Information processing device, information processing method, and information processing system

Abstract

It provides information processing that autonomously responds to various instructions from the user. [Solution] An information processing device 1 that performs information processing using a supervisor agent 11 and multiple subagents 12, each of which can execute different tasks. When a first instruction text indicating the content of the business to be processed is input, the supervisor agent 11 identifies a subagent 12 corresponding to the task to be processed that constitutes the business to be processed, and outputs a second instruction text indicating the content of the task to be processed to the identified subagent 12. In response to obtaining the second instruction text, the subagent 12 executes the task indicated by the second instruction text.

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing method, and an information processing system. [Background technology]

[0002] BACKGROUND ART There is known a device that performs business processing using generative AI (Artificial Intelligence) (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2025-026297 Summary of the Invention [Problem to be solved by the invention]

[0004] However, conventional generative AI had problems in that the tasks it could handle were limited and users had to provide instructions that matched the AI's specifications.

[0005] The present invention has been made in consideration of these points, and has as its object to provide information processing that autonomously responds to various instructions from a user. [Means for solving the problem]

[0006] In a first aspect of the information processing device of the present invention, information processing is performed by a first agent and a plurality of second agents, each of which can execute different tasks. When a first instruction text indicating the content of a business to be processed is input, the first agent identifies the second agent corresponding to the task to be processed that constitutes the business to be processed, and outputs second instruction text indicating the content of the task to be processed to the identified second agent. In response to obtaining the second instruction text, the second agent executes the task indicated by the second instruction text.

[0007] When the first instruction text is input, the first agent may refer to business specification data that associates a reference task with one or more processing methods for the reference task, and output the second instruction text that indicates a processing method for the reference task that corresponds to the task to be processed.

[0008] In the business specification data, the reference task is associated with one or more second agents that can execute the reference task, and the first agent may identify the second agent that can execute the reference task corresponding to the task to be processed by referring to the business specification data, and output the second instruction text to the identified second agent. The first agent may output the second instruction text to another second agent that is capable of executing the reference task corresponding to the task to be processed when the result of the second agent, to which the second instruction text was output, executing the task to be processed does not satisfy a predetermined condition.

[0009] The second agent may execute the task to be processed based on a first processing method for the task to be processed indicated by the second instruction text, and if the task to be processed is not completed based on the first processing method, may execute the task to be processed based on a second processing method different from the first processing method.

[0010] The second agent may output the execution result of the task to be processed to the first agent, and the first agent may determine whether the execution result is valid based on the first instruction text and the execution result, and if the execution result is not valid, may output a third instruction text to the second agent indicating a processing method different from the second instruction text.

[0011] The first agent may determine whether the number of times the task to be processed, for which the execution result has been input from the second agent in response to input of one of the first instruction texts, has been executed is a predetermined number of times or more, and if the number of times the task to be processed has been executed is the predetermined number of times or more, output a message indicating that the business to be processed cannot be executed.

[0012] The business to be processed is composed of a plurality of the tasks to be processed that are processed sequentially, the first agent identifies a plurality of second agents corresponding to each of the tasks to be processed, and outputs second instruction text indicating the content of each of the tasks to be processed to each of the identified second agents, and each of the second agents outputs to the first agent the execution result of the task to be processed that corresponds to the input second instruction text, and when the first agent receives the execution result of the task to be processed executed by the second agent from the second agent, the first agent may determine the content of the task to be processed that is subsequent to the task to be processed based on the execution result and output the second instruction text indicating the determined content.

[0013] The business to be processed may be composed of a plurality of the tasks to be processed, and when the complexity of the business to be processed is equal to or greater than a predetermined level, the first agent may cause a task plan agent for determining a processing order of the plurality of tasks to be processed to determine the processing order of the plurality of tasks to be processed, and based on the processing order, output a plurality of the second instruction texts to a plurality of second agents corresponding to each of the plurality of tasks to be processed, for causing each of the plurality of tasks to be processed.

[0014] In a second aspect of the information processing device of the present invention, information processing is performed by a first agent and a file loader agent capable of manipulating files, and when first instruction text indicating the content of the business to be processed is input, the first agent refers to business specification data that associates a reference task with one or more processing methods for the reference task, and if the reference task corresponding to the business to be processed includes file manipulation, outputs second instruction text to the file loader agent that includes the reference task corresponding to the business to be processed, and the file loader agent executes the task indicated by the second instruction text in response to acquiring the second instruction text.

[0015] In a third aspect of the information processing method of the present invention, a computer executes information processing using a first agent and a plurality of second agents, each of which can execute different tasks, and the information processing method includes the steps of: when a first instruction text indicating the content of a business to be processed is input, the first agent identifies the second agent corresponding to the task to be processed that constitutes the business to be processed, and outputting second instruction text indicating the content of the task to be processed to the identified second agent; and the second agent, in response to acquiring the second instruction text, executes the task indicated by the second instruction text.

[0016] In a fourth aspect of the information processing system of the present invention, information processing is performed by a first agent and a plurality of second agents, each of which can execute different tasks. When a first instruction text indicating the content of a business to be processed is input, the first agent identifies the second agent corresponding to the task to be processed that constitutes the business to be processed, and outputs second instruction text indicating the content of the task to be processed to the identified second agent. In response to obtaining the second instruction text, the second agent executes the task indicated by the second instruction text. [Effects of the Invention]

[0017] According to the present invention, it is possible to provide information processing that autonomously responds to various instructions from a user. [Brief explanation of the drawings]

[0018] [Figure 1] 1 is a diagram for explaining an overview of an information processing system S according to an embodiment. [Figure 2] 1 is a block diagram showing a configuration of an information processing device 1. FIG. [Figure 3] 3 is a diagram showing an example of the data structure of business specification data stored in a storage unit 14. FIG. [Figure 4] FIG. 2 is a block diagram showing the configuration of a supervisor agent 11. [Figure 5] FIG. 2 is a block diagram showing the configuration of a subagent 12. [Figure 6] FIG. 10 is a block diagram showing another example of the configuration of the supervisor agent 11. [Figure 7] 3 is a flowchart showing the flow of processing in the information processing device 1. [Figure 8] FIG. 10 is a diagram showing the configuration of an information processing system SA according to a modified example. DETAILED DESCRIPTION OF THE INVENTION

[0019] [Overview of information processing device 1] FIG. 1 is a diagram for explaining an overview of an information processing system S according to an embodiment. The information processing system S is a system for executing information processing related to business content by coordinating multiple agents. Businesses that the information processing system S can process include, for example, invoice processing, order processing, payment management (reconciliation processing), account item determination processing, and data format conversion processing for forms, etc. The information processing system S acquired from another information system has an information processing device 1, an external device 2, and an information terminal 3.

[0020] The information processing device 1 is a device for receiving instructions including work content and processing the instructed work. The information processing device 1 executes information processing using a supervisor agent 11 and multiple subagents 12. The supervisor agent 11 and the subagents 12 are programs for executing predetermined processing using a large-scale language model (LLM). Hereinafter, the supervisor agent 11 and the subagents 12 may be collectively referred to simply as "agents."

[0021] As will be described in detail later, based on text indicating the content of the business to be processed (hereinafter referred to as "first instruction text"), the supervisor agent 11 breaks down the business indicated by the first instruction text into tasks, and outputs text instructing the subagent 12 on the content of the tasks to be processed (hereinafter referred to as "second instruction text"). Specifically, the first instruction text includes an outline of the information processing required to achieve a predetermined purpose. The content of the business indicated by the first instruction text is typically realized by executing multiple tasks. Tasks include content that subdivides and embodies the content of the business indicated by the first instruction text. Specific examples of business and tasks will be described later.

[0022] The supervisor agent has a task generation model. The task generation model generates a second instruction text based on the input first instruction text, identifies the subagent 12 that will perform the task indicated by the second instruction text, and outputs the second instruction text and agent identification information. The agent identification information is information that indicates the subagent 12 that will perform the task indicated by the second instruction text.

[0023] In response to input of the second instruction text, the subagent 12 executes the task indicated by the second instruction text. The subagent 12 has an execution code generation model that outputs code for executing the task when the second instruction text is input. The code may be, for example, a program for executing the task using a predetermined processing method, or may be data defining a workflow indicating how to process the task. The subagent 12 is connected to an external device 2 so as to be able to communicate with the external device 2.

[0024] The subagents 12 each have a different executable task, and are classified into, for example, a communication agent 12A, a file loader agent 12B, a task plan agent 12C, a document analysis agent 12D, and an accounting specialist agent 12E.

[0025] The communication agent 12A is an agent for communicating with users (hereinafter simply referred to as "users") who use the information processing services provided by the information processing device 1, and with users of organizations outside the organization to which the user belongs, via email, chat, etc.

[0026] The file loader agent 12B is an agent for performing various file operations (upload, download, etc.), referencing or acquiring information provided by various information systems or information processing services, or inputting information into various information systems.

[0027] The task plan agent 12C adjusts the order in which tasks are processed and the timing at which tasks are processed when the relationships between tasks that make up the work indicated by the first instruction text are complex. For example, the task plan agent adjusts the order or timing at which tasks are processed when the completion of multiple tasks is a condition for starting a specific task or when the time at which a task can be executed is limited (for example, when a start time or completion deadline for a specific task is set).

[0028] The document analysis agent 12D is an agent for analyzing text contained in an input file when the file is input. Examples of analysis processing provided by the document analysis agent include summarizing text, checking consistency between texts, and checking consistency with various laws and regulations (e.g., checking legal issues in contracts and consistency with company rules). The files analyzed by the document analysis agent 12D are, for example, document files, spreadsheet files, presentation files, or image files that can be processed using OCR (Optical Character Recognition).

[0029] The accounting specialist agent 12E is an agent for responding to inquiries about accounting processes from users. For example, the accounting specialist agent 12E responds to inquiries such as inquiries about account items. For example, the accounting specialist agent 12E may output materials necessary to determine the inquiry content in response to second instruction text including the inquiry content from the supervisor agent 11. Furthermore, the accounting specialist agent 12E may output a response to the inquiry content using as input second instruction text including materials necessary to determine the inquiry content included in the second instruction text.

[0030] The external device 2 is a device that provides information necessary for each subagent to execute a task. Specifically, the external device 2 is a device that provides information systems, data, or various information processing services (e.g., OCR) used by the organization to which the user belongs.

[0031] The information terminal 3 is a terminal used by a user who utilizes information processing provided by the information processing device 1. The user operates the information terminal 3 to input a first instruction text.

[0032] An overview of the processing of the information processing device 1 will be explained using a specific use case as an example. This use case assumes processing for processing an invoice. The information processing device 1 obtains a first instruction text ((1) in Figure 1). In this use case, the first instruction text states, "Enter the invoice information into invoice management system B based on the PDF file of the invoice received by email from invoice management system A."

[0033] When the first instruction text is input, the supervisor agent 11 identifies the subagent 12 that corresponds to the task to be processed ((2) in FIG. 1). As will be described in detail later, the information processing device 1 refers to business specification data that associates a task, a task processing method, and a subagent that executes the task, inputs the content indicated by the business specification data and the first instruction text into a task generation model, and identifies the subagent 12 that executes the task to be processed.

[0034] In this use case, the supervisor agent 11 identifies the file loader agent 12B as the sub-agent 12 that handles the tasks of "receiving invoices from invoice management system A" and "entering invoices into invoice management system B."

[0035] The supervisor agent 11 generates a second instruction text ((3) in FIG. 1). In this use case, the second instruction text is written as follows, for example: "Perform the following process (1) Click the link in the email sent from A's invoice management system, enter the password provided in another email, and download the invoice PDF. (2) Next, enter the invoice information into Invoice Management System B based on the information contained in the downloaded invoice PDF file.

[0036] The supervisor agent 11 outputs the generated second instruction text to the identified subagent 12 (file loader agent 12B) ((4) in FIG. 1).

[0037] The subagent 12 executes the task based on the second instruction text ((5) in FIG. 1). In this use case, the file loader agent 12B inputs the second instruction text into the execution code generation model and causes the execution code generation model to output code for executing the task to be processed. Specifically, based on the second instruction text, the file loader agent 12B identifies that the following steps are required to execute the task, and outputs code for executing the following steps.

[0038] (a) Obtaining email identification information (ID, subject, date and time of sending, etc.) that identifies the email and a link to obtain a PDF file contained in the email sent from the A invoice management system. (b) Based on the email identification information obtained in (a), identify an email containing a password corresponding to the email identification information, and obtain the password contained in the identified email. (c) Downloading the PDF file using the link and password obtained. (d) Input the obtained PDF file into an OCR processing service and extract the information contained in the invoice. (O) B. Log in to the invoice management system. (a) Enter the information extracted in (b) into B's invoice management system.

[0039] The file loader agent 12B executes the code output by the execution code generation model. The subagent 12 outputs the result of the task execution to the supervisor agent 11 ((6) in FIG. 1). In this use case, the file loader agent 12B may output to the supervisor agent 11 that the task has been completed, or may output to the supervisor agent 11 the information entered into the invoice management system B. As will be described later, the information processing device 1 has multiple processors, and the subagent 12 may cause different processors to execute the process of generating code for executing the task and the process of executing the generated code.

[0040] The information processing device 1 is configured to perform information processing corresponding to the received instructions using a supervisor agent 11 that breaks down the work indicated by the received instructions into tasks, and a subagent 12 that executes tasks based on instructions from the supervisor agent 11, thereby providing information processing that autonomously responds to various instructions from the user.

[0041] [Configuration of information processing device 1] 2 is a block diagram showing the configuration of the information processing device 1. The information processing device 1 has a communication unit 13, a storage unit 14, and a control unit 15. The control unit 15 has a supervisor agent 11 and a subagent 12.

[0042] The communication unit 13 is a communication interface for transmitting and receiving data to and from other devices via a network. The storage unit 14 is a storage medium including a ROM (Read Only Memory), a RAM (Random Access Memory), an SSD (Solid State Drive), a hard disk drive, etc. The storage unit 14 stores in advance programs to be executed by the control unit 15.

[0043] The storage unit 14 stores a task generation model used by the supervisor agent 11 for inference and an execution code generation model used by each of the plurality of subagents 12 for inference.

[0044] The memory unit 14 stores business specification data. FIG. 3 is a diagram showing an example of the data structure of business specification data. In the business specification data, a "reference task," a "processing method," a "related system," and a "subagent" are associated with each other. A "reference task" indicates the type of task. A "processing method" indicates the method of processing the reference task. A "related system" indicates a system that needs to be accessed to process the task. A "related system" includes the name or ID of the system or an address for accessing the system (e.g., a URL (Uniform Resource Locator)). A "related system" may further include an ID, password, etc. for accessing the system. A "subagent" indicates a subagent that can execute the reference task.

[0045] The control unit 15 is a processor such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), etc. The control unit 15 functions as the supervisor agent 11 and the subagent 12 by executing the programs stored in the storage unit 14.

[0046] 2 shows an example in which one control unit 15 operates the agents, but this is not limiting. For example, the information processing device 1 may have an independent processor for each agent. Furthermore, the information processing device 1 may have multiple processors for executing each functional unit of the agent.

[0047] 4 is a block diagram showing the configuration of the supervisor agent 11. The supervisor agent 11 includes an acquisition unit 111, a task identification unit 112, and an output unit 113.

[0048] The acquisition unit 111 acquires first instruction text. As an example, the acquisition unit 111 may cause the information terminal 3 to display a chat interface and acquire the first instruction text input into the displayed chat interface from the information terminal 3. The acquisition unit 111 may also acquire the first instruction text stored in the storage unit 14, or may acquire the first instruction text from an external device (not shown).

[0049] When the first instruction text is input, the task identification unit 112 refers to the business specification data and outputs second instruction text indicating a processing method for a reference task corresponding to the processing target task. As an example, the task identification unit 112 refers to the storage unit 14 and acquires the business specification data. The task identification unit 112 may determine the similarity between each item included in the business specification data and the content included in the first instruction text, and acquire items of the business specification data that have a high degree of similarity. The task identification unit 112 inputs the first instruction text and the acquired business specification data to the task generation model, and causes it to output second instruction text. The output unit 113 outputs the second instruction text generated by the task generation model to the subagent 12.

[0050] Furthermore, the task identification unit 112 refers to the business specification data to identify a subagent that can execute a reference task corresponding to the target task. Specifically, the task identification unit 112 inputs the first instruction text and the acquired business specification data into the task generation model, and causes the task generation model to output agent identification information.

[0051] The task identification unit 112 outputs the second instruction text to the identified subagent. Specifically, the task identification unit 112 outputs the second instruction text output by the task generation model to the subagent indicated by the agent identification information output by the task generation model.

[0052] 5 is a block diagram showing the configuration of the subagent 12. The subagent 12 includes an acquisition unit 121, a task analysis unit 122, a task execution unit 123, and an output unit .

[0053] The acquisition unit 121 acquires the second instruction text. The task analysis unit 122 inputs the second instruction text into the execution code generation model, causing the execution code generation model to output code for executing the task.

[0054] The task execution unit 123 executes the task based on the code output by the task analysis unit 122. Specifically, the task execution unit 123 executes the code output by the task analysis unit 122. The output unit 124 outputs the execution result of the task executed by the task execution unit 123 to the supervisor agent 11.

[0055] For example, it is conceivable that a subagent 12 may not be able to execute a task as expected due to a problem with the system being accessed or a problem with the file being referenced, etc. Therefore, if the result of execution of the task to be processed by the subagent 12 that output the second instruction text does not satisfy a predetermined condition, the supervisor agent 11 may be configured to output the second instruction text to another subagent 12 that can execute a reference task corresponding to the task to be processed.

[0056] Fig. 6 is a block diagram showing another example of the configuration of the supervisor agent 11. The supervisor agent 11 shown in Fig. 6 has a judgment unit 114. As an example, the storage unit 14 stores a judgment model that receives as input a second instruction text and an execution result of a task based on the second instruction text and judges whether the execution result of the task is appropriate. The judgment model is a model that is machine-learned using examples of appropriate task execution results as training data.

[0057] The determination unit 114 acquires the task execution result from the subagent 12. In addition to the task execution result, the determination unit 114 may further acquire second instruction text input to the subagent 12 to output the execution result or information for identifying the second instruction text. The determination unit 114 inputs the task execution result and the second instruction text input to the subagent 12 to output the execution result into a determination model, and determines whether the task execution result is valid.

[0058] If the execution result of the task is determined to be valid, the determination unit 114 determines that the task has been completed, and if there is a subsequent task, the output unit 113 outputs a second instruction text to the subagent 12 for executing the subsequent task.

[0059] If the execution result of the task is invalid, the task identification unit 112 identifies another subagent 12 that can execute another task corresponding to the task corresponding to the execution result, and generates a new second instruction text to be output to the other subagent 12.

[0060] The processing of the information processing device 1 in this case will be described based on a specific use case. In this use case, it is assumed that order information is entered into a purchasing system, but the approval required to process the order information is in an unapproved status in the approval system.

[0061] In this case, the first instruction text states, "When an order is submitted in the purchasing system, if the order has been approved in the approval system, execute the order processing." The supervisor agent 11 inputs the acquired first instruction text and the business specification data into the task generation model, specifies the file loader agent 12B as the agent that will execute the task, and outputs the second instruction text to the task generation model.

[0062] In this case, the second instruction text includes the following content: (a) Log in to the purchasing system. (i) Obtaining order information (order details, approval number, orderer information, etc.) entered into the purchasing system. (c) Log in to the approval system. (d) Check whether the approval corresponding to the order information has been approved in the approval system. (E) To process orders based on order information when approval has been granted in the approval system.

[0063] The acquisition unit 121 acquires the second instruction text. The task analysis unit 122 inputs the second instruction text into the execution code generation model and causes the execution code generation model to output code for executing the task. The task execution unit 123 executes the task based on the code output by the execution code generation model. As a result, in the approval system, the approval corresponding to the order information has not been approved, so the output unit 124 outputs an execution result indicating that "the approval corresponding to the order information has not been approved" to the supervisor agent 11.

[0064] The task identification unit 112 of the supervisor agent 11 inputs the first instruction text, the execution result obtained from the output unit 124 of the file loader agent 12B, and the business specification data into the task generation model. The task identification unit 112 identifies the communication agent 12A as the agent that will execute the task, and generates new second instruction text indicating, "Request by email the orderer indicated by the order information to proceed with the approval process." The output unit 113 outputs the new second instruction text that has been generated to the communication agent 12A.

[0065] By configuring the information processing device 1 to judge the results of task execution and have another agent execute the task if the results are inappropriate, it becomes possible to complete a specified task without human intervention and to respond autonomously to various instructions from the user.

[0066] However, if the same subagent 12 can execute the same task using different processing methods and the execution result of the subagent 12 is invalid, the supervisor agent 11 may be configured to instruct the subagent 12 to execute the task using a processing method different from the processing method that it executed.

[0067] In this case, if the determination unit 114 determines that the execution result of the task output by the subagent 12 is invalid, the task identification unit 112 generates a third instruction text indicating that the task should be executed using a processing method different from that of the second instruction text. The output unit 113 outputs the third instruction text generated by the task identification unit 112 to the subagent 12.

[0068] Specifically, the processing in this case will be explained based on a use case in which the step of extracting invoice information from a PDF file of an invoice can be performed by either OCR processing service A or OCR processing service B. As an example, the task identification unit 112 outputs second instruction text instructing to "extract invoice information using OCR processing service A."

[0069] As already explained, the subagent 12 executes the task based on the input second instruction text. As a result, the subagent 12 outputs an execution result indicating that the task could not be executed to the supervisor agent. For example, the execution result indicates that the OCR processing service A could not be accessed for some reason.

[0070] As already explained, the determination unit 114 determines whether the execution result is valid. If it is determined that the execution result is invalid, the task identification unit 112 generates third instruction text indicating that the task should be executed using a different processing method (for example, performing OCR processing using OCR processing service B). The output unit 113 outputs the third instruction text to the subagent 12.

[0071] In this way, if the result of the task execution is invalid, the supervisor agent 11 is configured to generate a third instruction text to instruct the task to be executed using a processing method different from the processing method that was executed, thereby making it possible to complete the task autonomously.

[0072] In such a use case, a problem may occur in which the work is not completed due to a processing loop between the supervisor agent 11 and the subagent 12. Therefore, in order to avoid such a loop caused by a task not being completed, the information processing device 1 may be configured to output a message and terminate the processing when a task has been executed a predetermined number of times or more.

[0073] Specifically, the supervisor agent 11 determines whether the number of times the target task, for which the execution result has been input from the subagent 12 in response to the input of one first instruction text, has been executed is equal to or greater than a predetermined number. If the number of times the target task has been executed is equal to or greater than the predetermined number, the supervisor agent 11 outputs a message indicating that the target job cannot be executed.

[0074] As an example, the second instruction text and the third instruction text include the number of times the task to be processed has been executed. That is, when generating the third instruction text, the task identification unit 112 increments the number of times the task has been executed, and generates the third instruction text including the incremented number of times the task has been executed. The determination unit 114 determines whether the execution result of the task is valid based on the execution result output by the subagent 12 and the third instruction text.

[0075] If the execution result of the task is invalid, the determination unit 114 determines the number of times the task has been executed. If the number of times the task has been executed included in the third instruction text is equal to or greater than a predetermined number, the determination unit 114 determines to output a predetermined message. The predetermined number is determined so as not to place a burden on the information processing resources of the information processing device 1. The predetermined number is, for example, three times. The predetermined message is a message indicating that the task cannot be completed. The predetermined message may include information indicating which task was not completed. If the number of times the task has been executed is equal to or greater than the predetermined number, the output unit 113 outputs the predetermined message to the information terminal 3.

[0076] By configuring the information processing device 1 so that it does not execute a task more than a predetermined number of times, it is possible to avoid a processing loop between the information processing device 1 and the subagent 12.

[0077] In addition, when the same task can be executed by multiple processing methods, the subagent 12 may be configured to execute the task by another method if the task cannot be completed by the specified processing method. That is, the subagent 12 may execute the task to be processed based on a first processing method of the task to be processed indicated by the second instruction text, and if the task to be processed cannot be completed based on the first processing method, execute the task to be processed based on a second processing method different from the first processing method.

[0078] In the above-mentioned use case of invoice processing, the task analysis unit 122 identifies the processing method as follows and outputs code for executing the task. In the processing method below, if invoice information cannot be extracted by method (D1), invoice information is extracted by method (D3), and if invoice information cannot be extracted by method (D3), a task is executed to output an error message.

[0079] (a) Obtain the email identification information (ID, subject, date and time of sending, etc.) that identifies the email and the link to obtain the PDF file contained in the email sent from A's invoice management system. (b) Based on the email identification information obtained in (a), an email containing a password corresponding to the email identification information is identified, and the password contained in the identified email is obtained. (c) Download the PDF file using the link and password obtained. (D1) The acquired PDF file is input into OCR processing service A, and the information contained in the invoice is extracted. (D2) If the invoice information is successfully extracted, proceed to (E). (D3) Input the acquired PDF file into OCR processing service A and extract the information contained in the invoice. (D4) If the invoice information is successfully extracted, proceed to (E). (D5) Output an error message. (O) B Log in to the invoice management system. (K) Enter the information extracted in (D1) or (D3) into B's invoice management system.

[0080] The task execution unit 123 executes the task based on the code generated by the task analysis unit 122. By configuring the subagent 12 in this way, the information processing device 1 can autonomously execute tasks.

[0081] Incidentally, when a business to be processed is made up of a plurality of tasks to be processed sequentially, the supervisor agent 11 may be configured to sequentially generate a plurality of second instruction texts that differ for each task and output the second instruction texts to each of the subagents 12. The processing in this case will be explained using the use case of clearing processing of billing information as an example.

[0082] In this case, the first instruction text states, "For sent invoices, if payment has been made, perform a clearing process for the invoice information, and if payment has not been made, send a follow-up email to the purchaser." In this case, the business specification data is assumed to have registered a method for obtaining issued invoice information, a method for obtaining bank account payment information, and a clearing process method.

[0083] The supervisor agent 11 identifies the subagents 12 corresponding to each of the plurality of tasks to be processed, and outputs a second instruction text indicating the content of each of the plurality of tasks to be processed to each of the identified subagents 12.

[0084] Based on the first instruction text and the business specification data, the task identification unit 112 generates second instruction data indicating "obtain issued invoice data," "obtain payment information," "perform settlement processing for reversible invoice data," and "output non-reversible invoice data as the execution result." The output unit 113 then outputs the generated second instruction data to the file loader agent 12B.

[0085] The subagent 12 outputs the execution result of the processing target task corresponding to the input second instruction text to the supervisor agent 11. In this case, the execution result output by the file loader agent 12B includes information indicating the invoice data that could not be cleared.

[0086] When the execution result of the processing target task executed by the subagent 12 is input from the subagent 12, the supervisor agent 11 determines the content of the processing target task subsequent to the processing target task based on the execution result, and outputs second instruction text indicating the determined content. The task identification unit 112 generates second instruction text for instructing the communication agent 12A to send a follow-up email to the person in charge of the invoice data based on the invoice data that could not be cleared, which is indicated by the execution result. The output unit 113 outputs the generated second instruction text to the communication agent 12A.

[0087] When the relationships between tasks are complex, the information processing device 1 causes the task plan agent 12C to determine the processing order of the tasks. The storage unit 14 has an order determination model that has been trained to output the processing order of each of the multiple tasks based on second instruction texts that include the contents of the multiple tasks to be processed.

[0088] That is, when the complexity of the task to be processed is equal to or greater than a predetermined level, the supervisor agent 11 causes the task plan agent 12C to determine the processing order of the tasks to be processed. The complexity being equal to or greater than a predetermined level means, for example, when the number of tasks included in the task is equal to or greater than a threshold, when the number of types of tasks included in the task is equal to or greater than a threshold, or when the number of branch processes is equal to or greater than a threshold. In this case, the task identification unit 112 generates second instruction text that includes the content of the tasks that make up the task to be processed and instructs the task plan agent 12C to determine the processing order of the tasks. Then, the output unit 113 outputs the generated second instruction text to the task plan agent 12C.

[0089] The task analysis unit 122 of the task plan agent 12C inputs the second instruction text into the order determination model and causes the model to output an execution result indicating the execution order of each task. The output unit 124 outputs the execution result indicating the execution order of the tasks to the supervisor agent 11.

[0090] Then, the supervisor agent 11 outputs a plurality of second instruction texts to the subagents 12 corresponding to the plurality of tasks to be processed, respectively, based on the processing order, for causing the subagents 12 to execute the plurality of tasks to be processed, respectively.

[0091] By configuring the information processing device 1 in this way so that the task plan agent 12C adjusts the execution order of tasks, it becomes possible to autonomously execute complex tasks.

[0092] <Other use cases> An example of a use case of the information processing device 1 is to download a form input into an EDI system, convert the data format, and input the form into another system. In this case, the first instruction text states, "Input the form information input into the EDI system into the accounting system." In addition, the business specification data in this case is assumed to register a method for acquiring form information from the EDI system, a method for converting the data format, and a method for registering form data into the accounting system.

[0093] In this case, as already explained, supervisor agent 11, having acquired the first instruction text, inputs the first instruction text and the business specification data into the task identification model and generates second instruction text for causing file loader agent 12B to execute the task to be processed. In this case, the second instruction text includes tasks for causing file loader agent 12B to execute tasks such as downloading a form file from the EDI system, converting the format of the data included in the form file into a format for registration in the accounting system, and registering the data in the accounting system. Upon receiving the second instruction text, file loader agent 12B generates code corresponding to the second instruction text and executes the code corresponding to the second instruction text.

[0094] [Processing flow in information processing device 1] Fig. 7 is a flowchart showing the flow of processing in the information processing device 1. The flowchart shown in Fig. 7 starts from the point when the supervisor agent 11 acquires the first instruction text.

[0095] The acquisition unit 111 acquires a first instruction text (S01). The task identification unit 112 acquires task specification data (S02). The task identification unit 112 inputs the first instruction text and the task specification data into a task generation model, causes agent identification information to be output, and generates second instruction text (S03). The output unit 113 outputs the second instruction text to the subagent 12 indicated by the agent identification information (S04).

[0096] The task analysis unit 122 inputs the second instruction text into the execution code generation model and generates code for executing the task (S05). The task execution unit 123 executes the generated code (S06). The output unit 124 outputs the execution result of the task to the supervisor agent 11 (S07).

[0097] The determination unit 114 determines whether the execution result is valid based on the execution result and the second instruction text (S08). If the execution result of the task is valid (YES in S08), the information processing device 1 ends the process.

[0098] If the task execution result is invalid (NO in S08), the determination unit 114 determines whether the number of times the task has been executed, which is included in the task execution result, is equal to or greater than a predetermined number. If the number of times the task has been executed is equal to or greater than the predetermined number (YES in S09), a notification is issued that the task cannot be completed (S12), and the information processing device 1 ends the process.

[0099] If the number of executions of the task included in the task execution result is less than the predetermined number of times (NO in S09), the task identification unit 112 generates a third instruction text (S10). The output unit 113 outputs the third instruction text to the subagent 12 (S11), and the process proceeds to S05.

[0100] [Effects of information processing device 1] The information processing device 1 is configured to perform information processing corresponding to the received instructions using a supervisor agent 11 that breaks down the work indicated by the received instructions into tasks, and a subagent 12 that executes tasks based on instructions from the supervisor agent 11, thereby providing information processing that autonomously responds to various instructions from the user.

[0101] <Modification> In the above explanation, an example has been described in which a supervisor agent and a subagent operate in information processing device 1, but the information processing system may also be configured in which each agent operates in each of multiple information processing devices. The configuration of an information processing system SA according to a modified example is shown in FIG. 8. The information processing system SA is configured from an information processing device 1A and multiple information processing devices 1B. Information processing device 1A has a supervisor agent 11. Information processing device 1B has a subagent 12. Information processing device 1A and information processing device 1B are connected to each other so as to be able to communicate with each other via a network.

[0102] The present invention has been described above using embodiments, but the technical scope of the present invention is not limited to the scope described in the above embodiments, and various modifications and alterations are possible within the spirit and scope of the present invention. For example, although the information processing device 1 has been described as executing tasks using multiple subagents, a task may be executed by a single subagent. Furthermore, for example, all or part of the device may be configured by functionally or physically distributing or integrating in any unit. Furthermore, new embodiments resulting from any combination of multiple embodiments are also included in the embodiments of the present invention. The effects of the new embodiments resulting from the combination also have the effects of the original embodiments. [Explanation of symbols]

[0103] 1. Information processing equipment 2 External device 3. Information terminals 11 Supervisor Agent 12 Subagents 13 Communications Department 14 Storage section 15 Control Unit 111 Acquisition Department 112 Task Identification Unit 113 Output section 114 Judgment section 121 Acquisition Department 122 Task Analysis Unit 123 Task Execution Unit 124 Output section

Claims

1. An information processing device that executes information processing using a first agent and a plurality of second agents each capable of executing different tasks, When a first instruction text indicating the content of a business to be processed is input, the first agent identifies the second agent corresponding to a task to be processed that constitutes the business to be processed, references business specification data that associates a reference task with one or more processing methods of the reference task, and outputs to the second agent a second instruction text indicating the content of the task to be processed, the second instruction text indicating the processing method of the reference task that corresponds to the task to be processed; the second agent executes a task indicated by the second instruction text in response to receiving the second instruction text; Information processing device.

2. In the business specification data, the reference task is associated with one or more second agents capable of executing the reference task, the first agent refers to the business specification data to identify the second agent capable of executing the reference task corresponding to the target task, and outputs the second instruction text to the identified second agent; The information processing device according to claim 1 .

3. When a result of execution of the target task by the second agent to which the second instruction text has been output does not satisfy a predetermined condition, the first agent outputs the second instruction text to another second agent that can execute the reference task corresponding to the target task. The information processing device according to claim 2 .

4. An information processing device that executes information processing using a first agent and a plurality of second agents each capable of executing different tasks, when a first instruction text indicating the content of a task to be processed is input, the first agent identifies the second agent corresponding to a task to be processed that constitutes the task to be processed, and outputs a second instruction text indicating the content of the task to be processed to the identified second agent; the second agent, in response to acquiring the second instruction text, executes the task to be processed based on a first processing method of the task to be processed indicated by the second instruction text, and, if the task to be processed is not completed based on the first processing method, executes the task to be processed based on a second processing method different from the first processing method. Information processing device.

5. An information processing device that executes information processing using a first agent and a plurality of second agents each capable of executing different tasks, when a first instruction text indicating the content of a task to be processed is input, the first agent identifies the second agent corresponding to a task to be processed that constitutes the task to be processed, and outputs a second instruction text indicating the content of the task to be processed to the identified second agent; the second agent, in response to receiving the second instruction text, executes a task indicated by the second instruction text; the second agent outputs an execution result of the target task to the first agent; the first agent determines whether the execution result is valid based on the first instruction text and the execution result, and if the execution result is invalid, outputs to the second agent a third instruction text indicating a processing method different from that of the second instruction text; Information processing device.

6. The first agent determining whether the number of times the processing target task, for which the execution result has been input from the second agent in response to the input of one of the first instruction texts, has been executed is equal to or greater than a predetermined number of times; If the number of times that the task to be processed has been executed is equal to or greater than a predetermined number of times, a message is output indicating that the task to be processed cannot be executed. The information processing device according to claim 5 .

7. An information processing device that executes information processing using a first agent and a plurality of second agents each capable of executing different tasks, when a first instruction text indicating the content of a task to be processed is input, the first agent identifies the second agent corresponding to a task to be processed that constitutes the task to be processed, and outputs a second instruction text indicating the content of the task to be processed to the identified second agent; the second agent, in response to receiving the second instruction text, executes a task indicated by the second instruction text; the business to be processed is composed of a plurality of tasks to be processed that are sequentially processed, the first agent identifies a plurality of second agents corresponding to the plurality of tasks to be processed, respectively, and outputs second instruction text indicating the content of each of the plurality of tasks to be processed to each of the identified second agents; each of the plurality of second agents outputs to the first agent an execution result of the processing target task corresponding to the input second instruction text; when the first agent receives an execution result of the process target task executed by the second agent from the second agent, the first agent determines the content of the process target task subsequent to the process target task based on the execution result, and outputs the second instruction text indicating the determined content. Information processing device.

8. An information processing device that executes information processing using a first agent and a plurality of second agents each capable of executing different tasks, when a first instruction text indicating the content of a task to be processed is input, the first agent identifies the second agent corresponding to a task to be processed that constitutes the task to be processed, and outputs a second instruction text indicating the content of the task to be processed to the identified second agent; the second agent, in response to receiving the second instruction text, executes a task indicated by the second instruction text; the business to be processed is composed of a plurality of the tasks to be processed, The first agent when the complexity of the task to be processed is equal to or greater than a predetermined level, having a task plan agent for determining a processing order of the plurality of tasks to be processed determine a processing order of the plurality of tasks to be processed; outputting, based on the processing order, a plurality of second instruction texts to a plurality of second agents corresponding to the plurality of tasks to be processed, for causing the plurality of tasks to be processed, respectively; Information processing device.

9. An information processing device that executes information processing by a first agent and a file loader agent that can operate a file, When a first instruction text indicating the content of a task to be processed is input, the first agent refers to task specification data that associates a reference task with one or more processing methods for the reference task, and if the reference task corresponding to the task to be processed includes a file operation, outputs a second instruction text including the reference task corresponding to the task to be processed to the file loader agent; the file loader agent executes a task indicated by the second instruction text in response to receiving the second instruction text; Information processing device.

10. An information processing method in which a computer executes information processing using a first agent and a plurality of second agents each capable of executing a different task, the method comprising: When a first instruction text indicating the content of a business to be processed is input, the first agent identifies the second agent corresponding to a task to be processed that constitutes the business to be processed, references business specification data that associates a reference task with one or more processing methods of the reference task, and outputs to the second agent a second instruction text indicating the content of the task to be processed, the second instruction text indicating the processing method of the reference task that corresponds to the task to be processed; executing a task indicated by the second instruction text by the second agent in response to receiving the second instruction text; An information processing method comprising:

11. An information processing method in which a computer executes information processing using a first agent and a plurality of second agents each capable of executing a different task, the method comprising: a step in which, when a first instruction text indicating the content of a task to be processed is input to the first agent, the first agent identifies the second agent corresponding to a task to be processed that constitutes the task to be processed, and outputs a second instruction text indicating the content of the task to be processed to the identified second agent; a step in which the second agent, in response to acquiring the second instruction text, executes the task to be processed based on a first processing method of the task to be processed indicated by the second instruction text, and, if the task to be processed is not completed based on the first processing method, executes the task to be processed based on a second processing method different from the first processing method; An information processing method comprising:

12. An information processing method in which a computer executes information processing using a first agent and a plurality of second agents each capable of executing a different task, the method comprising: a step in which, when a first instruction text indicating the content of a task to be processed is input to the first agent, the first agent identifies the second agent corresponding to a task to be processed that constitutes the task to be processed, and outputs a second instruction text indicating the content of the task to be processed to the identified second agent; executing a task indicated by the second instruction text by the second agent in response to receiving the second instruction text; and the second agent outputs an execution result of the target task to the first agent; the first agent determines whether the execution result is valid based on the first instruction text and the execution result, and if the execution result is invalid, outputs to the second agent a third instruction text indicating a processing method different from that of the second instruction text; Information processing methods.

13. An information processing method in which a computer executes information processing using a first agent and a plurality of second agents each capable of executing a different task, the method comprising: a step in which, when a first instruction text indicating the content of a task to be processed is input to the first agent, the first agent identifies the second agent corresponding to a task to be processed that constitutes the task to be processed, and outputs a second instruction text indicating the content of the task to be processed to the identified second agent; executing a task indicated by the second instruction text by the second agent in response to receiving the second instruction text; and the business to be processed is composed of a plurality of tasks to be processed that are sequentially processed, the first agent identifies a plurality of second agents corresponding to the plurality of tasks to be processed, respectively, and outputs second instruction text indicating the content of each of the plurality of tasks to be processed to each of the identified second agents; each of the plurality of second agents outputs to the first agent an execution result of the processing target task corresponding to the input second instruction text; when the first agent receives an execution result of the process target task executed by the second agent from the second agent, the first agent determines the content of the process target task subsequent to the process target task based on the execution result, and outputs the second instruction text indicating the determined content. Information processing methods.

14. An information processing method in which a computer executes information processing using a first agent and a plurality of second agents each capable of executing a different task, the method comprising: a step in which, when a first instruction text indicating the content of a task to be processed is input to the first agent, the first agent identifies the second agent corresponding to a task to be processed that constitutes the task to be processed, and outputs a second instruction text indicating the content of the task to be processed to the identified second agent; executing a task indicated by the second instruction text by the second agent in response to receiving the second instruction text; and the business to be processed is composed of a plurality of the tasks to be processed, The first agent when the complexity of the task to be processed is equal to or greater than a predetermined level, having a task plan agent for determining a processing order of the plurality of tasks to be processed determine a processing order of the plurality of tasks to be processed; outputting, based on the processing order, a plurality of second instruction texts to a plurality of second agents corresponding to the plurality of tasks to be processed, for causing the plurality of tasks to be processed, respectively; Information processing methods.

15. An information processing method in which a computer executes information processing using a first agent and a file loader agent capable of manipulating files, comprising: a step in which, when a first instruction text indicating the content of a task to be processed is input, the first agent refers to task specification data that associates a reference task with one or more processing methods for the reference task, and, if the reference task corresponding to the task to be processed includes a file operation, outputs a second instruction text including the reference task corresponding to the task to be processed to the file loader agent; executing, by the file loader agent, a task indicated by the second instruction text in response to receiving the second instruction text; An information processing method comprising:

16. An information processing system comprising: an information processing device that executes information processing using a first agent and a plurality of second agents each capable of executing different tasks; and an information terminal that is capable of communicating with the information processing device, In the information processing device, When a first instruction text indicating the content of a business to be processed is input via the information terminal, the first agent identifies the second agent corresponding to a task to be processed that constitutes the business to be processed, references business specification data that associates a reference task with one or more processing methods of the reference task, and outputs to the second agent a second instruction text indicating the content of the task to be processed, the second instruction text indicating the processing method of the reference task that corresponds to the task to be processed, the second agent executes a task indicated by the second instruction text in response to receiving the second instruction text; Information processing system.

17. An information processing system comprising: an information processing device that executes information processing using a first agent and a plurality of second agents each capable of executing different tasks; and an information terminal that is capable of communicating with the information processing device, In the information processing device, when a first instruction text indicating the content of a business to be processed is input via the information terminal, the first agent identifies the second agent corresponding to a task to be processed that constitutes the business to be processed, and outputs a second instruction text indicating the content of the task to be processed to the identified second agent; the second agent, in response to acquiring the second instruction text, executes the task to be processed based on a first processing method of the task to be processed indicated by the second instruction text, and, if the task to be processed is not completed based on the first processing method, executes the task to be processed based on a second processing method different from the first processing method. Information processing system.

18. An information processing system comprising: an information processing device that executes information processing using a first agent and a plurality of second agents each capable of executing different tasks; and an information terminal that is capable of communicating with the information processing device, In the information processing device, when a first instruction text indicating the content of a business to be processed is input via the information terminal, the first agent identifies the second agent corresponding to a task to be processed that constitutes the business to be processed, and outputs a second instruction text indicating the content of the task to be processed to the identified second agent; the second agent, in response to receiving the second instruction text, executes a task indicated by the second instruction text; the second agent outputs an execution result of the target task to the first agent; the first agent determines whether the execution result is valid based on the first instruction text and the execution result, and if the execution result is invalid, outputs to the second agent a third instruction text indicating a processing method different from that of the second instruction text; Information processing system.

19. An information processing system comprising: an information processing device that executes information processing using a first agent and a plurality of second agents each capable of executing different tasks; and an information terminal that is capable of communicating with the information processing device, In the information processing device, when a first instruction text indicating the content of a business to be processed is input via the information terminal, the first agent identifies the second agent corresponding to a task to be processed that constitutes the business to be processed, and outputs a second instruction text indicating the content of the task to be processed to the identified second agent; the second agent, in response to receiving the second instruction text, executes a task indicated by the second instruction text; the business to be processed is composed of a plurality of tasks to be processed that are sequentially processed, the first agent identifies a plurality of second agents corresponding to the plurality of tasks to be processed, respectively, and outputs second instruction text indicating the content of each of the plurality of tasks to be processed to each of the identified second agents; each of the plurality of second agents outputs to the first agent an execution result of the processing target task corresponding to the input second instruction text; when the first agent receives an execution result of the process target task executed by the second agent from the second agent, the first agent determines the content of the process target task subsequent to the process target task based on the execution result, and outputs the second instruction text indicating the determined content. Information processing system.

20. An information processing system comprising: an information processing device that executes information processing using a first agent and a plurality of second agents each capable of executing different tasks; and an information terminal that is capable of communicating with the information processing device, In the information processing device, when a first instruction text indicating the content of a business to be processed is input via the information terminal, the first agent identifies the second agent corresponding to a task to be processed that constitutes the business to be processed, and outputs a second instruction text indicating the content of the task to be processed to the identified second agent; the second agent, in response to receiving the second instruction text, executes a task indicated by the second instruction text; the business to be processed is composed of a plurality of the tasks to be processed, The first agent when the complexity of the task to be processed is equal to or greater than a predetermined level, having a task plan agent for determining a processing order of the plurality of tasks to be processed determine a processing order of the plurality of tasks to be processed; outputting, based on the processing order, a plurality of second instruction texts to a plurality of second agents corresponding to the plurality of tasks to be processed, for causing the plurality of tasks to be processed, respectively; Information processing system.

21. An information processing system comprising: an information processing device that executes information processing using a first agent and a file loader agent that can operate files; and an information terminal that can communicate with the information processing device, In the information processing device, when a first instruction text indicating the content of a task to be processed is input via the information terminal, the first agent refers to task specification data that associates a reference task with one or more processing methods for the reference task, and if the reference task corresponding to the task to be processed includes a file operation, outputs a second instruction text including the reference task corresponding to the task to be processed to the file loader agent; the file loader agent executes a task indicated by the second instruction text in response to receiving the second instruction text; Information processing system.

Images