Business execution support system, business execution support method, and business execution support program

Abstract

To provide an operating environment that allows business objectives to be easily achieved. [Solution] A business execution support system comprising: a request receiving unit that receives natural language text containing business execution details; an overview planning unit that analyzes the text and instructs a specified generation AI service to generate an overview plan containing execution targets and execution order of a specified business program; a detailed planning unit that acquires a group of data related to the business as integrated data, and instructs the generation AI service to concretize and update the overview plan according to the content of the integrated data and generate a detailed plan; and a data manipulation unit that instructs the generation AI service to generate command information for executing the execution targets of the business program according to the execution order included in the detailed plan, and executes the business program using the command information.

Description

[Technical Field]

[0001] The present invention relates to a business execution support system, a business execution support method, and a business execution support program. [Background technology]

[0002] Various innovations have been made to support business operations. In recent years, improvements in the performance of smartphones and other devices, along with the expansion of communication infrastructure, have made it easier to manipulate business data. The operations being performed vary widely, including the manipulation of sales and marketing management information, the manipulation of production management information, communication via groupware and chat, the management of electronic medical records, or a combination of these, and the scope of the operations being performed is also diverse.

[0003] For example, the business support device disclosed in Patent Document 1 is characterized by having a purchasing data storage control unit that creates purchasing data in a memory unit, associating product information including the number of received products with a unique lot number for inventory management of the received products and a lot tracing number formed by adding a subnumber to the lot number; an inventory management unit that, when the product is moved or processed, creates an inventory file indicating the number of products moved or processed and the number obtained by subtracting the number of products moved or processed from the number of received products, and stores the created inventory data in the memory unit; and a state transition correspondence storage control unit that creates state transition data in which the number of moved or processed products is associated with the same product information as at the time of receiving the product, the same lot number as at the time of receiving the product, and a lot tracing number formed by adding a subnumber that is consecutive to the branch number at the time of receiving the product to the same lot number as at the time of receiving the product, and stores the state transition data together with the purchasing data and inventory file in the memory unit. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2023-034524 Summary of the Invention [Problem to be solved by the invention]

[0005] For example, to achieve a business objective by operating existing business software, an operator must actually imagine the necessary software operations and the order in which they should be performed, operate them quickly and without error, and coordinate all the necessary related processes. In other words, there are various difficulties, such as the need for expertise in operating the software according to the business content, and the need for an environment and time for operation.

[0006] SUMMARY OF THE INVENTION The present invention has been made in view of the above points, and has as its object to provide an operating environment that allows business objectives to be easily achieved. [Means for solving the problem]

[0007] The present application includes multiple means for solving at least part of the above-mentioned problems, examples of which are as follows: A business execution support system according to one aspect of the present invention comprises: a request receiving unit that receives natural language text including the execution content of a business; a general planning unit that analyzes the text and instructs a predetermined generation AI service to generate a general plan including execution targets and an execution order of a predetermined business program; a detailed planning unit that acquires a group of data related to the business as integrated data, and instructs the generation AI service to concretize and update the general plan according to the content of the integrated data and generate a detailed plan; and a data manipulation unit that instructs the generation AI service to generate command information for executing the execution targets of the business program according to the execution order included in the detailed plan, and executes the business program using the command information.

[0008] The business execution support system may further comprise a status reporting unit that records the execution status of the business program by the data manipulation unit and presents the status for reference.

[0009] The above business execution support system may also include a consistency processing unit that instructs the generation AI service to verify the consistency of the execution results of the business program by the data manipulation unit.

[0010] In addition, in the above-mentioned business execution support system, if the consistency verification result indicates an inconsistency, the consistency processing unit may discard the data changes made by the execution of the business program by the data manipulation unit and return to the state before execution.

[0011] In the above business execution support system, the general planning unit may obtain information on the API (Application Programming Interface) of the business program via an MCP (Model Context Protocol) server that stores information on the API of the business program, and use the information to generate the general plan.

[0012] In the above business execution support system, the detailed planning unit may denormalize a group of data related to the business and convert it into a predetermined integrated data format of a text description type to create the integrated data.

[0013] In the above business execution support system, the detailed planning unit may obtain a set of data related to the business by using an MCP (Model Context Protocol) token including information specifying a predetermined authority accessible to the user who provided the text.

[0014] Furthermore, in the above-mentioned business execution support system, when the request receiving unit receives natural language text including the execution content of the business, it may register the text in a FIFO asynchronous queue, and the overview planning unit may receive the text from the asynchronous queue.

[0015] The business execution support system may also include a status reporting unit that analyzes the execution status of the business program by the data manipulation unit and presents a report in natural language that includes at least the execution content and execution results of the business for reference.

[0016] In addition, another aspect of the present invention is a business execution support method that uses a computer system, wherein the computer system performs the following steps: a request receiving step that receives natural language text including the execution content of the business; an overview planning step that analyzes the text and instructs a predetermined generation AI service to generate an overview plan including execution targets of a predetermined business program and their execution order; a detailed planning step that acquires a group of data related to the business as integrated data, and instructs the generation AI service to concretize and update the overview plan according to the content of the integrated data and generate a detailed plan; and a data manipulation step that instructs the generation AI service to generate command information for executing the execution targets of the business program according to the execution order included in the detailed plan, and executes the business program using the command information.

[0017] In addition, another aspect of the present invention is a business execution support program that causes a computer system to support business execution, and is characterized in that the computer system executes the following steps: a request receiving step for receiving natural language text including the execution content of the business; an overview planning step for instructing a predetermined generation AI service to analyze the text and generate an overview plan including execution targets and execution order of a predetermined business program; a detailed planning step for acquiring a group of data related to the business as integrated data, and instructing the generation AI service to concretize and update the overview plan according to the content of the integrated data and generate a detailed plan; and a data manipulation step for instructing the generation AI service to generate command information for executing the execution targets of the business program according to the execution order included in the detailed plan, and executing the business program using the command information. [Effects of the Invention]

[0018] According to the present invention, an operating environment that allows business objectives to be easily achieved can be provided.

[0019] Problems, configurations, and effects other than those described above will become apparent from the following description of the embodiments. [Brief explanation of the drawings]

[0020] [Figure 1] 1 is an example of a block diagram of a business execution support system according to an embodiment; [Figure 2] FIG. 10 is a diagram illustrating an example of a data structure of access right information. [Figure 3] FIG. 10 is a diagram illustrating an example of a data structure of data operation specification information. [Figure 4] FIG. 2 illustrates an example of a hardware configuration of a task execution support device. [Figure 5] FIG. 2 illustrates an example of the hardware configuration of an operator terminal. [Figure 6] FIG. 10 is a diagram illustrating an example of the flow of a request reception process. [Figure 7] FIG. 10 is a diagram illustrating an example of a flow of detailed planning processing. [Figure 8] FIG. 10 is a diagram illustrating an example of the flow of an execution control process. [Figure 9] FIG. 10 is a diagram illustrating an example of the flow of a result recording process. [Figure 10] FIG. 10 illustrates an example of an instruction analysis prompt. [Figure 11] FIG. 10 illustrates an example of a blueprint prompt. [Figure 12] FIG. 10 is a diagram illustrating an example of an execution prompt. [Figure 13] FIG. 10 illustrates an example of a matching prompt. [Figure 14] FIG. 10 illustrates an example of a report creation prompt. DETAILED DESCRIPTION OF THE INVENTION

[0021] A task execution support system 1 to which an embodiment according to one aspect of the present invention is applied will be described below with reference to the drawings.

[0022] By using the business execution support system 1 according to this embodiment, business objectives can be achieved without the need for proficiency in software operation according to the business content, and without wasting time and the environment for operation. Specifically, the operator does not need to imagine the software operations and their sequence, and can easily achieve the coordination of all related necessary processes without omission.

[0023] 1 is a block diagram of a business execution support system according to an embodiment of the present invention. The business execution support system 1 includes a business execution support device 100, a generation AI service 200, and an operator terminal 300.

[0024] The operator terminal 300 is a terminal used by a user of the business execution support system 1, for example, a user who executes a business or an administrator related to the use and management of the business. The operator terminal 300 is communicably connected to the business execution support device 100 via a network 50.

[0025] The network 50 is a communication network such as a LAN (Local Area Network), a WAN (Wide Area Network), the Internet, a mobile phone network, etc. The network 50 may also be a VPN (Virtual Private Network) on a wireless communication network such as a mobile phone communication network.

[0026] The business execution support device 100 is a so-called server device, but is not limited to this, and may be various information processing devices such as a personal computer, a smartphone, a workstation, a PDA (Personal Data Assistant), or a tablet device.

[0027] The business execution support device 100 receives a request for a business to be executed from the operator terminal 300 via the network 50, generates an appropriate execution plan for the business processing in accordance with the request, executes the business processing in accordance with the plan, verifies the consistency of the execution results, and prepares a report on the execution results.

[0028] At this time, the business execution support device 100 delegates the creation, execution, consistency check, and report creation of an execution plan based on the business processing request to the generation AI service 200 via the network 50.

[0029] The operator terminal 300 performs output processing when it receives a predetermined response from the task execution support device 100 via the network 50. In addition, the operator terminal 300 transmits input information received from an input device or the like (to be described later) to the task execution support device 100 via the network 50.

[0030] The operator terminal 300 includes a processing unit 310, an input receiving unit 320, a display unit 330, and a communication unit 340. The processing unit 310 includes a request processing unit 311.

[0031] The request processing unit 311 receives input of information on the request content via the input receiving unit 320, and transfers the information on the request content (textual task instructions) to the task execution support device 100. Furthermore, upon receiving a predetermined response from the task execution support device 100, the request processing unit 311 outputs the response via the display unit 330.

[0032] The input receiving unit 320 receives input from an operator of the operator terminal 300, for example, a user or an administrator, and passes the input to the request processing unit 311. The input receiving unit 320 receives text from a hardware keyboard, a software keyboard, or the like, but is not limited to this. For example, the input receiving unit 320 may receive voice input from a microphone to identify text, or may identify text by eye-gaze input, or the like. When the display unit 330 receives screen information in a predetermined format instructed to be displayed by the task execution support device 100 from the communication unit 340, the display unit 330 draws the screen information on the display device. The communication unit 340 communicates with the task execution support device 100 via the network 50.

[0033] The generation AI service 200 is a service that provides the functions of so-called generation AI, such as GPT, Gemini, and Claude, via an API (Application Programming Interface). The generation AI service 200 provides commands (prompts) in natural language to the generation AI to generate desired results. The generation AI service 200 includes an AI processing unit 210 and a communication unit 220.

[0034] In this embodiment, when the generation AI service 200 receives an instruction via, for example, an API, it causes the generation AI to generate answer information and sends the result to the instruction source as the API return value. At that time, the generation AI accepts information related to the answer, adds that information, reads it as part of the information to be considered, and reflects it in generating the answer.

[0035] For example, when the generation AI service 200 receives an instruction to execute a certain business operation, it receives API details via an MCP (Model Context Protocol) server to which the API (preferably compliant with the OpenAPI standard) of the required business process is pre-associated, and plans the extraction and execution order of the APIs required to execute the instruction. For example, if the instruction to execute the business operation is "please discount the estimated cost of the X system construction project for company A by 20%, the generation AI service 200 extracts the current estimated cost of the X system construction project for company A, obtains API specification information from an MCP server (not shown) provided in the business operation execution support device 100, and creates a plan to generate new quotation data with a 20% discount.

[0036] Then, as a task that requires consistency, the generation AI service 200 obtains API specification information from the MCP server and incorporates it into the plan so that the sales amount planned in the sales management system is updated with a 20% discount.

[0037] The business execution support device 100 includes a processing unit 110, a storage unit 120, and a communication unit 130. The processing unit 110 includes a request receiving unit 111, a general planner 112, an integrated data generator 113, a detailed planner 114, a data manipulation unit 115, a consistency processor 116, and a status report unit 117. The storage unit 120 includes access right information 121, business information 122, and data manipulation specification information 123. Note that the data structures and data management methods described below are merely examples and are not intended to be limiting. For example, structured data management methods such as a relational database, a graph database, a vector database, text data, or HTML (HyperText Markup Language) may also be used.

[0038] 2 is a diagram showing an example of the data structure of access right information. The access right information 121 stores a handleable data range 121b and specific operation authority 121c for the handleable data range 121b, in association with a user identifier 121a.

[0039] The user identifier 121a is information for distinguishing a user from other users. Here, the user is a user or administrator of the above-mentioned business execution support system 1. The user is granted authority and has access restrictions for use and management of the business execution support system 1 based on the user identifier 121a.

[0040] The handleable data range 121b is information that specifies the range of customers and data that can be handled by the user identified by the user identifier 121a. For example, it is assumed that the business information 122 is assigned a predetermined identifier in advance according to a combination of the customer and the data type, and one or more of the identifiers are stored in the handleable data range 121b.

[0041] The operation authority 121c is information that specifies the operations that a user identified by the user identifier 121a can or cannot perform on data within the range specified by the available data range 121b. For example, the operation authority 121c is information that specifies, using an operation identifier, which operations are permitted and which are not permitted among data operations such as referencing, creating a new record (adding a record), changing (updating a record), and deleting. Each of these data operations is implemented as an API on the OpenAPI.

[0042] FIG. 3 is a diagram illustrating an example of the data structure of data operation specification information. The data operation specification information 123 is stored in such a manner that a description 123b is associated with each operation identifier 123a that identifies an operation on data included in the business information 122. The description 123b includes various information about the operation identified by the operation identifier 123a. For example, the description 123b includes information such as parameters required for the operation, options for customizing the operation, a description and usage method of the operation, prerequisite operations, and subsequent operations. Such description 123b conforms to the OpenAPI specification. In other words, the data operation specification information 123 is associated in advance with an MCP server provided in the business execution support device 100 and can be referenced by other processing units and the AI ​​processing unit 210 via the MCP server.

[0043] Returning to the explanation of FIG. 1 , the request receiving unit 111 receives, via the network 50, text serving as a task instruction from a user, which is received via the request processing unit 311 of the operator terminal 300. Specifically, the request receiving unit 111 receives text in natural language including the execution content of the task from the operator terminal 300. Furthermore, the request receiving unit 111 acquires information including the range of data that can be handled by referring to the access right information 121 for a user who is in a usage state (logged in), creates a predetermined token (e.g., a 36-digit shortened MCP token, etc.), and stores the token in the token. The request receiving unit 111 then registers the text and the token in a FIFO (First In First Out) asynchronous queue (not shown). The asynchronous queue is provided in advance in the task execution support device 100. Furthermore, although a FIFO is illustrated as the asynchronous queue, it is not limited to a FIFO and various mechanisms for realizing asynchronous processing may be used.

[0044] The general planning unit 112 obtains the text and tokens received by the request receiving unit 111 from the asynchronous queue. The general planning unit 112 then analyzes the text and instructs the generation AI service 200 to generate a general plan including the execution targets (APIs corresponding to the data operations to be executed) of the specified business program and the execution order. Specifically, the general planning unit 112 analyzes the text, creates an instruction analysis prompt for creating a general plan, and instructs the generation AI service 200. The general planning unit 112 obtains API information for the business processing (data operation program) via an MCP server that holds API information for the business program, and passes it to the generation AI service 200 to use in generating the general plan. However, this is not limited to this, and the generation AI service 200 may obtain API specifications by referencing data operation specification information 123 via the MCP server, and generate a general plan.

[0045] The integrated data generation unit 113 functions as part of the processing of the detailed plan unit 114. The integrated data generation unit 113 extracts data from the business information 122 that is included in the range defined by the handleable data range 121b for the logged-in user, denormalizes the data as a data group related to the business processing to be executed, and converts it into a predetermined integrated data format of text description type (for example, JSON (JavaScript Object Notation) format or XML (Extensible Markup Language) format) to create integrated data.

[0046] The detailed planning unit 114 instructs the integrated data generation unit 113 to acquire a group of data related to the business as integrated data. Then, the detailed planning unit 114 instructs the generation AI service 200 to concretize and update the overview plan according to the contents of the integrated data and generate a detailed plan. Specifically, the detailed planning unit 114 creates a detailed planning prompt to review the overview plan including the integrated data and instructs the generation AI service 200. The detailed planning unit 114 extracts a group of data related to the business from the business information 122 using an MCP token containing information identifying the specified permissions accessible by the user who provided the text that serves as the business instructions.

[0047] The data manipulation unit 115 instructs the generation AI service 200 to generate command information for executing the execution targets of the business program according to the execution order included in the detailed plan, and controls the execution of the business program (calling the API corresponding to the data operation to be executed) using the command information generated by the generation AI service 200. Specifically, the data manipulation unit 115 refers to past execution results, creates an execution prompt that generates command information (information necessary to call the API), and instructs the generation AI service 200.

[0048] The consistency processing unit 116 creates a consistency prompt to verify the consistency of the execution results of the business program by the data manipulation unit 115, and instructs the generation AI service 200. If the consistency verification result indicates inconsistency, the consistency processing unit 116 discards the data changes made by the execution of the business program by the data manipulation unit 115, and returns to the state before execution.

[0049] The status reporting unit 117 records and presents for reference the execution status of the business program by the data manipulation unit 115. The status reporting unit 117 also analyzes the execution status of the business program by the data manipulation unit 115, and presents for reference a report in natural language that includes at least the execution content and execution results of the business.

[0050] The communication unit 130 communicates with other devices, such as the generation AI service 200 and the operator terminal 300, via the network 50.

[0051] 4 is a diagram showing an example of the hardware configuration of a business task execution support device 100. The business task execution support device 100 includes a processor 101, a memory 102, a storage 103, a communication device 104, and a bus 105 connecting the devices. The business task execution support device 100 may also include an input device.

[0052] The processor 101 is an arithmetic device such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit), and executes processing in accordance with a program recorded in the memory 102 or the storage 103. The program is, for example, an application program that can be executed on an OS (Operating System) program. In the business execution support device 100, processing is performed by the processor 101 that operates in accordance with a program read out onto the memory 102 or the storage 103. The processing unit 110, the request receiving unit 111, the general planning unit 112, the integrated data generating unit 113, the detailed planning unit 114, the data operating unit 115, the consistency processing unit 116, and the status reporting unit 117 each realize their respective functions by the processor 101 executing the program.

[0053] The memory 102 is a storage device such as a RAM (Random Access Memory) or a flash memory, and functions as a storage area from which programs and data are temporarily read. The storage 103 is a writable and readable storage device. The functions of the storage unit 120, the access right information 121, the business information 122, and the data operation specification information 123 are realized by the memory 102 or the storage 103. Note that the functions of the storage unit 120 may be realized by a storage device connected via the communication device 104.

[0054] The communication device 104 is an interface for connecting the task execution support device 100 to an external device for communication. For example, the communication device 104 is a network card that performs wired communication via a wired connection. Alternatively, the communication device 104 performs wireless communication using an antenna that can use a predetermined radio wave (e.g., 5 GHz band, 2.4 GHz band, etc.) that establishes a connection with the generation AI service 200 and the operator terminal 300 according to the Wi-Fi standard.

[0055] The processing of each component of the task execution support device 100 may be executed by one piece of hardware or by multiple pieces of hardware. Furthermore, the processing of each component of the task execution support device 100 may be realized by one program or by multiple programs.

[0056] The communication unit 130 of the task execution support device 100 is realized by the communication device 104. An example of the hardware configuration of the task execution support device 100 has been described above.

[0057] Each configuration of the task execution support device 100 can be further classified into more components depending on the processing content, or each component can be classified so as to perform more processes.

[0058] Furthermore, each processing unit (processing unit 110, request receiving unit 111, general planning unit 112, integrated data generation unit 113, detailed planning unit 114, data manipulation unit 115, consistency processing unit 116, and status reporting unit 117) may be constructed using dedicated hardware (ASIC, GPU, etc.) that realizes each function. Furthermore, the processing of each processing unit may be executed by a single piece of hardware or by multiple pieces of hardware.

[0059] 5 is a diagram showing an example of the hardware configuration of an operator terminal 300. The operator terminal 300 includes an input device 301, a processor 302, a storage 303, a memory 304, a display 305, a communication device 306, and a bus 307 connecting the devices together.

[0060] The input device 301 is one of various input devices such as a keyboard, a mouse, a touch panel, etc. The input reception unit 320 of the operator terminal 300 is realized by the input device 301 and the processor 302. The processor 302 is an arithmetic unit such as a CPU or a GPU, and executes processing according to a program recorded in the memory 304 or the storage 303. The program is, for example, an application program that can be executed on an OS program. In the operator terminal 300, processing is performed by the processor 302 that operates according to a program read onto the memory 304 or the storage 303. The processing unit 310 and the request processing unit 311 realize their respective functions by the processor 302 executing the program.

[0061] The memory 304 is a storage device such as a RAM (Random Access Memory) or a flash memory, and functions as a storage area from which programs and data are temporarily read. The storage 303 is a writable and readable storage device.

[0062] The display 305 is a display device such as a liquid crystal display, an organic EL display, etc. The display unit 330 is realized by the display 305 and the processor 302.

[0063] The communication device 306 is an interface for connecting the operator terminal 300 to an external device for communication. For example, the communication device 306 is a network card that performs wired communication via a wired connection. Alternatively, the communication device 306 performs wireless communication using an antenna that can use a predetermined radio wave (e.g., 5 GHz band, 2.4 GHz band, etc.) that establishes a connection with the task execution support device 100 according to the Wi-Fi standard. The communication unit 340 of the operator terminal 300 is realized by the communication device 306 and the processor 302.

[0064] The processing of each component of the operator terminal 300 may be executed by one piece of hardware or by multiple pieces of hardware. Furthermore, the processing of each component of the operator terminal 300 may be realized by one program or by multiple programs.

[0065] The input device 301, processor 302, storage 303, memory 304, display 305, and communication device 306 are connected to one another by connecting wires such as a bus 307. An example of the hardware configuration of the operator terminal 300 has been described above.

[0066] Each configuration of the operator terminal 300 can be further classified into more components depending on the processing content, and each component can be classified so that it performs more processes.

[0067] Furthermore, each processing unit (processing unit 310, request processing unit 311, input receiving unit 320, display unit 330, communication unit 340) may be constructed using dedicated hardware (ASIC, GPU, etc.) that realizes the respective functions. Furthermore, the processing of each processing unit may be executed by one piece of hardware or by multiple pieces of hardware.

[0068] The generation AI service 200 is a service provided outside the business execution support system 1 that uses a generation AI. However, it is not limited to this, and may use a generation AI held by the business execution support system 1. In this case, the generation AI includes a generation AI that uses a large-scale language model (LLM).

[0069] The generative AI is a trained model that has been previously trained using language data to perform machine learning or deep learning on a large-scale language model based on a neural network (NN). The AI ​​processing unit 210 is realized by the generative AI. The generative AI is not limited to a neural network (NN), and other known techniques may be used. It is also desirable to use a generative AI that has been tuned to achieve higher accuracy by few-shot learning, fine tuning, or the like.

[0070] The communication unit 220 communicates with the task execution support device 100, which is another device, via the network 50.

[0071] Next, the operation of the task execution support system 1 in this embodiment will be described.

[0072] First, in this embodiment, a user or administrator of the task execution support system 1 acts as an operator, and inputs task instructions to the task execution support device 100 as text in a natural language (or text obtained by performing speech recognition on an utterance) using the operator terminal 300. This becomes request information, and the operator terminal 300 requests (asynchronously) a response from the task execution support device 100. The task execution support device 100 then creates a plan to execute the necessary task processes (APIs for data manipulation) in an appropriate order, executes the plan, generates a response including the execution results, or the generation AI service 200 supplies the response to the task execution support device 100, and the task execution support device 100 responds with the response to the operator terminal 300, thereby completing the task instructions.

[0073] For example, when an operator uses the operator terminal 300 to input a command such as "Please discount the estimated cost of the X system construction project for Company A" as text in a natural language, the business execution support device 100 registers the text and tokens in an asynchronous queue. Then, after retrieving the text and tokens from the asynchronous queue, the business execution support device 100 instructs the generation AI service 200 to analyze the text and create a general plan. The generation AI service 200 references past implementation results to extract data operation APIs that are likely to be needed and creates a general plan including their execution order. The business execution support device 100 stores the general plan and its creation status in a predetermined storage area.

[0074] The process up to this point is referred to as the request reception process, which will be described later. The business execution support device 100 then creates integrated data for Company A's X system construction project according to the specific user's data handling range, and instructs the generation AI service 200 to revise the overview plan and create a specific detailed plan in order to manipulate the integrated data. The generation AI service 200 specifically queries the MCP server to refer to OpenAPI information, refines the data manipulation APIs included in the overview plan, and creates a detailed plan. For example, the generation AI service 200 extracts the current estimated amount for Company A's X system construction project, obtains an API from the MCP server to generate new estimate data with a 20% discount, and creates a detailed plan. The generation AI service 200 then obtains an API from the MCP server to update the sales amount projected in the sales management system with a 20% discount, as a task that requires consistency, and incorporates this into the detailed plan. Furthermore, the task execution support device 100 stores the creation status of the detailed plan in a predetermined storage area as needed, and when an inquiry about the execution status is received from the operator terminal 300, the task execution support device 100 responds with the execution history and the execution status.

[0075] The process up to this point is referred to as the detailed plan processing, which will be described later. The business execution support device 100 then instructs the generation AI service 200 to specify parameters to be given to the data operation API according to the detailed plan, acquires them, executes the data operation API including the specified parameters in the execution order of the detailed plan (requests execution from the API server), and saves the execution status, repeating this process until the detailed plan is completed. The business execution support device 100 then instructs the generation AI service 200 to perform a consistency check. The generation AI service 200 specifically compares the results of the data operation with the data before the operation to check for consistency. If there is no consistency, the business execution support device 100 restores the data to the state before the execution of the detailed plan. The business execution support device 100 also stores the execution status of the detailed plan in a specified memory area as needed, and responds with the execution history and execution status when an inquiry about the execution status is received from the operator terminal 300.

[0076] The process up to this point is referred to as the execution control process, which will be described later. The business execution support device 100 then instructs the generation AI service 200 to create a report on the execution results. The generation AI service 200 creates report data including the execution results of the implemented API. Based on the report data obtained from the generation AI service 200, the business execution support device 100 generates a predetermined business instruction execution result screen and displays it on the operator terminal 300.

[0077] The process up to this point is the result recording process described later.

[0078] 6 is a diagram showing an example of the flow of the request reception process. The flow of the request reception process starts when the operator terminal 300 receives a start process from the operator.

[0079] First, the request processing unit 311 of the operator terminal 300 accepts input of text that is a task instruction and transmits the text to the task execution support device 100 (step S101). Specifically, after performing user login and authentication, the request processing unit 311 of the operator terminal 300 accepts the text and transmits it to the task execution support device 100 together with information such as a user identifier.

[0080] Then, the request receiving unit 111 of the task execution support device 100 creates a token including the user and the handleable data range (step S102). Specifically, when the request receiving unit 111 receives text input from the user, it refers to the access right information 121 for the user who is in a usage state (logged in) to obtain information including the handleable data range, creates a predetermined token (for example, a 36-digit shortened MCP token, etc.), and stores it in the token.

[0081] Then, the request receiving unit 111 registers the text and the token in a FIFO asynchronous queue (step S103).

[0082] The general planning unit 112 obtains the text and tokens received by the request receiving unit 111 from the FIFO asynchronous queue. Note that because the queue is asynchronous, the general planning unit 112 does not necessarily obtain the text and tokens received by the request receiving unit 111 from the asynchronous queue immediately after execution of step S103. In other words, because it is asynchronous, the work request from the operator terminal 300 is temporarily terminated, allowing the operator to execute another task in parallel.

[0083] Then, the general planning unit 112 creates an instruction analysis prompt and instructs the generation AI service 200 to analyze the text and create a general plan (step S104). Specifically, the general planning unit 112 creates an instruction analysis prompt to analyze the text and create a general plan, and instructs the generation AI service 200 to do so. The general planning unit 112 obtains API information of the business program (data manipulation program) via an MCP server that stores information about the API of the business program, and uses this information to generate the general plan. However, this is not limited to this, and the generation AI service 200 may obtain the API specifications by referring to the data manipulation specification information 123 via the MCP server, and generate the general plan.

[0084] The generation AI service 200 creates an overview plan in accordance with the instruction analysis prompt (step S105). Specifically, the generation AI service 200 analyzes the content of the text and identifies the business content desired by the user. Then, using information about the API of the business program and information about past implementation results, the generation AI service 200 extracts the data operation APIs to be used to realize the business content, organizes the processing procedures including the execution order, and creates an overview plan.

[0085] Then, the general planner 112 stores the general plan and its creation status in a predetermined storage area so that the general plan can be referenced (step S106).

[0086] The above is an example of the flow of the request reception process. According to the request reception process, it is possible to automatically generate a general plan for carrying out business processing in response to a business instruction in natural language from a user.

[0087] 7 is a diagram showing an example of the flow of detailed planning processing. The flow of detailed planning processing starts immediately after the request acceptance processing ends.

[0088] First, the detailed planner 114 acquires the user and the handleable data range from the token and creates integrated data in JSON format (step S201). Specifically, the detailed planner 114 instructs the integrated data generator 113 to acquire a group of data related to the business as integrated data. The integrated data generator 113 extracts data included in the range defined by the handleable data range 121b for a user who is in a usage state (logged in) from the business information 122, denormalizes the group of data related to the business to be executed, and converts it into a text-description type JSON integrated data format to create integrated data.

[0089] Then, the detailed plan unit 114 instructs the user to review the general plan using a detailed plan prompt that includes the integrated data (step S202).

[0090] The generation AI service 200 then reviews the overview plan in accordance with the detailed plan prompt and creates a detailed plan (step S203). Specifically, the generation AI service 200 queries the MCP server to reference the OpenAPI information and integrated data, refines the data manipulation APIs included in the overview plan, and creates a detailed plan. For example, the generation AI service 200 extracts the current estimated amount for the X system construction project for Company A, obtains an API from the MCP server to generate new estimate data with a 20% discount, and creates a detailed plan. The generation AI service 200 then obtains an API from the MCP server to update the sales amount projected in the sales management system with the 20% discount amount, as a task that requires consistency, and includes this in the detailed plan.

[0091] Then, the detailed plan unit 114 stores the detailed plan and its creation state in a referenceable manner in a predetermined storage area (step S204).

[0092] The above is an example of the flow of detailed planning processing. According to the detailed planning processing, a general plan for carrying out business processing can be automatically concretized to generate a detailed plan while specific data can be referenced.

[0093] 8 is a diagram showing an example of the flow of the execution control process. The flow of the execution control process starts immediately after the detailed plan process ends.

[0094] First, the data operation unit 115 saves the integrated data as pre-processing backup data (step S301).

[0095] Then, the data manipulation unit 115 creates an execution prompt and instructs the generation of request data (step S302). Specifically, the data manipulation unit 115 instructs the generation AI service 200 to generate request data including parameters and option specifications for executing the execution targets of the business program according to the execution order included in the detailed plan. At that time, the data manipulation unit 115 creates an execution prompt that instructs the generation of command information (information required to call the API) by referring to the execution results of the API of the past data manipulation, and instructs the generation AI service 200.

[0096] Then, the generation AI service 200 refers to the execution results of past data operation APIs in accordance with the execution prompt and generates command information (information required to call the data operation API) (step S303). Specifically, the generation AI service 200 queries the MCP server to refer to the OpenAPI information and integrated data, and refers to the API execution history to detail the parameters and options of the data operation API and create request data.

[0097] Then, the data operation unit 115 invokes the API for the data operation using the request data (step S304). Specifically, the data operation unit 115 requests the API server (not shown) to execute the data operation (business program, business process) via the MCP server. Then, the data operation unit 115 saves the executed API, parameters, options, execution status, and execution results as the execution state so that they can be referenced (step S305). Note that steps S302 to S305 are performed for each business program included in the detailed plan, according to its step.

[0098] Then, the consistency processing unit 116 generates a consistency prompt and instructs a data consistency check (step S306). Specifically, the consistency processing unit 116 creates a consistency prompt that verifies the consistency of the execution results of the business program by the data operation unit 115 based on the backup data acquired in step S301, and instructs the generation AI service 200 to do so.

[0099] Then, the generation AI service 200 determines the consistency of the data in accordance with the consistency prompt (step S307). Specifically, the generation AI service 200 determines whether the data changes made by the execution of the business program are consistent in accordance with the consistency prompt using a predetermined method, and returns to the business execution support device 100 whether the verification result of the data consistency is consistent.

[0100] If the determination result indicates inconsistency (inconsistency), the consistency processing unit 116 restores the data using the backup data (step S308). Specifically, the consistency processing unit 116 discards the data changes made by the execution of the business program by the data operation unit 115, and returns the data to the state before the execution, i.e., the state of the backup data saved in step S301.

[0101] The above is an example of the flow of the execution control process. The execution control process allows data manipulation to be performed according to the detailed plan by specifying specific data manipulation parameters and options, and then automates the control of business processing to perform data manipulation in accordance with the detailed plan.

[0102] 9 is a diagram showing an example of the flow of the result recording process. The flow of the result recording process starts immediately after the execution control process ends.

[0103] First, if the determination result in step S307 of the execution control process indicates that the consistency is OK (consistency), the consistency processing unit 116 records the data changes made by the execution of the business program (step S401).

[0104] Then, the status reporting unit 117 creates a report creation prompt and instructs the generation of a report (step S402). Specifically, the status reporting unit 117 analyzes the execution result, creates a report creation prompt that creates a report in a predetermined format in plain language, and instructs the generation AI service 200 to create the report.

[0105] Then, the generation AI service 200 generates report data in accordance with the report generation prompt (step S403). Specifically, the generation AI service 200 analyzes the execution status of the business program in accordance with the report generation prompt, and generates report data in natural language that includes at least the execution content and execution results of the business.

[0106] Then, the status reporting unit 117 stores the report data (step S404).

[0107] Then, the status reporting unit 117 generates a predetermined work instruction execution result screen based on the report data obtained from the generation AI service 200 (step S405), and displays it on the operator terminal 300 (step S406).

[0108] The above is an example of the flow of the result recording process. The result recording process makes it possible to display the content of data manipulations performed according to the detailed plan, as well as the execution status and execution results.

[0109] Furthermore, when an inquiry about the execution status is received from the operator terminal 300 at any time, the status reporting unit 117 responds with the execution history and execution status. This makes it possible to eliminate the drawback of an execution model that uses an asynchronous queue, namely, the difficulty in grasping the execution results in real time.

[0110] 10 is a diagram showing an example of an instruction analysis prompt. The example instruction analysis prompt 600 includes a command 601 that instructs the generation AI service 200 to use the MCP to create an execution plan (overview plan) for realizing a business instruction (denoted as "XX" in the example instruction analysis prompt 600). The example instruction analysis prompt 600 also includes an example output format 602.

[0111] 11 is a diagram showing an example of a detailed planning prompt. The example detailed planning prompt 700 includes an instruction 701 that instructs the generation AI service 200 to review the overview plan based on the acquired integrated data and create a detailed plan. The example detailed planning prompt 700 also includes an example output format 702.

[0112] 12 is a diagram showing an example of an execution prompt. The example execution prompt 800 includes an instruction 801 that instructs the generation AI service 200 to create specific request data (API parameters and options) for the next step to be executed in the detailed plan based on the execution results. The example execution prompt 800 also includes an example output format 802.

[0113] 13 is a diagram showing an example of a consistency prompt. The example consistency prompt 900 includes instructions 901 that instruct the generative AI service 200 to analyze the work results, check for completeness, and detect anomalies. The example consistency prompt 900 also includes an example output format 902.

[0114] 14 is a diagram showing an example of a report creation prompt. The example report creation prompt 1000 includes instructions 1001 that instruct the generation AI service 200 to analyze the execution results and create a report of approximately 100 to 300 characters in plain language. The example report creation prompt 1000 also includes example elements 1002 to be included in the output.

[0115] The above has described the business execution support system 1 to which the embodiment of the present invention is applied. According to this embodiment, an operating environment that allows business objectives to be easily achieved can be provided.

[0116] The present invention is not limited to the above-described embodiment. Various modifications of the above-described embodiment are possible within the scope of the technical concept of the present invention. For example, in the above-described embodiment, the generative AI service 200 includes a generative AI that uses a large-scale language model (LLM). However, the present invention is not limited to this, and the generative AI may be created using other modeling techniques.

[0117] Furthermore, the processing in the above embodiment can support business processing in companies, etc., but is not limited to this. For example, even individuals can create data processing required for asset management, life's work, and solving various problems using APIs, and automatically achieve consistency and obtain final output.

[0118] Furthermore, the technical elements of the above-described embodiments may be applied independently, or may be divided into multiple parts such as program parts and hardware parts and applied.

[0119] The present invention has been described above mainly with reference to the embodiments. [Explanation of symbols]

[0120] 1 Business execution support system, 50 Network, 100 Business execution support device, 110 Processing unit, 111 Request reception unit, 112 Overview planning unit, 113 Integrated data generation unit, 114 Detailed planning unit, 115 Data operation unit, 116 Consistency processing unit, 117 Status reporting unit, 120 Storage unit, 121 Access right information, 122 Business information, 123 Data operation specification information, 130 Communication unit, 200 Generated AI service, 210 AI processing unit, 220 Communication unit, 300 Operator terminal, 310 Processing unit, 311 Request processing unit, 320 Input reception unit, 330 Display unit, 340 Communication unit.

Claims

1. a request receiving unit that receives text in a natural language including details of the business to be executed; an overview planning unit that analyzes the text and instructs a predetermined generation AI service to generate an overview plan including execution targets and execution sequences of a predetermined business program; a detailed planning unit that acquires a group of data related to the business as integrated data, and instructs the generation AI service to concretize and update the outline plan according to the content of the integrated data and generate a detailed plan; a data manipulation unit that instructs the generation AI service to generate command information for executing the execution target of the business program in accordance with the execution order included in the detailed plan, and executes the business program using the command information; Equipped with the general planning unit obtains information on an API (Application Programming Interface) of the business program via an MCP (Model Context Protocol) server that stores information on the API of the business program, and uses the information in generating the general plan; A business execution support system characterized by:

2. 2. The business execution support system according to claim 1, a status reporting unit that records the execution status of the business program by the data manipulation unit and presents it for reference; A business execution support system comprising:

3. 2. The business execution support system according to claim 1, a consistency processing unit that instructs the generated AI service to verify the consistency of the execution result of the business program by the data manipulation unit; A business execution support system comprising:

4. 4. The business execution support system according to claim 3, when the verification result of the consistency indicates inconsistency, the consistency processing unit discards the data changes made by the execution of the business program by the data operation unit, and returns the data to the state before the execution; A business execution support system characterized by:

5. 2. The business execution support system according to claim 1, the detailed planning unit denormalizes the data group related to the business and converts it into a predetermined integrated data format of text description type to create the integrated data; A business execution support system characterized by:

6. 2. The business execution support system according to claim 1, The detailed planning unit obtains a data set related to the business using a Model Context Protocol (MCP) token including information specifying a predetermined access right of the user who provided the text. A business execution support system characterized by:

7. 2. The business execution support system according to claim 1, When the request receiving unit receives a natural language text including the execution content of the business, it registers the text in a FIFO asynchronous queue; the high level planner receives the text from the asynchronous queue; A business execution support system characterized by:

8. 2. The business execution support system according to claim 1, a status reporting unit that analyzes the execution status of the business program by the data manipulation unit and presents a report in natural language that includes at least the execution content and execution results of the business in a referable manner; A business execution support system comprising:

9. A business execution support method using a computer system, comprising: The computer system includes: a request receiving step of receiving text in a natural language including details of the business to be executed; an outline planning step of instructing a predetermined generation AI service to analyze the text and generate an outline plan including execution targets and execution sequences of a predetermined business program; a detailed planning step of acquiring a group of data related to the business as integrated data, and instructing the generation AI service to concretize and update the outline plan according to the content of the integrated data and generate a detailed plan; a data manipulation step of instructing the generation AI service to generate command information for executing the execution target of the business program in accordance with the execution order included in the detailed plan, and executing the business program using the command information; and In the general planning step, information on an API (Application Programming Interface) of the business program is obtained via an MCP (Model Context Protocol) server that holds information on the API of the business program, and the information is used in generating the general plan. A business execution support method comprising:

10. A business execution support program that causes a computer system to support business execution, The computer system includes: a request receiving step of receiving text in a natural language including details of the business to be executed; an outline planning step of instructing a predetermined generation AI service to analyze the text and generate an outline plan including execution targets and execution sequences of a predetermined business program; a detailed planning step of acquiring a group of data related to the business as integrated data, and instructing the generation AI service to concretize and update the outline plan according to the content of the integrated data and generate a detailed plan; a data manipulation step of instructing the generation AI service to generate command information for executing the execution target of the business program in accordance with the execution order included in the detailed plan, and executing the business program using the command information; Execute In the general planning step, information on an API (Application Programming Interface) of the business program is obtained via an MCP (Model Context Protocol) server that holds information on the API of the business program, and the information is used in generating the general plan. A business execution support program characterized by:

Images