AI Agent planning enhancement system based on enterprise information architecture (4A), control device and equipment

Abstract

The embodiment of the invention provides an enterprise information architecture (4A)-based AI Agent planning enhancement system, a control device and equipment, and the system comprises a Prompt construction module which is used for analyzing an unstructured demand to obtain a structured demand, and cooperating with other modules to generate a standardized Prompt comprising a business process, a data entity and a tool strategy; the knowledge matching module is used for matching related knowledge based on a knowledge graph employment number fusion knowledge base and sorting the related knowledge into a knowledge list containing business processes, practical experience, a data list and a tool list; and the context reasoning module is used for generating a reasoning result based on the historical data and the knowledge list and feeding back the reasoning result to the Prompt construction module so as to assist in generating the normalized Prompt. According to the system, the integration capability of the AI Agent on business logic and data resources in the task planning process is remarkably improved, so that the AI Agent can more accurately understand enterprise-level business scenes, and standardized and intelligent planning support is provided for practical application of the AI technology in enterprise digital transformation.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of AI Agent planning, and particularly relates to an AI Agent planning enhancement system based on enterprise information architecture (4A), a control device and equipment. BACKGROUND

[0002] As an intelligent entity capable of perceiving the environment, making autonomous decisions and performing actions, AI Agent (Artificial Intelligence Agent) has the ability of memory storage, logical analysis, task decomposition and problem solving, and is regarded as the core path to achieve general artificial intelligence. In the process of digital construction, AI Agent has become a high-efficiency tool for empowering various industries, and its planning module undertakes the key functions of task decomposition, execution sequencing and dynamic adjustment, which directly determines the reliability of the system. Therefore, improving the accuracy and adaptability of the planning module is of great significance to promote the landing application of AI Agent in complex business environment.

[0003] Although the AI Agent system based on large models has developed planning, memory, tool calling and other components, the existing technology still has significant deficiencies: 1) Task target analysis defects: The planning process highly depends on the quality of the Prompt input by the user, but most users lack the ability to construct accurate instructions, resulting in vague task requirement expression and unclear intent, making it difficult for the planning module to effectively analyze the ambiguous target and causing task decomposition deviation and execution direction error; 2) Lack of dynamic knowledge matching: In the face of complex business scenarios, the existing system lacks automatic knowledge retrieval and screening mechanism, and cannot match core resources such as process specifications and business rules in real time, resulting in disconnection between knowledge base content and business scenarios, and the planning scheme is difficult to support effective decision-making; 3) Insufficient integration capability: The existing memory mechanism does not extract historical interaction records and business background information completely, and the key details are missing. The multi-source and heterogeneous context information cannot be effectively integrated, which makes the decision-making basis insufficient, and seriously affects the task execution effect.

[0004] The present application proposes an AI Agent planning enhancement system based on enterprise information architecture (4A) to solve the above problems. SUMMARY

[0005] The application provides an AI Agent planning enhancement system based on enterprise information architecture (4A), a control device and equipment, which is constructed based on enterprise information architecture (4A) and realizes systematic processing of complex business requirements, can efficiently analyze unstructured requirements and generate standardized planning instructions, and significantly improves the integration ability of AI Agent to business logic and data resources in the task planning process with the help of industry data fusion knowledge base matching and context reasoning, so that AI Agent can more accurately understand enterprise-level business scenarios, providing standardized and intelligent planning support for the practical application of AI technology in enterprise digital transformation.

[0006] In a first aspect, an AI Agent planning enhancement system based on enterprise information architecture (4A) is provided, which comprises: A Prompt construction module is configured to analyze unstructured requirements, obtain structured requirements, and generate standardized prompts in cooperation with other modules of the system, wherein the standardized prompts include business processes, data entities and tool strategies. A knowledge matching module is configured to match knowledge related to the structured requirements from an industry data fusion knowledge base based on a knowledge graph, and organize the knowledge into a knowledge list, wherein the knowledge list includes one or more of business processes, business practices, data lists and tool lists. A context reasoning module is configured to generate reasoning results based on historical data and the knowledge list, and feed the reasoning results back to the Prompt construction module to finally generate the standardized prompts.

[0007] It should be understood that the Prompt construction module converts unstructured requirements into structured requirements, providing standardized input for subsequent modules. The knowledge matching module accurately obtains relevant business, data and tool knowledge based on a knowledge graph, avoiding the one-sidedness of single requirement analysis. The context reasoning module complements implicit information and decision logic in complex business scenarios through deep fusion reasoning of historical data and knowledge list. The combination of the three modules enables the system to break through the limitations of traditional AI Agent relying on preset rules or simple prompts to generate standardized prompts containing business processes, data entities and tool strategies, significantly improving the autonomy, accuracy and scenario adaptability of AI Agent planning. Especially in complex tasks across departments and multiple data sources, the system exhibits high planning efficiency that cannot be achieved by traditional single module architecture, resulting in a technical improvement beyond the simple addition of individual module effects.

[0008] In some implementations of the first aspect, the business integration knowledge base is constructed based on a hierarchical integration framework of business architecture (BA) and information architecture (IA), the BA hierarchy includes L1 process class, L2 process group, L3 business process, L4 business activity, and L5 task, and the IA hierarchy includes L1 subject domain, L2 subject, L3 business object, L4 logical entity, and L5 attribute.

[0009] It should be understood that the enterprise information architecture (4A) described in the present application focuses on the integration design of business and information: the BA (business architecture) defines the business logic from L1 process to L5 task (such as business process, role function), the IA (information architecture) specifies the data rules from L1 subject domain to L5 attribute (such as data entity, association relationship), and the two are associated through the "business and data integration classification" to provide a bottom framework of business process and data resources for task planning of AI Agent.

[0010] In some implementations of the first aspect, the Prompt construction module includes: a requirement analysis unit, configured to extract requirements and related key business information from the unstructured requirements to generate structured requirements, the key business information including one or more of business board, role function, skill list, and historical processing experience; a business confirmation unit, configured to confirm business definition and role function based on the key business information; a Prompt generation unit, configured to generate the standardized Prompt based on the knowledge list and the reasoning result.

[0011] It should be understood that the requirement analysis unit extracts key business information such as business board and role function from unstructured requirements, the business confirmation unit further clarifies business definition and role function based on the information, forming a standardized business description, and the Prompt generation unit fuses the output of the first two with the knowledge list and the reasoning result to generate a standardized Prompt containing business process, data entity, and tool strategy. This progressive design breaks through the limitation of traditional NLP requirement analysis which only stays on the surface of text, solves the problems of term ambiguity and role ambiguity in unstructured requirements through the "second definition" of the business confirmation unit, and enables the system to accurately understand the business semantics and responsibility boundaries within the enterprise, especially in cross-department collaboration scenarios, to automatically identify and complete implicit business relationships and data dependencies.

[0012] In some implementations of the first aspect, the knowledge matching module includes: A knowledge graph construction unit, wherein the knowledge graph construction unit maps the internal data in the industry-data fusion knowledge base to knowledge graph nodes; A matching unit performs vectorization processing on the structured requirements and obtains the knowledge that matches the structured requirements with the knowledge graph nodes through a matching algorithm; The knowledge list generation unit organizes the matched knowledge into a knowledge list, which includes one or more of the following: business processes, business practice experience, data list, and tool list.

[0013] In conjunction with the first aspect, in some implementations of the first aspect, the mapping method of the knowledge graph construction unit includes: Extract the content from the Business Architecture (BA) and the Information Architecture (IA) respectively; The elements are categorized and defined as nodes; Associate the nodes of the Business Architecture (BA) with the nodes of the Information Architecture (IA).

[0014] In conjunction with the first aspect, in some implementations of the first aspect, the matching method of the matching unit includes: Vectorize the structured requirements; The similarity between the knowledge in the industry-data fusion knowledge base and the structured requirements is calculated using the cosine similarity algorithm. A dynamic threshold filtering mechanism is used to select knowledge that is highly similar to the structured requirements.

[0015] It should be understood that the knowledge graph construction unit maps the content in the industry-data fusion knowledge base to knowledge graph nodes, the matching unit vectorizes structured requirements and obtains matching knowledge through matching algorithms, and the knowledge list generation unit organizes the matched knowledge into a knowledge list containing business processes, business practice experience, data lists, and tool lists. This design breaks through the limitations of traditional knowledge retrieval based solely on text keyword matching. Through the semantic association of the knowledge graph and vectorized matching algorithms, it can uncover the implicit relationships between requirements and knowledge. Especially when dealing with complex business scenarios, it can automatically identify and associate cross-domain knowledge resources.

[0016] In conjunction with the first aspect, in some implementations of the first aspect, the context reasoning module includes: A historical problem processing unit extracts feature information and processing experience from historical problems, and vectorizes and stores the feature information and processing experience in the industry data fusion knowledge base; The reasoning unit invokes procedural logic, sequential logic, and empirical logic according to priority to perform reasoning and generate the reasoning result. The verification unit organizes the reasoning results into a structured form and performs secondary matching and verification between the reasoning results and the industry data fusion knowledge base to ensure the completeness and accuracy of the reasoning results.

[0017] In conjunction with the first aspect, in some implementations of the first aspect, the reasoning method of the reasoning unit includes: Process-oriented logical reasoning: Based on the L3 business process sequence of the Business Architecture (BA), the business activity nodes in the knowledge list are linked together to generate the first reasoning path; Temporal logical reasoning: Based on the L5 attribute temporal relationship of the information architecture (IA), data verification points and tool call timing are inserted into the first reasoning path to generate a second reasoning path; Empirical logical reasoning: Match historical problem-solving experience vectors, replace or optimize the weights of conflicting nodes in the second path, and generate a third reasoning path.

[0018] It should be understood that the first reasoning path is based on the L3 business process sequence of the Business Architecture (BA), which connects the business activity nodes in the knowledge list according to the preset process logic to form a preliminary task execution sequence framework; the second reasoning path, based on the first path, introduces the L5 attribute sequence relationship of the Information Architecture (IA), inserts data verification points and tool call sequences between business activity nodes, and binds the process with the data operation sequence; the third reasoning path is to replace or optimize the weight of nodes with execution conflicts or inefficiencies in the second path by matching historical problem handling experience vectors, and finally generate an optimized execution path that integrates business processes, data sequences, and historical experience.

[0019] It should be understood that the historical problem processing unit vectorizes the characteristic information and processing experience of historical problems and stores them in the industry-data fusion knowledge base, providing basic data for reasoning; the reasoning unit calls procedural logic, temporal logic, and empirical logic according to priority to perform multi-dimensional reasoning and generate initial reasoning results; the verification unit then structures and verifies the reasoning results with the industry-data fusion knowledge base. This design breaks through the limitations of traditional AI reasoning that relies only on a single logic or static rules. Through the dynamic combination of multi-dimensional logic and result verification, it can automatically identify and correct potential errors in the reasoning process.

[0020] In a second aspect, an AI Agent planning enhancement control device based on an enterprise information architecture (4A) is provided. The control device includes a processor and a memory, the processor being coupled to the memory, the memory being used to store computer programs or instructions, and the processor being used to execute the computer programs or instructions in the memory, such that any implementation of the first aspect is executed.

[0021] Thirdly, an apparatus is provided that includes the control device as described in the second aspect.

[0022] The beneficial effects of the technical solution in this application include: 1. Efficient and accurate requirement processing: Based on the enterprise information architecture (4A) and the business data integration knowledge base, the system automatically transforms unstructured requirements into structured instructions, breaking through the limitations of manual writing. Through intelligent association and adaptation of knowledge and tools, it significantly improves the efficiency of task parsing and the accuracy of planning, reduces execution errors caused by misunderstanding of requirements, and reduces the enterprise's human and time costs. 2. Comprehensive and in-depth knowledge matching: By leveraging the enterprise information architecture (4A) to build a knowledge graph, a deep mapping between business logic and data entities is achieved, accurately mining the relationship between needs and knowledge and tools, avoiding the omission of key information, ensuring that the planning scheme fully covers business scenarios, and improving the utilization efficiency of enterprise-level knowledge resources and the AI ​​Agent's ability to adapt to complex scenarios. 3. Complete and coherent contextual reasoning: Relying on the enterprise information architecture (4A) to integrate multi-source information, the AI ​​Agent automatically completes the business context through multi-dimensional logical reasoning, eliminates information gaps, and provides comprehensive and coherent data support for planning and decision-making. This enables the AI ​​Agent to dynamically optimize the planning path based on complete business logic, effectively improve the accuracy and execution effect of complex task processing, and enhance the stability of intelligent applications. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of an AI Agent planning enhancement system structure based on Enterprise Information Architecture (4A) provided in an embodiment of this application.

[0024] Figure 2 This is a schematic diagram of a Prompt building module structure provided in an embodiment of this application.

[0025] Figure 3 This is a schematic diagram of a knowledge matching module structure provided in an embodiment of this application.

[0026] Figure 4 This is a schematic diagram of a context reasoning module structure provided in an embodiment of this application.

[0027] Figure 5This is a schematic diagram of the operation of an AI Agent planning enhancement system based on Enterprise Information Architecture (4A) provided in an embodiment of this application.

[0028] Figure 6 This is a standardized Prompt diagram provided in an embodiment of this application. Detailed Implementation

[0029] The terminology used in the following embodiments is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to also include expressions such as “one or more,” unless the context clearly indicates otherwise. It should also be understood that in the following embodiments of this application, “at least one” and “one or more” refer to one, two, or more than two. The term “and / or” is used to describe the relationship between related objects, indicating that three relationships may exist; for example, A and / or B can indicate: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character “ / ” generally indicates that the preceding and following related objects are in an “or” relationship.

[0030] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0031] As a core pathway to achieving general artificial intelligence, the planning module of an AI agent undertakes key functions such as task decomposition and dynamic adjustment. Currently, the planning module faces bottlenecks such as vague task objectives, inefficient knowledge matching, and contextual gaps, leading to planning schemes deviating from requirements and high uncertainty in execution results. Although systems based on large models have basic components, significant shortcomings remain: 1) They rely on high-quality prompt input, and vague user descriptions can easily cause deviations in task decomposition; 2) They lack dynamic knowledge matching mechanisms, resulting in a disconnect between the knowledge base and business scenarios; 3) Historical information extraction is incomplete, and the ability to integrate multi-source contexts is insufficient.

[0032] This application proposes an AI Agent planning enhancement system based on Enterprise Information Architecture (4A) to address the aforementioned issues and improve planning accuracy and adaptability.

[0033] The technical solutions of the embodiments of this application will now be described with reference to the accompanying drawings.

[0034] Figure 1 This is a schematic diagram of an AI Agent planning enhancement system architecture based on Enterprise Information Architecture (4A) provided in an embodiment of this application. In some examples, the system includes: The Prompt building module is used to parse unstructured requirements, obtain structured requirements, and work with other modules of the system to generate a standardized Prompt, which includes business processes, data entities, and tool strategies. The knowledge matching module matches knowledge related to the structured requirements based on the knowledge graph industry data fusion knowledge base, and organizes the knowledge into a knowledge list, which includes one or more of the following: business processes, business practice experience, data list, and tool list. The context reasoning module generates reasoning results based on historical data and the knowledge list. The reasoning results are fed back to the Prompt construction module, which finally generates the normalized Prompt.

[0035] In some examples, the business-data integration knowledge base is built on a layered integration framework of business architecture (BA) and information architecture (IA). The BA layer includes L1 process class, L2 process group, L3 business process, L4 business activity, and L5 task. The IA layer includes L1 subject domain, L2 subject, L3 business object, L4 logical entity, and L5 attribute.

[0036] Figure 2 This is a schematic diagram of a Prompt building module structure provided in an embodiment of this application. In some examples, the Prompt building module includes: The requirement parsing unit is used to extract requirements and related key business information from the unstructured requirements to generate structured requirements. The key business information includes one or more of the following: business segments, roles and functions, skill lists, and historical processing experience. A business confirmation unit, which confirms the business definition and role functions based on the key business information; The Prompt generation unit generates the normalized Prompt based on the knowledge list and the reasoning result.

[0037] In one possible implementation, the requirements parsing unit uses natural language processing (NLP) technology to segment, identify entities, and perform semantic analysis on unstructured requirements text, extracting key business information such as business segments and roles. For example, named entity recognition (NID) technology is used to locate business segments and role-related terms such as "sales department" and "customer service representative," and these are mapped to the L3 business process level of the Business Architecture (BA) to generate structured requirements containing requirements objectives and key elements. Based on the extracted key business information, the business confirmation unit retrieves the corresponding BA definitions for business segments from the business data fusion knowledge base, matches the business segments to L2 process groups and L3 business processes, and simultaneously establishes "roles" by associating them with the L4 business activity level in the BA. - The "activity" mapping relationship is established, for example, associating the "customer service representative" role with the "customer inquiry response" business activity, and verifying the accessibility of the data entities required for this business activity in the information architecture (IA) to ensure the consistency between business definitions and data entities; The Prompt generation unit integrates the standardized business definitions output by the business confirmation unit, the knowledge list provided by the role and function and knowledge matching module, and the reasoning results generated by the context reasoning module, and generates standardized instructions according to the preset Prompt template. These instructions include business process steps at the L3-L5 level in BA, data entities and operation attributes at the L3-L4 level in IA, and tool call strategies for the corresponding roles in the tool list, ultimately forming structured instructions that can directly drive the AI ​​Agent to execute task planning.

[0038] Figure 3 This is a schematic diagram of a knowledge matching module structure provided in an embodiment of this application. In some examples, the knowledge matching module includes: A knowledge graph construction unit, wherein the knowledge graph construction unit maps the internal data in the industry-data fusion knowledge base to knowledge graph nodes; A matching unit performs vectorization processing on the structured requirements and obtains the knowledge that matches the structured requirements with the knowledge graph nodes through a matching algorithm; The knowledge list generation unit organizes the matched knowledge into a knowledge list, which includes one or more of the following: business processes, business practice experience, data list, and tool list.

[0039] In one possible implementation, the knowledge graph construction unit uses graph database technology to map the Business Architecture (BA) and Information Architecture (IA) content in the business-data fusion knowledge base into knowledge graph nodes. For example, the L3 business process "order processing" and L4 business activity "inventory verification" of BA, as well as the L3 business object "order information" and L4 logical entity "inventory status" of IA, are defined as graph nodes, and the connections between nodes are established through relational edges, such as "order processing → dependency → inventory verification" and "inventory verification → operation → inventory status". The matching unit first uses a word embedding model to vectorize the structured requirements, and then uses a cosine similarity algorithm to calculate the relationship between the requirement vector and the data. The similarity of knowledge graph node vectors is calculated. For example, when the requirement involves "inventory shortage handling", the similarity between the node vector and nodes such as "inventory verification" and "out-of-stock warning" in the knowledge graph is calculated. Then, a dynamic threshold filtering mechanism is used to select nodes with high similarity and their associated knowledge. Based on the filtering results, the knowledge list generation unit extracts corresponding knowledge from the industry data fusion knowledge base. For example, it extracts the business process steps of "inventory shortage handling" from BA, extracts business practice experience of similar scenarios from the historical case library, extracts data lists such as "inventory status" and "supplier information" from IA, and extracts tool lists such as "inventory management system" and "supplier collaboration tools" from the tool resource library.

[0040] In some examples, the mapping method for the knowledge graph construction unit includes: Extract the content from the Business Architecture (BA) and the Information Architecture (IA) respectively; The elements are categorized and defined as nodes; Associate the nodes of the Business Architecture (BA) with the nodes of the Information Architecture (IA).

[0041] In some examples, the matching method of the matching unit includes: Vectorize the structured requirements; The similarity between the knowledge in the industry-data fusion knowledge base and the structured requirements is calculated using the cosine similarity algorithm. A dynamic threshold filtering mechanism is used to select knowledge that is highly similar to the structured requirements.

[0042] In one possible implementation, structured requirements are vectorized and integrated with an industry-specific knowledge base to calculate cosine similarity. A threshold 'a' is set, combining historical data or a user-defined tolerance, and information with similarity below the threshold is filtered out. The formula for calculating similarity(A, B) is as follows:

[0043] Figure 4This is a schematic diagram of a context reasoning module structure provided in an embodiment of this application. In some examples, the context reasoning module includes: A historical problem processing unit extracts feature information and processing experience from historical problems, and vectorizes and stores the feature information and processing experience in the industry data fusion knowledge base; The reasoning unit invokes procedural logic, sequential logic, and empirical logic according to priority to perform reasoning and generate the reasoning result. The verification unit organizes the reasoning results into a structured form and performs secondary matching and verification between the reasoning results and the industry data fusion knowledge base to ensure the completeness and accuracy of the reasoning results.

[0044] In one possible implementation, the historical issue processing unit extracts feature information such as "customer complaint type" and "processing steps" from historical work orders using natural language processing technology, vectorizes it using a word vector model, and stores it in the business data fusion knowledge base. The reasoning unit first calls the process logic defined by the business architecture (such as the order processing sequence), then combines it with the temporal logic of the data in the information architecture (such as inventory update timestamps), and finally matches it with the empirical logic in the historical issue vector library (such as similar complaint handling solutions) to generate a reasoning result containing the step sequence and risk points. The verification unit organizes the reasoning result into a structured flowchart, compares it a second time with the latest business rules and historical successful cases in the business data fusion knowledge base, corrects conflicting nodes, and outputs the final reasoning result to ensure that the logical chain is complete and conforms to the actual business specifications of the enterprise.

[0045] In some examples, the reasoning method of the reasoning unit includes: Process-oriented logical reasoning: Based on the L3 business process sequence of the Business Architecture (BA), the business activity nodes in the knowledge list are linked together to generate the first reasoning path; Temporal logical reasoning: Based on the L5 attribute temporal relationship of the information architecture (IA), data verification points and tool call timing are inserted into the first reasoning path to generate a second reasoning path; Empirical logical reasoning: Match historical problem-solving experience vectors, replace or optimize the weights of conflicting nodes in the second path, and generate a third reasoning path.

[0046] Optionally, refer to Figure 5The Business Architecture (BA) and Information Architecture (IA) are integrated and classified through business data fusion. The Prompt building module parses requirements to generate a standardized Prompt, the knowledge matching module extracts business, data, and tool knowledge from the business data fusion knowledge base, and the context reasoning module optimizes reasoning based on historical data and knowledge lists, ultimately generating a standardized Prompt.

[0047] Figure 6 This document describes a standardized Prompt template that includes information on roles, business processes, data, and tools. Roles clearly define AI responsibilities, business knowledge describes the workflow, the data list defines data and relationships, and the tool list provides database query tools. Through module collaboration, the system integrates the process logic of the Business Architecture (BA) and the data timeline and historical experience of the Information Architecture (IA) to generate a standardized Prompt. This enables the AI ​​Agent to accurately integrate business logic and data resources, enhancing enterprise-level task planning capabilities and achieving end-to-end enhancement from requirements analysis to knowledge matching and inference optimization, providing intelligent planning support for complex business scenarios.

[0048] This application provides an AI Agent planning enhancement control device based on Enterprise Information Architecture (4A). The control device includes a processor and a memory. The processor is coupled to the memory. The memory is used to store computer programs or instructions. The processor is used to execute the computer programs or instructions in the memory, such that the method described in any of the foregoing embodiments is executed.

[0049] This application also provides an apparatus that includes the control device as described in the above embodiments.

[0050] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. Any equivalent modifications or variations made by those skilled in the art based on the content disclosed in the present invention should be included within the scope of protection set forth in the claims.

Claims

1. An AI Agent planning enhancement system based on Enterprise Information Architecture (4A), characterized in that, The system includes: The Prompt building module is used to parse unstructured requirements, obtain structured requirements, and work with other modules of the system to generate a standardized Prompt, which includes business processes, data entities, and tool strategies. The knowledge matching module matches knowledge related to the structured requirements based on the knowledge graph industry data fusion knowledge base, and organizes the knowledge into a knowledge list, which includes one or more of the following: business processes, business practice experience, data list, and tool list. The context reasoning module generates reasoning results based on historical data and the knowledge list. The reasoning results are fed back to the Prompt construction module, which finally generates the normalized Prompt.

2. The system according to claim 1, characterized in that, The business-data fusion knowledge base is built on a layered fusion framework of Business Architecture (BA) and Information Architecture (IA). The BA layer includes L1 process class, L2 process group, L3 business process, L4 business activity, and L5 task. The IA layer includes L1 subject domain, L2 subject, L3 business object, L4 logical entity, and L5 attribute.

3. The system according to claim 1, characterized in that, The Prompt building module includes: The requirement parsing unit is used to extract requirements and related key business information from the unstructured requirements to generate structured requirements. The key business information includes one or more of the following: business segments, roles and functions, skill lists, and historical processing experience. A business confirmation unit, which confirms the business definition and role functions based on the key business information; The Prompt generation unit generates the normalized Prompt based on the knowledge list and the reasoning result.

4. The system according to claim 1, characterized in that, The knowledge matching module includes: A knowledge graph construction unit, wherein the knowledge graph construction unit maps the internal data in the industry-data fusion knowledge base to knowledge graph nodes; A matching unit performs vectorization processing on the structured requirements and obtains the knowledge that matches the structured requirements with the knowledge graph nodes through a matching algorithm; The knowledge list generation unit organizes the matched knowledge into a knowledge list, which includes one or more of the following: business processes, business practice experience, data list, and tool list.

5. The system according to claim 4, characterized in that, The mapping method for the knowledge graph construction unit includes: Extract the content from the Business Architecture (BA) and the Information Architecture (IA) respectively; The elements are categorized and defined as nodes; Associate the nodes of the Business Architecture (BA) with the nodes of the Information Architecture (IA).

6. The system according to claim 4, characterized in that, The matching method of the matching unit includes: Vectorize the structured requirements; The similarity between the knowledge in the industry-data fusion knowledge base and the structured requirements is calculated using the cosine similarity algorithm. A dynamic threshold filtering mechanism is used to select knowledge that is highly similar to the structured requirements.

7. The system according to claim 1, characterized in that, The context reasoning module includes: A historical problem processing unit extracts feature information and processing experience from historical problems, and vectorizes and stores the feature information and processing experience in the industry data fusion knowledge base; The reasoning unit invokes procedural logic, sequential logic, and empirical logic according to priority to perform reasoning and generate the reasoning result. The verification unit organizes the reasoning results into a structured form and performs secondary matching and verification between the reasoning results and the industry data fusion knowledge base to ensure the completeness and accuracy of the reasoning results.

8. The system according to claim 7, characterized in that, The reasoning method of the reasoning unit includes: Process-oriented logical reasoning: Based on the L3 business process sequence of the Business Architecture (BA), the business activity nodes in the knowledge list are linked together to generate the first reasoning path; Temporal logical reasoning: Based on the L5 attribute temporal relationship of the information architecture (IA), data verification points and tool call timing are inserted into the first reasoning path to generate a second reasoning path; Empirical logical reasoning: Match historical problem-solving experience vectors, replace or optimize the weights of conflicting nodes in the second path, and generate a third reasoning path.

9. An AI Agent planning enhancement control device based on Enterprise Information Architecture (4A), characterized in that, The control device includes a processor and a memory, the processor being coupled to the memory, the memory being used to store computer programs or instructions, and the processor being used to execute the computer programs or instructions in the memory such that the method as described in any one of claims 1 to 8 is performed.

10. A device, characterized in that, The device includes the control unit as described in claim 9.

Images