based on the Transformer architecture and pre-trained on a large scale, such as the GPT series, BERT, etc., to enhance support); and constructing the ego that acts on the executor intelligent agent bound to the workstation. The system comprises several contexts: a task context for defining tasks for the executor agent, and an ego context for determining the agent's identity, forming the basis for identity recognition and decision-making. It establishes a hierarchical data uplink channel from the bottom-level workstations to the top-level workstations, and a file downlink channel from the top-level workstations to the bottom-level workstations. An authorization link is also established based on this hierarchical structure, where higher-level workstations have the authority to issue instructions to lower-level workstations, and lower-level workstations have access to data insights from higher-level workstations. The executor agent inherits the authorization relationship of its assigned workstation, and data is isolated by default between workstations at the same level.

[0009] Optionally, in the step of acquiring and parsing the panoramic HR map reflecting human resources information and constructing multiple workstation-based office space nodes with hierarchical relationships, a topological sorting algorithm can be used to automatically construct multiple workstation-based office space nodes with hierarchical relationships based on reporting relationships. Furthermore, the constructed workstation-based office space nodes can be initialized with functional partitions, resulting in the following non-hierarchical topological relationships: a clone area for carrying the personal information and digital identity of the workstation-based office space node owner; a management area for carrying the management configuration of the decision-making module and the workstation-based office space node-level settings; a workstation area for carrying the execution agent intelligence of all workstations under this workstation-based office space node; and an inspiration area for carrying the creative generation and knowledge exploration functions of the decision-making module. Through functional partitioning, the internal structure of the workstation-based office space nodes becomes clearer, facilitating management and functional expansion.

[0010] Optionally, the panoramic HR map includes the following fields: fields for the workstation-based office space node category, namely, workstation-based office space node name field, workstation-based office space node type field, parent workstation-based office space node identifier field, and department code field; fields for the workstation category, namely, workstation name field, workstation type field, workstation responsibility description field, and the identifier of the workstation-based office space node to which it belongs; and fields for the executor agent category, namely, executor agent name field, executor agent role type field, and bound workstation identifier field. Accordingly, the steps for creating decision modules and workstations may specifically include: for the creation of decision modules, creating a decision module within each workstation-based office space node and loading the corresponding strategy template based on the workstation-based office space node type field to obtain the corresponding capabilities; for the creation of workstations, creating workstations in batches within each workstation-based office space node based on the workstation category field under that workstation-based office space node in the panoramic HR map. By utilizing the detailed fields in the HR map, the automated and refined creation of decision modules and workstations is achieved.

[0011] Optionally, in the step of generating the corresponding task context, a task context in a corresponding task relationship graph format can be generated for each workstation based on the following information: a job description obtained based on the panoramic HR graph, the type and hierarchical position of the workstation-based office space node to which it belongs, historical task relationship graph templates for workstations of the same type, and strategic documents of the superior workstation-based office space node. The task relationship graph can contain the following standard node structure: a task meta-node representing the starting point of task deduction, used to reflect the basic and core description of user needs; task nodes, used as intermediate or final results in the task deduction process; task deduction conditions connecting task nodes or task meta-nodes, used to reflect the logical reasons for the reasoning; and operable nodes, used to reflect the user's expected phased or final demands. By generating a structured task relationship graph, the executor agent can clearly understand its task objectives and execution path.

[0012] Optionally, the construction of the intrinsic context for the executor agent bound to the workstation can include the following five layers: Organizational Context Layer: used to automatically inherit the organizational goals, strategic direction, and departmental positioning of the superior workstation-based office space node based on the hierarchical position of the workstation to which the executor agent belongs; File System Layer: used to automatically mount the file directories related to the executor agent's responsibilities; DTV Data Layer: used to configure the data dimensions and indicators to be monitored according to the workstation type; Permission Context Layer: automatically sets the data range and operation permissions that the executor agent can access based on the authorization link; Workstation-based Office Space Node-Specific Context: records the identifier of the workstation-based office space node to which the executor agent belongs, as well as the collaboration strategy of that workstation-based office space node. By constructing a multi-layered intrinsic context, the executor agent possesses rich environmental awareness and identity recognition capabilities, providing a solid foundation for its decision-making.

[0013] Optionally, in the steps of establishing the data uplink channel and the file downlink distribution channel, the data uplink aggregation method based on the data uplink channel may include: the decision module, as the data user, uses and parses the data aggregated by the executor agents (data acquisition units) within its workstation-based office space node, to achieve data aggregation for executor agents that do not have communication channels with each other; the data aggregated by the workstation-based office space node is passed upwards along the hierarchical structure by its decision module to the decision module of the superior workstation-based office space node. The downlink file distribution method based on the file downlink distribution channel may include: based on the hierarchical structure, the decision module distributes files from the superior workstation-based office space node to its directly subordinate subordinate workstation-based office space nodes; wherein, the executor agents at the workstations within the workstation-based office space nodes obtain files by directly referencing them or by referencing the support of the decision module. This design clarifies the data and file flow path, ensuring the orderliness and isolation of information transmission.

[0014] Optionally, based on the authorization chain, the executor agent inherits the authorization relationship of the workstation-based office space node to obtain data visibility to the superior workstation-based office space node. This ensures automatic inheritance of permissions, simplifies configuration, and guarantees compliance of data access.

[0015] Optionally, after establishing the authorization link, the process may further include: cross-validating the established authorization link with the reporting relationships contained in the panoramic HR map to verify the correctness of the authorization link, and generating a verification report for administrator review. Cross-validation can effectively identify and correct potential errors in the organizational structure building process, improving the accuracy and reliability of the building results.

[0016] Optionally, the method further includes: using a large language model to perform a pre-inspection on the construction results, the pre-inspection including at least one of the following: hierarchical depth and management span checks, permission configuration consistency checks, data path integrity checks, task context coverage checks, and compliance checks on isolation between executor agents within the same workstation-based office space node; and outputting optimization suggestions based on the pre-inspection results. Through AI-assisted pre-inspection, the rationality and completeness of the constructed agent organization can be evaluated from multiple dimensions, and optimization directions can be provided, further improving the quality of automated construction.

[0017] This invention also provides an automated construction system for intelligent agents based on a panoramic HR graph, comprising: a hierarchical construction module for acquiring and parsing a panoramic HR graph reflecting human resource information, and constructing multiple workstation-based office space nodes with hierarchical structures based on the reporting relationships therein; an intelligent agent construction module for creating decision modules and workstations within the workstation-based office space nodes; wherein the decision module is used for decision analysis and serves as the external collaboration interface for the workstation-based office space nodes, and the workstation is an organizational placeholder node for binding corresponding executor intelligent agents as needed; a context generation module for generating a corresponding task context for each workstation based on a large language model; and constructing an intrinsic context acting on the executor intelligent agents bound to the workstations; wherein the task context is used to define tasks for the executor intelligent agents. The ego context is used to determine the identity of the executor agent, forming the basis for the executor agent's identity cognition and decision-making; the channel establishment module is used to establish data uplink channels from the bottom-level workstations to the top-level workstations based on hierarchical relationships, and file downlink distribution channels from the top-level workstations to the bottom-level workstations; the authorization establishment module is used to establish authorization links based on hierarchical relationships, where higher-level workstations have the authority to issue instructions to lower-level workstations, and lower-level workstations have the authority to gain data insights from higher-level workstations; the executor agent inherits the authorization relationship of its affiliated workstations, and data is isolated between workstations at the same level by default.

[0018] Compared with existing technologies, this invention has the following beneficial effects: 1. It achieves automated and accurate mapping of organizational structure: By directly parsing the panoramic HR map and constructing hierarchical workstation-based office space nodes based on reporting relationships, this invention can automatically and accurately map the real organizational structure of an enterprise into a three-level digital entity architecture containing organizational units, decision-making cores, and execution units, solving the problem of the disconnect between digital organization and real organizational structure in existing technologies. 2. It achieves "ready to use upon creation" for intelligent agents: By automatically generating structured task contexts (such as task relationship graphs) and multi-dimensional self-contexts for each workstation, newly created executor intelligent agents immediately possess task understanding, environmental awareness, and identity recognition capabilities after being bound to a workstation, eliminating the need for cumbersome manual secondary configuration and significantly improving deployment efficiency. 3. It achieves full-process automation of organizational construction and permission configuration: This invention integrates HR map parsing, organizational hierarchy construction, data channel establishment, and authorization link generation into a unified automated process, and automatically establishes authorization relationships using reporting relationships in the HR map, eliminating the burden and potential error risks of manually creating organizations and configuring permissions, and achieving seamless integration with the enterprise HR system. 4. Ensures isolation and orderly collaboration within the organization: By creating a decision module as a unified collaboration interface within the workstation-based office space nodes and establishing hierarchical data uplink and file downlink channels, this invention ensures data isolation between workstation-based office space nodes at the same level and between different executor agents within the same node. At the same time, it standardizes cross-level and cross-departmental collaboration and data flow paths, improving the security and compliance of organizational operations. Attached Figure Description

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

[0020] Figure 1 This is a flowchart illustrating an automated construction method for intelligent agent organization based on a panoramic HR map according to an embodiment of the present invention.

[0021] Figure 2 This is a structural block diagram of an intelligent agent organization automation construction system based on a panoramic HR map according to an embodiment of the present invention.

[0022] Figure 3 This is a schematic diagram of the node hierarchy structure of a workstation-based office space according to an embodiment of the present invention.

[0023] In the diagram: 101 - Obtain a panoramic HR map; 102 - Construct workstation-based office space nodes; 103 - Create decision-making modules and workstations; 104 - Generate context; 105 - Establish channels; 106 - Establish authorization links; 107 - Cross-validation; 108 - AI pre-check; 109 - Output intelligent agent organization; 200 - Intelligent agent organization automated construction system; 201 - Panoramic HR map; 202 - Hierarchical construction module; 203 - Intelligent agent construction module; 204 - Context generation module; 205 - Channel establishment module; 206 - Authorization establishment module; 210 - Intelligent agent organization; 301 - CEO Office; 302 - CTO Office; 303 - CMO Office; 304 - Architecture group; 305 - Development group; 306 - Testing group. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this application. Unless otherwise defined, all technical and scientific terms used in the embodiments of this application have the same meaning as commonly understood by those skilled in the art. The terminology used in the embodiments of this application is for the purpose of describing the embodiments of this application only and is not intended to limit this application. Before further detailed description of the embodiments of this application, the nouns and terms involved in the embodiments of this application are explained, and the nouns and terms involved in the embodiments of this application are subject to the following interpretations.

[0025] (1) Panoramic HR Map: refers to a data structure that reflects human resources information. It includes at least reporting relationship information for building organizational hierarchical relationships, and may include relevant fields such as workstation-based office space node categories, workstation categories, and executor intelligent agent categories.

[0026] (2) Workstation-based office space node: refers to a basic organizational unit corresponding to a position in the real-world organizational structure of an enterprise. It can contain decision-making modules and multiple workstations, and can form a hierarchical structure in the organization. It can also be called "Office".

[0027] (3) Decision module: This refers to the collective intelligence engine set up inside the workstation-based office space node, which is used for decision analysis and task arrangement, and serves as the only external collaboration interface for the workstation-based office space node. It can also be called "DeepBrain".

[0028] (4) Workstation: refers to the organizational occupancy node set inside the workstation-based office space node. It is a persistent task definition unit used to bind the corresponding executor intelligent agent instance as needed.

[0029] (5) Executor intelligent agent: refers to an artificial intelligence instance that acts as an independent task execution unit and is responsible for executing the tasks defined by the workstation after being bound to it. It can also be called "Agent".

[0030] (6) Task Context: This refers to a data structure used to define the tasks of an agent, such as a task relationship graph. It can also be called "Mission Context".

[0031] (7) EgoContext: refers to a data structure used to determine the identity of an agent, which is the basis for its identity cognition and decision-making. It can contain multi-dimensional information such as organizational context, file system, and data permissions. It can also be called "EgoContext".

[0032] (8) Authorization Link: This refers to a permission relationship chain established based on the hierarchical structure between nodes in a workstation-based office space, defining the instruction-issuing permissions of the superior node and the data insight permissions of the subordinate node. It can also be called "TrustLink".

[0033] (9) DTV Data Layer: DTV stands for "Data flow, Trust link, Value point," and is one of the components of the Ego Context. It is used to configure the data dimensions and metrics that need to be focused on based on the workstation type (e.g., sales workstations focus on revenue data, and production workstations focus on yield data). The data in this layer comes from the execution data transmitted by agents at all workstations under this Office. It is aggregated into the DTV dataset of this Office through Trust Link authorization, providing the executor agents with real-time perception of business reality.

[0034] Please see Figure 1 and Figure 2 This application aims to provide an automated construction scheme for intelligent agent organizations, addressing issues such as structural disconnect, insufficient initialization of agent capabilities, lack of collaboration and isolation mechanisms, disconnection from HR systems, and poor data flow in existing technologies when mapping real-world organizational structures to digital intelligent agent organizations. This scheme, through automated processes, transforms an enterprise's human resource information into a structurally complete, clearly defined, and data-connected intelligent agent organization.

[0035] In one specific implementation, the method first acquires and parses a panoramic HR map reflecting human resources information, and then constructs multiple workstation-based office space nodes with hierarchical relationships based on the reporting relationships within it. This step aims to visualize and digitize the organizational structure tree of the real world, laying the foundation for subsequent intelligent agency organization construction. By parsing the hierarchical relationships in HR data, the command and reporting chains within the enterprise can be automatically and accurately reproduced, avoiding the tediousness and errors of manual configuration.

[0036] Next, within the established workstation-based office space nodes, decision-making modules and workstations are created. The decision-making module is designed as a collective intelligence engine, responsible for the node's decision analysis and serving as its external collaboration interface. This solves the problem of lacking a unified command and coordination center in traditional flat intelligent agent teams. Simultaneously, workstations are created as organizational placeholder nodes for subsequent on-demand binding of executor intelligent agents, decoupling the organizational structure from specific execution units and enhancing organizational flexibility.

[0037] Subsequently, based on a large language model, the method generates a corresponding task context for each workstation and constructs an ego context for the executor agent bound to that workstation. The task context clearly defines the executor agent's work tasks and goals, while the ego context gives the executor agent a clear identity, including its position within the organization, permissions, and the data it should focus on. Through this step, the newly created executor agent is no longer a functionally empty "shell," but rather a "ready-to-use" entity with preliminary working capabilities.

[0038] To facilitate information flow within the organization, this method further establishes a data uplink channel from workstations at the bottom level to those at the top level, and a document downlink channel from those at the top level to those at the bottom level, based on hierarchical relationships. The data uplink channel ensures that lower-level execution data is accessible to senior management, while the document downlink channel guarantees that top-level strategies and instructions are effectively communicated to the execution end, thus achieving data connectivity within the organization.

[0039] Finally, an authorization chain is established based on hierarchical relationships. This authorization chain clearly defines that higher-level workstation-based office space nodes have the authority to issue instructions to lower-level workstation-based office space nodes, while lower-level nodes have data insight permissions to higher-level nodes. Simultaneously, the executor agent inherits the authorization relationship of its parent node, and data isolation is implemented by default between nodes at the same level. This design ensures the effectiveness of the command chain while also guaranteeing data security within the organization through a default isolation strategy, resolving the problems of chaotic permissions and disordered collaboration among agents in existing technologies.

[0040] Furthermore, in a preferred embodiment, to more accurately construct the organizational hierarchy, a topological sorting algorithm can be used to automatically construct the workstation-based office space nodes based on reporting relationships. This algorithm ensures that parent nodes are always created before their child nodes, thereby guaranteeing the correctness of the hierarchical structure and the orderly nature of the construction process. Simultaneously, each constructed workstation-based office space node is initialized with functional partitions, resulting in an agent area, management area, workstation area, and inspiration area. These partitions respectively carry different functions such as node owner information, management configuration, agent operating space, and creative exploration, making the internal structure of the nodes clearer and the functions more complete. However, there is no hierarchical relationship between these partitions, and they do not affect the main organizational structure determined by reporting relationships.

[0041] In another preferred embodiment, the panoramic HR map can include richer field information, such as: for workstation-based office space node categories, the workstation-based office space node name field, the workstation-based office space node type field, the parent workstation-based office space node identifier field, and the department code field; for workstation categories, the workstation name field, the workstation type field, the workstation job description field, and the workstation-based office space node identifier field; and for executor agent categories, the executor agent name field, the executor agent role type field, and the bound workstation identifier field. Based on these detailed fields, the process of creating decision modules and workstations can be more automated and precise. For example, the creation of decision modules can load corresponding strategy templates based on the workstation-based office space node type field to obtain specific capabilities; while the creation of workstations can be achieved in batches and automatically based on the workstation category field information under that node in the map.

[0042] In one alternative implementation, the process of generating task context for each workstation can be further refined. Specifically, based on various information such as the job description obtained from the panoramic HR map, the type and hierarchical position of the workstation-based office space node, historical task relationship map templates for similar workstations, and strategic documents from higher-level workstation-based office space nodes, a task context in a task relationship graph format can be generated for each workstation. This task relationship graph can contain a standardized node structure, such as: task meta-nodes as the starting point of the task, task nodes as intermediate or final results, task derivation conditions reflecting logical reasoning, and actionable nodes reflecting the user's final demands. Through this structured task definition, the executor agent can more deeply understand the ins and outs and ultimate goals of the task.

[0043] Furthermore, the self-context of the executor intelligent agent can be concretized into a five-layer structure. The first layer is the organizational context layer: enabling it to automatically inherit the organizational goals and strategic direction of its superior nodes. The second layer is the file system layer: automatically mounting responsibility-related file directories for it. The third layer is the DTV data layer: configuring the data dimensions and metrics it needs to focus on. The fourth layer is the permission context layer: automatically setting its data access and operation permissions based on the authorization chain. The fifth layer is the workstation-based office space node-specific context: recording its node identifier and collaboration strategy. These five layers together constitute the executor intelligent agent's complete identity cognition, ensuring its behavior is highly aligned with organizational goals.

[0044] In a preferred embodiment, the data uplink and file downlink channel mechanism can be implemented more specifically. During data uplink aggregation, the decision module, as the data user, uses the data aggregated by the executor agents within its node (acting as data acquisition units) and passes it up the hierarchical structure to the decision module of the upper-level node. This enables data aggregation to be completed even without a direct communication channel between the executor agents. During file downlink distribution, files are distributed from the upper-level node to the lower-level node through the decision module. The executor agents in the lower-level nodes obtain files by directly referencing them or by referencing methods supported by the decision module, ensuring the orderliness and controllability of file distribution.

[0045] In another preferred embodiment, the executor agent can gain data visibility to its superior workstation-based office space nodes by inheriting the authorization link relationship of its subordinate workstation-based office space node. This inheritance mechanism ensures the automatic transfer and consistency of permissions, enabling lower-level execution units to understand the upper-level business data and background within the scope of their permissions, thereby making more reasonable decisions and actions.

[0046] To ensure the accuracy of the build process, a verification step can be added after establishing the authorization link. Specifically, this involves cross-validating the established authorization link with the original reporting relationships contained in the panoramic HR map. By comparing the consistency between the two, potential configuration errors can be identified and located, such as discrepancies between authorization direction and reporting relationship. A verification report is then generated for administrator review, thereby improving the reliability of the entire automated build process.

[0047] Furthermore, to further improve the quality of the build process, this method can include a pre-check step. Before formally creating the organization, the pre-built result is checked using a large language model. This check may include verifying the reasonableness of hierarchical depth and management span, the consistency of permission configurations, the completeness of data pathways, the adequacy of task context coverage, and the compliance of isolation between executor agents within the same node. Based on the pre-check results, the system can output optimization suggestions to assist administrators in making adjustments before final creation, thereby ensuring that the final generated agent organization achieves optimal performance in terms of structure, permissions, data, and compliance.

[0048] Please see Figure 2 This application also provides an automated construction system 200 for intelligent agent organizations based on a panoramic HR map. This system aims to materialize the above-mentioned methods, achieving automated transformation from human resource data to usable intelligent agent organizations. System 200 includes a hierarchical construction module 202, an intelligent agent construction module 203, a context generation module 204, a channel establishment module 205, and an authorization establishment module 206.

[0049] Specifically, in one embodiment, an IT administrator can export a panoramic HR map 201 containing information such as organizational structure, positions, and personnel from the company's human resources system (such as an enterprise communication and collaboration platform) and provide it as input to this system 200. For example, the map can be a JSON file.

[0050] First, execute step S1, corresponding to Figure 1 The system acquires the panoramic HR map 101 and constructs workstation-based office space nodes 102. The hierarchical construction module 202 is responsible for acquiring and parsing the uploaded panoramic HR map 201. For example, the system parses an organizational structure containing 3 top-level workstation-based office space nodes (e.g., CEO Office 301, CTO Office 302, CMO Office 303), 6 workstation groups (e.g., Architecture Group 304, Development Group 305, Testing Group 306, etc.), and 29 specific workstations. During parsing, if key fields such as "workstation responsibility description" are missing from the map, the large language model can automatically supplement them based on job title, department, and other information. Subsequently, the hierarchical construction module 202 automatically constructs multiple workstation-based office space nodes with hierarchical relationships based on the reporting relationships in the map, using a topological sorting algorithm to form an Office tree consistent with the real organization. Figure 3As shown, a hierarchical structure is constructed with CEO Office 301 as the root node, overseeing CTO Office 302 and CMO Office 303. Simultaneously, each newly created workstation-based office space node is initialized with functional partitions, such as a personal area, management area, workstation area, and inspiration area, and is assigned a corresponding strategy template based on its type (e.g., strategy, management, execution). For example, CEO Office 301, as a strategy-type node, receives a strategic planning template.

[0051] Next, step S2 is executed, corresponding to Figure 1 The system creates a decision-making module and workstation 103. Within each workstation-based office space node created by the hierarchical construction module 202, the intelligent agent construction module 203 automatically creates a decision-making module instance and loads the corresponding strategy template based on the node's type. For example, it creates a CTO-DeepBrain instance for CTO Office 302 and loads a team management template, enabling it to perform responsibilities such as technical task orchestration and cross-team coordination. Simultaneously, the intelligent agent construction module 203 creates workstation entities in batches within the corresponding workstation-based office space nodes based on the workstation information defined in the panoramic HR map 201. These workstations serve as persistent task definition units, awaiting subsequent binding with executor intelligent agents. This design ensures that executor intelligent agents within the same workstation-based office space node operate independently and do not directly collaborate; all cross-workstation collaborations are routed and managed through the upper-level decision-making module.

[0052] Then, step S3 is executed, corresponding to Figure 1 The context generation module 204 utilizes a large-scale artificial intelligence model to automatically generate a task context and an ego context for each workstation created by the agent building module 203. For example, for the "backend development engineer" workstation, the context generation module 204 will generate a structured task relationship graph (MRG) based on information such as the job description, the hierarchical position in the CTO Office 302, and the strategic documents in the CEO Office 301. This graph includes task meta-nodes (such as "Technology R&D Domain - Backend Engineering"), task nodes, task derivation conditions, and actionable nodes (such as "Daily Code Submission Record"). As a preferred embodiment of MRG construction, the MRG is first initialized based on knowledge resources on the Internet. The solution can be based on existing scenarios or combinations of scenarios and applicable users, along with statistical results from product categories and brands, to uncover relationships within the internet; further analysis can be conducted... Continuously updated.

[0053] Simultaneously, a five-layered Ego Context is constructed for the executor agent bound to this workstation, including: an organizational context layer inherited from the CTO Office 302 strategic goals; a file system layer that automatically mounts the code repository; a DTV data layer that monitors code commit volume and bug fix rate; a permission context layer that sets read and write permissions for the code repository; and a workstation-specific field context that records the node identifier to which it belongs. In this way, the newly created executor agent acquires a complete task definition and identity awareness. The construction of the Ego Context follows these mechanisms: inheritance mechanism, inheriting the organizational-level context from the Ego Context of the parent Office link; enhancement mechanism, enhancing specific domain knowledge based on workstation type and MRG definition; and customization mechanism, fine-tuning based on personalized configurations in the HR graph.

[0054] Subsequently, step S4 is executed, corresponding to Figure 1 The channel establishment module 205 is responsible for configuring the data and file flow paths. It establishes an upward data channel from the bottom-level workstations to the top-level workstation-based office space nodes, ensuring that DTV data (such as code commit records) generated by the executor agents can be aggregated level by level to the decision modules (such as CTO-DeepBrain) of their respective workstation-based office space nodes, and can be further passed to higher levels (such as CEO-DeepBrain). For example, aggregation functions such as SUM, AVG, and MAX can be configured during upward data aggregation. Simultaneously, the channel establishment module 205 also establishes a downward file distribution channel, allowing strategic documents from higher-level nodes to be distributed level by level. For example, the downward file distribution process is: CEOOffice strategic document → CTO and CMO Office DeepBrain → each workstation agent. The executor agents obtain files by directly referencing them or by referencing the support of the decision modules, rather than the decision modules actively distributing them.

[0055] Next, step S5 is executed, corresponding to Figure 1The authorization link establishment module 206 automatically establishes a Trust Link authorization link based on the hierarchical structure established by the hierarchical construction module 202. For example, establishing an authorization relationship: CEOOffice → authorizes → CTO Office, indicating that CEO Office 301 has the authority to issue instructions to CTO Office 302 and can view its aggregated DTV data. At the same time, CTO Office 302 and CMO Office 303 are sibling nodes with default data isolation, represented as: CTO Office ↔ CMO Office. Workstations and the executor intelligent agents bound to them will automatically inherit the authorization relationship of their respective workstation-based office space nodes. For example, CTO Office → authorizes → Architecture Group, Development Group, Testing Group. CMO Office → authorizes → Brand Group, Growth Group, Customer Success Group.

[0056] In a preferred embodiment, the system can also perform verification and pre-inspection steps. After the authorization link is established, the authorization establishment module 206 can perform cross-validation 107, comparing the generated authorization link with the original reporting relationship in the panoramic HR map 201 to generate a verification report. Before all construction steps are completed, the system can also perform AI pre-inspection 108, checking the organizational structure rationality, permission configuration consistency, DTV data path integrity, MRG coverage, and isolation compliance of the construction results, and output optimization suggestions, such as "It is recommended to add a DevOps (Development and Operations Integration) workstation in the CTO Office" or "The confidence level of the automatically generated MRG for two workstations in the Customer Success Group is less than 70%, and manual review is recommended."

[0057] Finally, after administrator confirmation, the system completed the creation of all entities and output a fully structured, data-connected, and clearly defined intelligent agent organization 210, realizing the automated construction from HR data to a usable intelligent agent organization.

Claims

1. A method for automated construction of intelligent agent organizations based on a panoramic HR map, characterized in that, Includes the following steps: S1: Obtain and analyze a panoramic HR map reflecting human resources information, and construct multiple workstation-based office space nodes with hierarchical structure based on the reporting relationships within it; S2: Within the workstation-based office space node, create a decision module and workstations; the decision module is used for decision analysis and serves as the external collaboration interface for the workstation-based office space node, while the workstation is an organizational placeholder node used to bind corresponding executor intelligent agents as needed; S3: Based on a large language model, generate a corresponding task context for each workstation; and construct an ego context that acts on the executor agent bound to the workstation; wherein, the task context is used to define the task for the executor agent, and the ego context is used to determine the identity of the executor agent, so as to form the basis for the executor agent's identity cognition and decision-making. S4: Based on the hierarchical structure, establish a data uplink channel from the workstations of the bottom-level workstation-based office space nodes to the top-level workstation-based office space nodes, and a file downlink distribution channel from the top-level workstation-based office space nodes to the workstations of the bottom-level workstation-based office space nodes. S5: Based on the hierarchical structure, an authorization link is established, in which the superior workstation-based office space node has the authority to issue instructions to the subordinate workstation-based office space node, and the subordinate workstation-based office space node has the authority to gain data insight to the superior workstation-based office space node; the executor intelligent agent inherits the authorization relationship of its subordinate workstation-based office space node, and data is isolated between workstation-based office space nodes of the same level by default.

2. The method for automated construction of intelligent agent organization based on panoramic HR map according to claim 1, characterized in that, In step S1: Based on the reporting relationships, a topological sorting algorithm is used to automatically construct multiple workstation-based office space nodes with hierarchical relationships; The constructed workstation-based office space nodes are initialized with functional partitions, resulting in the following hierarchical topology: a clone area for carrying the personal information and digital identity of the workstation-based office space node owner; a management area for carrying the management configuration of the decision-making module and the workstation-based office space node-level settings; a workstation area for carrying the execution agent intelligence of all workstations under this workstation-based office space node; and an inspiration area for carrying the creative generation and knowledge exploration functions of the decision-making module.

3. The method for automated construction of intelligent agent organizations based on panoramic HR maps according to claim 1, characterized in that, The panoramic HR map includes the following fields: fields for workstation-based office space node categories, namely, workstation-based office space node name, workstation-based office space node type, parent workstation-based office space node identifier, and department code; fields for workstation categories, namely, workstation name, workstation type, workstation job description, and the identifier of the workstation-based office space node to which it belongs; and fields for executor intelligent agent categories, namely, executor intelligent agent name, executor intelligent agent role type, and bound workstation identifier. Step S2 specifically includes: Decision module creation: Within each workstation-based office space node, a decision module is created, and the corresponding strategy template is loaded based on the workstation-based office space node type field to obtain the corresponding capabilities; Batch creation of workstations: Within each workstation-based office space node, workstations are created in batches based on the workstation category field under that workstation-based office space node in the panoramic HR map.

4. The method for automated construction of intelligent agent organization based on panoramic HR map according to claim 1, characterized in that, In step S3: Based on the following information, a corresponding task context in task relationship graph format is generated for each workstation, including: Based on the panoramic HR map, the job description, the type and hierarchical position of the workstation node, the historical task relationship map template of the same type of workstation, and the strategic documents of the superior workstation node. The task relationship graph contains the following standard node structure: The task meta node, which is the starting point of the task deduction, is used to reflect the basic and core description of user needs; Task nodes are used as intermediate or final results in the task simulation process; Task derivation conditions used to connect task nodes or task meta nodes, reflecting the logical reasons for the reasoning; Actionable nodes are used to reflect the user's expected stage or final demands.

5. The method for automated construction of intelligent agent organization based on panoramic HR map according to claim 1, characterized in that, The ego context for the executor agent bound to the workstation is constructed, including the following five layers: The organizational context layer is used to automatically inherit the organizational goals, strategic direction, and departmental positioning of the superior workstation-based office space node based on the hierarchical position of the workstation to which the executor intelligent agent belongs. File system layer: Used to automatically mount the file directories related to the responsibilities of the executor agent; The DTV data layer is used to configure the data dimensions and metrics that need to be monitored based on the workstation type. Permission Context Layer: Based on the authorization chain, automatically sets the data range and operation permissions that the executor agent can access; Workstation-based office space node-specific context: Records the identifier of the workstation-based office space node to which the executor agent belongs, as well as the collaboration strategy of that workstation-based office space node.

6. The method for automated construction of intelligent agent organization based on panoramic HR map according to claim 1, characterized in that, In step S4, the data uplink aggregation method based on the data uplink channel includes: The decision module, as the user of the data, uses and parses the data collected by the executor intelligent agents within its workstation-based office space node, which act as data acquisition units, to achieve data aggregation among executor intelligent agents that have no communication channels with each other; the data aggregated by the workstation-based office space node is passed up the hierarchical structure by its decision module to the decision module of the upper-level workstation-based office space node. Downlink file distribution methods based on file downlink distribution channels include: Based on a hierarchical structure, documents are distributed from higher-level workstation-based office space nodes to their directly subordinate workstation-based office space nodes through a decision-making module. Among them, the executor agents at the workstations within the workstation-based office space nodes obtain documents by directly referencing them or by referencing the support of the decision-making module.

7. The method for automated construction of intelligent agent organization based on panoramic HR map according to claim 1, characterized in that, Based on the authorization link, the executor agent inherits the authorization relationship of the authorization link of the workstation-based office space node, which is used to obtain data visibility to the superior workstation-based office space node.

8. The method for automated construction of intelligent agent organization based on panoramic HR map according to claim 1, characterized in that, Step S5 further includes: The established authorization links are cross-validated with the reporting relationships contained in the panoramic HR map to verify the correctness of the authorization links, and a verification report is generated for administrator review.

9. The method for automated construction of intelligent agent organization based on panoramic HR map according to claim 1, characterized in that, The method also includes: using a large language model to perform a pre-inspection on the construction results, which includes at least one of the following: hierarchical depth and management span checks, permission configuration consistency checks, data path integrity checks, task context coverage checks, and compliance checks on isolation between executor agents within the same workstation-based office space node; and outputting optimization suggestions based on the pre-inspection results.

10. An automated intelligent agent organization construction system based on a panoramic HR map, characterized in that, include: The hierarchical construction module is used to acquire and parse a panoramic HR map that reflects human resources information, and based on the reporting relationships therein, construct multiple workstation-based office space nodes with hierarchical structure. The intelligent agent building module is used to create decision modules and workstations within the workstation-based office space node. The decision module is used for decision analysis and serves as the external collaboration interface for the workstation-based office space node. The workstation is an organizational placeholder node used to bind the corresponding executor intelligent agent as needed. The context generation module is used to generate corresponding task contexts for each workstation based on a large language model; and to construct the ego context for the executor agent bound to the workstation. The task context is used to define the task for the executor agent, and the ego context is used to determine the identity of the executor agent, so as to form the basis for the executor agent's identity cognition and decision-making. The channel establishment module is used to establish a data uplink channel from the workstations of the bottom-level workstation-based office space nodes to the top-level workstation-based office space nodes based on the hierarchical structure relationship, as well as a file downlink distribution channel from the top-level workstation-based office space nodes to the workstations of the bottom-level workstation-based office space nodes. The authorization establishment module is used to establish authorization links based on hierarchical structure relationships. In this module, higher-level workstation-based office space nodes have the authority to issue instructions to lower-level workstation-based office space nodes, while lower-level workstation-based office space nodes have the authority to gain data insight from higher-level workstation-based office space nodes. The executor agent inherits the authorization relationship of its affiliated workstation-based office space node, and data is isolated between workstation-based office space nodes at the same level by default.

Images