Information processing system, information processing method, information processing program, and AI agent

Abstract

An information processing system, an information processing method, an information processing program, and an AI agent are provided that improve the accuracy of task execution by allocating tasks to a plurality of tasks and AI agents that process the tasks, and evaluating and reviewing the content of the AI ​​agents that process the tasks. [Solution] In the information processing system 10, the AI ​​agent comprises an agent allocation means for allocating an AI agent (subagent) corresponding to each of a plurality of tasks that make up a specified task, a processing execution means for executing the specified task by having the subagent execute the task, an agent evaluation means for evaluating the subagent based on the execution results of the specified task, and an agent generation means for generating a new subagent based on the evaluation of the subagent, and the agent allocation means updates the allocation of the subagent based on the evaluation of the subagent.

Description

[Technical Field]

[0001] The present invention relates to an information processing system using an AI agent, an information processing method, an information processing program, and an AI agent. [Background technology]

[0002] Conventionally, a management system has been proposed that uses a generative AI interface to visualize and classify tasks across the entire company and appropriately assign them to each employee. The AI ​​monitors the progress of tasks in real time and reallocates tasks as necessary (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2025-52500 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, while conventional systems can assign tasks using AI, they were designed for company management purposes, which meant that their scope of application was limited.

[0005] In view of the above situation, the present invention aims to provide an information processing system, an information processing method, an information processing program, and an AI agent that can dramatically improve the accuracy of executing a specified task by allocating the specified task to multiple tasks and AI agents that process the tasks, and by evaluating and reviewing the content of the AI ​​agents that process the tasks. [Means for solving the problem]

[0006] The information processing system of the present invention is an information processing system configured with an AI agent, wherein the AI ​​agent comprises an agent allocation means for assigning an AI agent (hereinafter referred to as a "subagent") corresponding to each of a plurality of tasks that make up a predetermined task, a processing execution means for executing the predetermined task by having the subagent execute the task, an agent evaluation means for evaluating the subagent based on the execution result of the predetermined task, and an agent generation means for generating a new subagent based on the evaluation of the subagent, wherein the agent allocation means updates the allocation of the subagent based on the evaluation of the subagent.

[0007] Furthermore, the information processing method of the present invention is an information processing method executed using an AI agent, which comprises an agent allocation step of assigning AI agents (hereinafter referred to as "subagents") corresponding to each of a plurality of tasks that make up a predetermined task, a processing execution step of executing the predetermined task by having the subagents execute the tasks, an agent evaluation step of evaluating the subagents based on the execution results of the predetermined tasks, and an agent generation step of generating new subagents based on the evaluation of the subagents, wherein the agent allocation step updates the allocation of the subagents based on the evaluation of the subagents.

[0008] In addition, the information processing program of the present invention is a program for an information processing system configured with an AI agent, and causes the AI ​​agent, which is a computer, to function as an agent allocation means that allocates AI agents (hereinafter referred to as "subagents") corresponding to each of a plurality of tasks that make up a predetermined task, a processing execution means that executes the predetermined task by having the subagents execute the tasks, an agent evaluation means that evaluates the subagents based on the results of execution of the predetermined tasks, and an agent generation means that generates new subagents based on the evaluation of the subagents, and the agent allocation means updates the allocation of the subagents based on the evaluation of the subagents.

[0009] Furthermore, the AI ​​agent of the present invention is an AI agent that acts on behalf of a human and has the ability to learn and make decisions on its own, and is characterized in that it causes a computer to function as an agent allocation means that allocates AI agents (hereinafter referred to as "subagents") corresponding to each of a plurality of tasks that make up a predetermined task, a processing execution means that executes the predetermined task by having the subagents execute the tasks, an agent evaluation means that evaluates the subagents based on the results of execution of the predetermined tasks, and an agent generation means that generates new subagents based on the evaluation of the subagents, and the agent allocation means updates the allocation of the subagents based on the evaluation of the subagents. [Effects of the Invention]

[0010] The information processing system, information processing method, information processing program, and AI agent according to the present invention can achieve the excellent effect of dramatically improving the accuracy of execution of a specified task by allocating the specified task to multiple tasks and the AI ​​agents that process the tasks, and by evaluating and reviewing the content of the AI ​​agents that process the tasks. [Brief explanation of the drawings]

[0011] [Figure 1] 1 is a schematic diagram showing an overview of an information processing system 10 according to the present embodiment. [Figure 2] 1 is a system configuration diagram showing an example of the configuration of an information processing system 10 according to the present embodiment. [Figure 3] 1A is a diagram showing an example of task allocation, FIG. 1B is a diagram showing an example of a combination of tasks and subagents, and FIG. 1C is a diagram showing an example of the execution order of tasks. [Figure 4] (a) is a diagram showing an example of evaluation of subagents and reassignment of subagents, (b) is a diagram showing another example of evaluation of subagents and reassignment of subagents. DETAILED DESCRIPTION OF THE INVENTION

[0012] An information processing system 10 according to an embodiment of the present invention will be described below with reference to the drawings.

[0013] <System Overview> First, an overview of an information processing system 10 according to this embodiment will be described with reference to Fig. 1. Fig. 1 is a schematic diagram showing an overview of the information processing system 10 according to this embodiment.

[0014] The information processing system 10 is an information processing system configured with an AI agent 12, which comprises task allocation means 12a that allocates a predetermined task to a plurality of tasks, agent allocation means 12b that assigns an AI agent (hereinafter referred to as a "subagent") corresponding to the task, process execution means 12c that executes the predetermined task by having the subagent execute the task, agent evaluation means 12d that evaluates the subagent based on the execution result of the predetermined task, and agent generation means 12e that generates a new subagent based on the evaluation of the subagent, and the agent allocation means 12b is an information processing system characterized in that it updates the allocation of the subagent based on the evaluation of the subagent.

[0015] According to the information processing system 10, the accuracy of execution of a specified task can be dramatically improved by allocating the specified task to multiple tasks and AI agents that process the tasks, and by evaluating and reviewing the content of the AI ​​agents that process the tasks.

[0016] Here, "AI agent" refers to a system that acts as a proxy (agent) for humans, has the ability to learn and make decisions on its own, and integrates various AI technologies to solve complex problems.

[0017] When a task occurs based on an external event (e.g., a question (prompt) from a user, a signal input from a sensor) or an internal event (e.g., activation by an internal scheduler, the occurrence of an abnormality), the AI ​​agent executes the task and outputs the specified result.

[0018] For example, when a question (prompt) is given as an external event, the system executes tasks such as searching for an answer to the question, generating an answer, and outputs the answer to the question. Also, when a problem (prompt) is given as an external event, the system executes tasks such as searching for a solution to the problem, generating a solution, and outputs the solution to the problem.

[0019] Here, "prompt" refers to an instruction or question given by a user to the system, and includes, for example, an AI prompt given to a generating AI, or a command prompt that gives instructions using a command.

[0020] A "task" is an operation required to achieve a certain purpose or task. Multiple or all "tasks" may be executed in parallel, each in a predetermined order, each in a predetermined priority order, or each in a random order.

[0021] The "predetermined task" according to the present invention may be a task generated based on an external event (e.g., a question (prompt) from a user, a signal input from a sensor), or may be a task generated (defined) by the AI ​​agent itself in response to an ambiguous or abstract question or task (prompt) input by a user. It may also be a task generated based on an internal event (e.g., activation by an internal scheduler, occurrence of an abnormality).

[0022] The "task" according to the present invention may be a task relating to business figures or a task relating to purchasing history.

[0023] With this configuration, tasks related to business figures and purchasing history can be divided into multiple tasks, and the content of the tasks can be evaluated and reviewed, thereby dramatically improving the accuracy of tasks related to business figures and purchasing history.

[0024] Here, "management figures" refer to figures that indicate the business condition of a company and are used to evaluate the performance of a company through management indicators and financial indicators. In addition, "data related to business figures" refers to data (information such as numbers and character strings) that indicate the business status of a company, and "data related to business figures" includes, for example, (1) sales and profit-related data (e.g., sales data, profit data), (2) sales and product management-related data (e.g., product master and specifications, shelf layout data, sales performance and sales promotion effects), (3) inventory and purchasing-related data (e.g., inventory status, order and purchasing data), (4) customer and marketing-related data (e.g., customer ID data (ID-POS), marketing measure results), (5) store operation and personnel management-related data (e.g., store performance, staff and shifts), (6) expense and finance-related data (e.g., store and headquarters expenses, financial indicators), (7) external environment data (e.g., trade area / population data, location and price information of competing stores, weather, temperature, and disaster information), (8) local event / school event information, and (9) instruction and communication-related data (e.g., work instructions and policies from superiors, reports and proposals from the field, and collaboration history with headquarters).

[0025] Furthermore, among the "data related to management figures," (2) sales and product management related data (e.g., product master and specifications, shelf layout data, sales performance and promotional effects) may be referred to as "data related to sales promotion," and among the "data related to management figures," (4) customer and marketing related data (e.g., customer ID data (ID-POS), results of marketing measures) may be referred to as "data related to purchase history."

[0026] <System configuration example> Next, a configuration example of the information processing system 10 according to this embodiment will be described with reference to Fig. 2. Fig. 2 is a system configuration diagram showing a configuration example of the information processing system 10 according to this embodiment.

[0027] The information processing system 10 can be configured to include, for example, a system terminal 12 that controls the entire system, and an external terminal 16 that is connected to the system terminal 12 via a network NW so that they can communicate with each other.

[0028] The system terminal 12 is a terminal that constitutes an AI agent and controls the entire information processing system 10, and is configured from a conventionally known server, personal computer, etc. In this example, the system terminal 12 is configured from one server, but it may also be configured from multiple servers, personal computers, etc. The hardware configuration of the system terminal 12 and the programs executed by the system terminal 12 will be described later.

[0029] The external terminal 16 is a terminal used by a user (an individual or a specific group consisting of multiple people (a team, a department, a company, etc.)) who uses the information processing system 10, and is configured as a personal computer, a tablet, a smartphone, etc. The type of the external terminal 16 is not particularly limited, but examples thereof include a smartphone, a personal computer, a tablet, etc. used by an individual, and a smartphone, a personal computer, a tablet, etc. used by a company employee or a specific group, etc.

[0030] The network NW is a line that allows the system terminal 12 and the external terminal 16 to communicate with each other, and is typically configured as a WAN (Wide Area Network), also known as the Internet. The network NW may be wired or wireless, may be a LAN (Local Area Network), may be a dedicated line such as a VPN (Virtual Private Network), or may be a combination of these lines.

[0031] <System terminal / hardware configuration example> Next, an example of the hardware configuration of the system terminal 12 will be described.

[0032] As shown in FIG. 2, the system terminal 12 is configured to include, for example, a CPU 21, a ROM 22, a RAM 23, an external storage drive 25, a storage device 26, an input device 27, a display device 28, a communication unit 29, etc., all of which are connected to the CPU 21 via a bus.

[0033] The CPU 21 is a control means for controlling the entire system terminal 12, and performs processes such as executing application programs and operating systems (OS) stored in the ROM 22, storage device 26, etc., and storing data and files necessary for executing programs in the RAM 23, storage device 26, etc.

[0034] The ROM 22 is a storage means for storing basic I / O programs and various data, and is configured, for example, by a PROM, flash memory, etc. The RAM 23 is a storage means for temporarily storing data, and is configured, for example, by an SDRAM, DRAM, etc. The external storage drive 25 is a control means capable of reading and writing data from and to a recording medium 24 such as a magnetic tape or DVD, and is configured, for example, by a magnetic tape storage, DVD drive, etc.

[0035] The storage device 26 is a storage means for storing application programs, OS, control programs, related programs, various information, etc., and is configured, for example, by a hard disk (HDD), solid state drive (SDD), etc. The input device 27 is used to input commands (instructions) etc. to the system terminal 12, and is configured, for example, by a keyboard, a pointing device (mouse, etc.), a touch panel, etc.

[0036] The display device 28 displays commands input by the input device 27, response outputs from the system terminal 12 to the commands, various displays, etc., and is configured, for example, by a liquid crystal display, a plasma display, an organic EL display, etc. The communication unit 29 is control means for controlling communication with the external terminal 16, etc. via the network NW, and is configured, for example, by a communication card, etc.

[0037] <Management terminal / function> Next, the functions of the system terminal 12 will be described.

[0038] The storage device 26 of the system terminal 12 stores a program (information processing program) that causes the system terminal 12 to function as a task allocation means 12a, an agent allocation means 12b, a process execution means 12c, an agent evaluation means 12d, and an agent generation means 12e.

[0039] <Management terminal / function / task allocation method> Next, the task allocation means 12a will be described.

[0040] The task allocation means 12a is a means for allocating a predetermined task to multiple tasks and updating the allocation of the multiple tasks based on the evaluation of the tasks, and in this example, is composed of a program stored in the memory device 26 of the system terminal 12, the memory device 26, etc.

[0041] The task allocation means 12a allocates predetermined tasks given to the AI ​​agent 12 from inside or outside into multiple tasks. As mentioned above, a "task" is an operation required to achieve a certain purpose or task. Multiple or all of the "tasks" may be executed in parallel, each in a predetermined order, each in a predetermined priority order, or each in a random order.

[0042] For example, when a task is given to the AI ​​agent 12 from outside the AI ​​agent 12, the task allocation means 12a allocates the processing required to achieve the task to multiple processes based on the knowledge acquired through machine learning, and stores each allocated process as a task (subtask) in the storage device 26 in association with the user's ID.

[0043] For example, when a task (prompt) such as "Predict sales for product A" is given from the external terminal 16 as a predetermined task, the task allocation means 12a breaks down the processing required to achieve the task (predict sales for product A) into multiple processes and allocates them to tasks based on knowledge acquired through machine learning and knowledge obtained from the external terminal 16.

[0044] The multiple processes required to realize a sales forecast for product a include, for example, (process 1) collecting sales performance data for product a (e.g., collecting POS data), (process 2) collecting external factor data (weather, season, etc.), (process 3) collecting market data (market trends, sales trends of similar products, etc.), (process 4) collecting competitive data (prices of competing products, etc.), (process 5) selecting a forecast model (moving average method, exponential smoothing method, etc.), (process 6) creating a sales forecast, and (process 7) visualizing the sales forecast (creating a forecast graph, etc.).

[0045] The task allocation means 12a allocates each of the multiple processes (Process 1) to (Process 7) thus decomposed into tasks 1 to 7 as shown in Figure 3(a), and then stores the tasks 1 to 7 in the storage device 26 in association with the user's ID.

[0046] <Management terminal / function / agent allocation method> Next, the agent allocation means 12b will be described.

[0047] The agent allocation means 12b is a means for allocating AI agents (subagents) corresponding to tasks and updating the allocation of the subagents based on the evaluation of the subagents, and in this example is composed of a program stored in the memory device 26 of the system terminal 12, the memory device 26, etc.

[0048] The agent allocation means 12b refers to the multiple tasks stored in the storage device 26 by the task allocation means 12a, assigns an appropriate subagent to each of the multiple tasks, and stores the combination of the subagent and task in the storage device 26 in association with the user's ID.

[0049] For example, consider a case where a predetermined task is given from the external terminal 16 as a prompt to "make a sales forecast for product A," and multiple tasks are stored in the storage device 26, including (task 1) collection of sales performance data for product A, (task 2) collection of external factor data (weather, season, etc.), (task 3) collection of market data (market trends, sales trends of similar products, etc.), (task 4) collection of competitive data (prices of competing products, etc.), (task 5) selection of a forecast model (moving average method, exponential smoothing method, etc.), (task 6) creation of a sales forecast, and (task 7) visualization of the sales forecast (creating a forecast graph, etc.).

[0050] In this case, as shown in Figure 3(b), the agent allocation means 12b assigns, for example, subagent 1 that has learned to collect POS data through machine learning to "(Task 1) Collect sales performance data for product a," and stores the combination of subagent 1 and task 1 in the storage device 26 in association with the user's ID.

[0051] Next, the agent allocation means 12b assigns, for example, subagent 2 that has learned to collect external factor data through machine learning to "(Task 2) Collection of external factor data (weather, season, etc.)", and stores the combination of subagent 2 and task 2 in the storage device 26 in association with the user's ID.

[0052] Next, the agent allocation means 12b assigns, for example, a subagent 3 that has learned to collect market data through machine learning to "(Task 3) Collect market data (market trends, sales trends of similar products, etc.)", and stores the combination of the subagent 3 and task 3 in the storage device 26 in association with the user's ID.

[0053] Next, the agent allocation means 12b assigns, for example, a subagent 4 that has learned to collect competitive data through machine learning to "(Task 4) Collect competitive data (prices of competing products, etc.)", and stores the combination of the subagent 4 and task 4 in the storage device 26 in association with the user's ID.

[0054] Next, the agent allocation means 12b assigns, for example, a subagent 5 that has learned how to select a prediction model through machine learning to "(Task 5) Selection of a prediction model (moving average method, exponential smoothing method, etc.)", and stores the combination of the subagent 5 and task 5 in the storage device 26 in association with the user's ID.

[0055] Next, the agent allocation means 12b assigns, for example, a subagent 6 that has learned how to create sales forecasts through machine learning to "(Task 6) Creating a sales forecast," and stores the combination of the subagent 6 and task 6 in the storage device 26 in association with the user's ID.

[0056] Next, the agent allocation means 12b assigns, for example, a subagent 7 that has learned how to visualize sales forecasts through machine learning to "(Task 7) Visualizing sales forecasts (creating forecast graphs, etc.)", and stores the combination of the subagent 7 and task 7 in the storage device 26 in association with the user's ID.

[0057] The type of subagent assigned to each task is not limited to this example; for example, the same (or similar) subagent may be assigned to different tasks, or one subagent may be assigned to multiple tasks.

[0058] Therefore, for example, since tasks 1 to 4 have in common the fact that they involve data collection, a single subagent that has learned data collection through machine learning may be assigned to tasks 1 to 4, and the combination of the subagent and tasks 1 to 4 may be stored in storage device 26 in association with the user's ID.

[0059] In addition, when the agent evaluation means 12d (described later) has evaluated a subagent, or the agent generation means 12e (described later) has generated a subagent, the agent allocation means 12b executes a process to update the allocation of the subagent based on the evaluation of the subagent; this process will be described later.

[0060] <Management terminal / function / processing execution means> Next, the process execution means 12c will be described.

[0061] The process execution means 12c is a means for executing a predetermined task by having a subagent execute the task, and in this example is configured by a program stored in the storage device 26 of the system terminal 12, the storage device 26, etc.

[0062] The process execution means 12c refers to the combinations of tasks and subagents stored in the storage device 26 by the agent allocation means 12b, and causes each subagent to execute the corresponding task, thereby executing a predetermined task.

[0063] For example, consider a case where a given task (prompt) is given from external terminal 16 to "predict sales for product A," and the following combinations of tasks and subagents are stored in storage device 26, as shown in Figure 3(b): "(Task 1) Collecting sales performance data for product A" with subagent 1; "(Task 2) Collecting external factor data (weather, season, etc.)" with subagent 2; "(Task 3) Collecting market data (market trends, sales trends of similar products, etc.)" with subagent 3; "(Task 4) Collecting competitive data (prices of competing products, etc.)" with subagent 4; "(Task 5) Selecting a forecast model (moving average method, exponential smoothing method, etc.)" with subagent 5; "(Task 6) Creating a sales forecast" with subagent 6; and "(Task 7) Visualizing the sales forecast (creating a forecast graph, etc.)" with subagent 7.

[0064] First, the processing execution means 12c refers to the multiple tasks stored in the storage device 26, determines whether each task can be executed in parallel or whether each task should be executed in a predetermined order, and determines the optimal execution method that will enable the multiple tasks to be executed most efficiently.

[0065] In the previous case, tasks 1 to 4 are not dependent on each other, so they are judged to be tasks that can be executed in parallel. However, tasks 5 to 7 are dependent on each other, so they are judged to be executed in the order of their task numbers (task 5 → task 6 → task 7) after all tasks 1 to 4 have been completed. As the optimal execution method, it is selected to execute tasks 1 to 4 in parallel, and then execute the remaining tasks 5 to 7 in this order.

[0066] Subsequently, the process execution means 12c executes the predetermined task by having each subagent execute the corresponding task in accordance with the determined optimal execution method.

[0067] In the previous case, as shown in FIG. 3(c), the processing execution means 12c executes the tasks 1 to 4 in parallel to the subagents 1 to 4, and after all tasks 1 to 4 are completed, executes the remaining tasks 5 to 7 to the subagents 5 to 7 in that order, thereby executing the processing for the assignment (predetermined task) of "Predict sales of product A."

[0068] In addition, when the agent evaluation means 12d, which will be described later, completes the process of updating the allocation of the subagent based on the evaluation of the subagent, the process execution means 12c executes the updated task again and outputs a solution (answer) to the specified task (problem); this process will be described later.

[0069] <Management terminal / function / agent evaluation method> Next, the agent evaluation means 12d will be described.

[0070] The agent evaluation means 12d is a means for evaluating a subagent based on the execution result of a predetermined task, and in this example is configured by a program stored in the storage device 26 of the system terminal 12, the storage device 26, and the like.

[0071] First, the agent evaluation means 12d evaluates multiple subagents based on the results of the processing executed by the processing execution means 12c, and stores the combination of the subagents and the evaluations in the storage device 26 in association with the user's ID.

[0072] Here, the evaluation of the subagent is performed using indicators such as the task execution time, the difference between the scheduled task execution time and the task execution time (e.g., whether there is a delay, the number of delays), the utilization rate of resources required to execute the task (e.g., server, CPU, memory, etc.), the accuracy and precision of task execution, the amount of information obtained by executing the task, and the variation in task processing results.

[0073] For example, consider a case where the process executing means 12c executes tasks 1 to 7 to execute a process for the assignment (predetermined task) of "Predict sales of product a" and generate the results of the sales forecast.

[0074] In this case, for example, if the agent evaluation means 12d analyzes the results of the sales forecast based on knowledge acquired through machine learning or knowledge obtained from the external terminal 16 and determines that the market trend is not sufficiently reflected, it determines that the amount of information in (Task 3) "Market data collection" is small and the accuracy is low, and sets the evaluation score of the subagent 3 that executed (Task 3) to, for example, 3 out of 10, as shown in Figure 4(a).

[0075] On the other hand, if the agent evaluation means 12d analyzes the results of the sales forecast based on knowledge acquired through machine learning or knowledge obtained from the external terminal 16 and determines that the accuracy of the sales performance data is high, it determines that the accuracy of (Task 1) ``Collecting sales performance data of product a'' is high, and sets the evaluation score of the subagent 1 that executed (Task 1) to, for example, 9 out of 10, as shown in Figure 4(a).

[0076] The agent evaluation means 12d also evaluates subagents other than subagent 1 and subagent 3 based on knowledge acquired through machine learning and knowledge obtained from the external terminal 16, and sets evaluation scores.Here, we assume that the evaluation scores for other tasks are set to, for example, 6 out of 10, as shown in Figure 4(a).

[0077] <Management terminal / function / agent generation method> Next, the agent generating means 12e will be described.

[0078] The agent generating means 12e is a means for generating a new subagent based on the evaluation of the subagent, and in this example is made up of a program stored in the storage device 26 of the system terminal 12, the storage device 26, and the like.

[0079] The agent generating means 12e refers to the combination of the subagent and the evaluation stored in the storage device 26 by the agent evaluating means 12d, and generates a new subagent based on the evaluation of the subagent.

[0080] For example, in the previous case, the evaluation scores of subagents other than subagent 3 are equal to or greater than a predetermined reference score (e.g., 6), so no new subagents are generated, whereas the evaluation score of subagent 3 is less than the predetermined reference score (e.g., 6), so a new subagent 8 is generated in place of subagent 3.

[0081] Here, in generating a subagent, measures are taken to improve the index (e.g., task execution time, etc.) used in evaluating the subagent. Examples of newly generated subagents include a subagent that has newly learned through machine learning how to improve the index (e.g., task execution time, etc.) used in evaluating the subagent, a subagent connected to a highly capable device (e.g., a server or DB), a different subagent that can process the same task, a subagent previously generated by the agent generation means 12e, and a subagent (e.g., LMM) specialized in a specific function (e.g., language processing).

[0082] For example, in the previous case, a new subagent 8 is generated that executes only part of the task "(Task 3) Collect market data (market trends, sales trends of similar products, etc.)," ​​thereby shortening the execution time of Task 3.

[0083] As described above, when the agent evaluation means 12d has completed evaluation of a subagent or the agent generation means 12e has completed generation of a subagent, the agent allocation means 12b executes processing to update the allocation of the subagent based on the evaluation of the subagent.

[0084] For example, in the previous case, the agent allocation means 12b replaces the subagent 3 assigned to task 3 with the newly created subagent 8, thereby completing the process of updating the subagent allocation. By replacing subagent 3 with subagent 8, which shortens the execution time of task 3, it is possible to increase the processing speed of the entire system.

[0085] Furthermore, as described above, when the agent evaluation means 12d completes the process of updating the allocation of the subagent based on the evaluation of the subagent, the process execution means 12c executes the updated task again and outputs a solution (answer) to the specified task (problem).

[0086] For example, in the previous case, as shown in Figure 4(a), when the agent evaluation means 12d completes the process of updating the allocation of subagents 1 to 7 to the allocation of subagents 1, 2, 8, 4 to 7, the processing execution means 12c causes the updated subagents 1, 2, 8, 4 to 7 to execute tasks 1 to 7, and transmits the sales forecast results (solutions) to the external terminal 16 used by the user as the execution results (answers) of the specified task (issue).

[0087] According to this example, by allocating a given task to multiple tasks and the AI ​​agents that process those tasks, and by evaluating and reviewing the content of the AI ​​agents that process the tasks, the accuracy of executing the given task can be dramatically improved.

[0088] Note that updating (reviewing) the allocation of subagents is not limited to replacing subagents, but may also include, for example, changing the processing content of tasks to be executed by subagents, deleting tasks to be executed by subagents, adding new tasks to be executed by subagents, changing the order of tasks to be executed by subagents, changing the priority of tasks to be executed by subagents, or replacing all of the tasks to be executed by subagents.

[0089] Therefore, for example, in the previous case, as shown in Figure 4(b), the processing of (Task 3) can be divided into (Task 3-1) collecting market trends and (Task 3-2) collecting sales trends of similar products, and the original subagent 3 can be assigned to "(Task 3-1) collecting market trends" and the new subagent 8 can be assigned to "(Task 3-2) collecting sales trends of similar products."

[0090] Furthermore, for example, if the processing time for "(Task 4) collecting competing data" is long, then Task 4 and subagent 4 may be deleted. Furthermore, if the accuracy of the sales forecast for Product A is determined to be low as a result of evaluation based on knowledge acquired through machine learning or knowledge acquired from external terminal 16, then the accuracy of the sales forecast may be improved by generating a combination of tasks and subagents that collect additional data other than the data collected in Tasks 1 to 4. Furthermore, tasks that were executed in parallel may be executed sequentially, or conversely, tasks that were executed sequentially may be executed in parallel.

[0091] <Information Processing System / Summary> As explained above, the information processing system according to this embodiment (for example, the information processing system 10 shown in FIGS. 1 and 2) is an information processing system configured to have AI agents (for example, the AI ​​agent 12 shown in FIG. 1 and the system terminal 12 shown in FIG. 2), and the AI ​​agents are assigned to a plurality of tasks (for example, tasks 1 to 7 shown in FIG. 3(a)) constituting a predetermined task (for example, a task for realizing the subject of "perform sales forecast for product a"), respectively, by agent assignment means (for example, the agent assignment means shown in FIGS. 1 and 2) that assigns AI agents (hereinafter referred to as "subagents") corresponding to each of the plurality of tasks (for example, tasks 1 to 7 shown in FIG. 3(a)) (for example, subagents 1 to 7 shown in FIG. 3(b)). a processing execution means (e.g., processing execution means 12c shown in Figures 1 and 2) that executes the predetermined task by having the subagent execute the task; an agent evaluation means (e.g., agent evaluation means 12d shown in Figures 1 and 2) that evaluates the subagent based on the execution result of the predetermined task; and an agent generation means (e.g., agent generation means 12e shown in Figures 1 and 2) that generates a new subagent based on the evaluation of the subagent, wherein the agent allocation means updates the allocation of the subagent based on the evaluation of the subagent.

[0092] Furthermore, the information processing method according to this embodiment (for example, a method executed by the information processing system 10 shown in FIGS. 1 and 2) is an information processing method executed using an AI agent (for example, the AI ​​agent 12 shown in FIG. 1, the system terminal 12 shown in FIG. 2), and the AI ​​agent includes an agent allocation step (for example, a process executed by the agent allocation means 12b shown in FIGS. 1 and 2) of allocating AI agents (hereinafter referred to as "subagents") corresponding to a plurality of tasks (for example, tasks 1 to 7 shown in FIG. 3(a)) constituting a predetermined task (for example, a task for realizing the subject of "perform sales forecast for product a") (for example, tasks 1 to 7 shown in FIG. 3(b)); the agent evaluation step (for example, a process executed by the agent evaluation means 12d shown in FIGS. 1 and 2) of evaluating the subagent based on the execution result of the predetermined task; and the agent generation step (for example, a process executed by the agent generation means 12e shown in FIGS. 1 and 2) of generating a new subagent based on the evaluation of the subagent, wherein the agent allocation step updates the allocation of the subagent based on the evaluation of the subagent.

[0093] Furthermore, the program of the information processing system according to this embodiment (for example, the information processing system 10 shown in FIGS. 1 and 2) is a program of an information processing system configured to have AI agents (for example, the AI ​​agent 12 shown in FIG. 1, the system terminal 12 shown in FIG. 2), and the AI ​​agent is a computer, and the AI ​​agent is assigned to an agent allocation means (for example, the AI ​​agent allocation means shown in FIGS. 1 and 2) that assigns AI agents (hereinafter referred to as "subagents") corresponding to each of a plurality of tasks (for example, tasks 1 to 7 shown in FIG. 3(a)) that constitute a predetermined task (for example, a task for achieving the subject of "perform sales forecast for product a"). and a process execution means (e.g., process execution means 12c shown in FIGS. 1 and 2) that executes the predetermined task by having the subagent execute the task, an agent evaluation means (e.g., agent evaluation means 12d shown in FIGS. 1 and 2) that evaluates the subagent based on the execution result of the predetermined task, and an agent generation means (e.g., agent generation means 12e shown in FIGS. 1 and 2) that generates a new subagent based on the evaluation of the subagent, and the agent allocation means updates the allocation of the subagent based on the evaluation of the subagent.

[0094] Furthermore, the AI ​​agent according to this embodiment (for example, the AI ​​agent 12 shown in FIG. 1, the system terminal 12 shown in FIG. 2) is an AI agent that acts on behalf of a human being and has the ability to learn and make decisions on its own, and the AI ​​agent comprises a computer, an agent allocation means (for example, the agent allocation means 12b shown in FIGS. 1 and 2) that allocates AI agents (hereinafter referred to as "subagents") corresponding to a plurality of tasks (for example, tasks 1 to 7 shown in FIG. 3(a)) that constitute a predetermined task (for example, a task for achieving the objective of "perform sales forecast for product a") (for example, tasks 1 to 7 shown in FIG. 3(b)), The AI ​​agent functions as a processing execution means (e.g., processing execution means 12c shown in Figures 1 and 2) that executes the specified task by having the subagent execute the task, an agent evaluation means (e.g., agent evaluation means 12d shown in Figures 1 and 2) that evaluates the subagent based on the execution result of the specified task, and an agent generation means (e.g., agent generation means 12e shown in Figures 1 and 2) that generates a new subagent based on the evaluation of the subagent, and the agent allocation means updates the allocation of the subagent based on the evaluation of the subagent.

[0095] According to the information processing system, information processing method, information processing program, and AI agent of this embodiment, the accuracy of executing a specified task can be dramatically improved by allocating the specified task to multiple tasks and the AI ​​agents that process the tasks, and by evaluating and reviewing the content of the AI ​​agents that process the tasks.

[0096] Furthermore, the system may also include a task allocation means (for example, the task allocation means 12a shown in FIGS. 1 and 2) that allocates the predetermined task to the plurality of tasks.

[0097] With this configuration, agents can be smoothly assigned to tasks, and the speed at which a given task is executed can be increased.

[0098] In addition, the agent allocation means , a plurality of said sub None of the agents have reached the predetermined level of evaluation. Sub If there is an agent, Sub It may also be something that updates the contents of the agent.

[0099] With this configuration, it is possible to reuse highly rated agents while reviewing the content of low-rated agents, thereby improving the accuracy of task processing.

[0100] The agent allocation means also Sub The agent is not allowed to continue until all of the agent's ratings reach the predetermined level. Sub It may also be something that updates the contents of the agent.

[0101] With this configuration, all agents that execute a plurality of tasks that make up a specific task can be raised to a predetermined level, thereby improving the accuracy of the overall processing.

[0102] The system may also be configured to learn the correspondence between the predetermined task, the plurality of tasks, and the subagents through machine learning.

[0103] With this configuration, multiple tasks and subagents can be quickly assigned to a given task, thereby increasing the processing speed for a particular task.

[0104] It should be noted that the information processing system, information processing method, information processing program, and AI agent according to the present invention are not limited to the above-described embodiments, and various modifications can be made within the scope of the gist of the present invention.

[0105] Therefore, for example, the "predetermined task" according to the present invention is not limited to a prompt given from the external terminal 16, but may be, for example, a task generated based on an external event (e.g., a signal input from a sensor), or a task generated based on an internal event (e.g., activation by an internal scheduler, occurrence of an abnormality).

[0106] Furthermore, in the above embodiment, an example was shown in which the information processing system 10 includes the task allocation means 12a, but the task allocation means 12a is not an essential component, and the information processing system 10 does not necessarily have to include the task allocation means 12a. [Industrial Applicability]

[0107] The information processing system, information processing method, information processing program, and AI agent according to the present invention can be widely applied in industries such as manufacturing, service, and retail. [Explanation of symbols]

[0108] 10 Information Processing Systems 12 System Terminal 12a Task allocation method 12b Agent Allocation Method 12c Processing Execution Means 12d Agent Evaluation Methods 12e Agent Creation Method 16 External Terminal 21 CPU 22 ROM 23 RAM 24 Recording media 25 External Storage Drive 26 Storage device 27 Input Devices 28 Display device 29 Communications Department

Claims

1. An information processing system configured with an AI agent, The AI ​​agent: agent allocation means for allocating AI agents (hereinafter referred to as "subagents") corresponding to each of a plurality of tasks constituting a predetermined task; a process execution means for causing the subagent to execute the task, thereby executing the predetermined task; agent evaluation means for evaluating the subagent based on the execution result of the predetermined task; an agent generating means for generating a new subagent based on the evaluation of the subagent; the agent allocation means updates the allocation of the subagents based on the evaluation of the subagents; An information processing system comprising:

2. 2. The information processing system according to claim 1, the agent allocation means updates the content of a subagent when the evaluation of the subagent does not reach a predetermined level among the plurality of subagents; An information processing system comprising:

3. 3. The information processing system according to claim 2, the agent allocation means updates the contents of the subagents until the evaluations of all of the plurality of subagents reach the predetermined level; An information processing system comprising:

4. An information processing method performed using an AI agent, The AI ​​agent: an agent allocation step of allocating AI agents (hereinafter referred to as "subagents") corresponding to each of a plurality of tasks constituting a predetermined task; a processing execution step of causing the subagent to execute the task, thereby executing the predetermined task; an agent evaluation step of evaluating the subagent based on the execution result of the predetermined task; an agent generation step of generating a new subagent based on the evaluation of the subagent; the agent allocation step updates the allocation of the subagents based on the evaluation of the subagents; 1. An information processing method comprising:

5. 5. The information processing method according to claim 4, The agent allocation step updates the contents of the subagents until the evaluations of all of the subagents reach a predetermined level.

1. An information processing method comprising:

6. A program for an information processing system configured with an AI agent, The AI ​​agent is a computer. agent allocation means for allocating AI agents (hereinafter referred to as "subagents") corresponding to each of a plurality of tasks constituting a predetermined task; a process execution means for causing the subagent to execute the task, thereby executing the predetermined task; agent evaluation means for evaluating the subagent based on the execution result of the predetermined task; functioning as an agent generation means for generating a new subagent based on the evaluation of the subagent; the agent allocation means updates the allocation of the subagents based on the evaluation of the subagents; An information processing program characterized by:

7. 7. The information processing program according to claim 6, the agent allocation means updates the contents of the subagents until the evaluations of all of the plurality of subagents reach a predetermined level; An information processing program characterized by:

8. An AI agent that acts on behalf of a human and has the ability to learn and make decisions on its own, Computer, agent allocation means for allocating AI agents (hereinafter referred to as "subagents") corresponding to each of a plurality of tasks constituting a predetermined task; a process execution means for causing the subagent to execute the task, thereby executing the predetermined task; agent evaluation means for evaluating the subagent based on the execution result of the predetermined task; functioning as an agent generation means for generating a new subagent based on the evaluation of the subagent; the agent allocation means updates the allocation of the subagents based on the evaluation of the subagents; An AI agent characterized by:

9. 9. The AI ​​agent of claim 8, The system is configured to learn the correspondence between the predetermined task, the plurality of tasks, and the subagents by machine learning. An AI agent characterized by:

Images