Information processing device and program

Abstract

This technology provides a way to realize a multi-agent system that allows for more appropriate assignment of AI agents to each task. [Solution] Service provider 1 operates a sales site 1A for AI agent 1C and provides a platform 1B on which the purchased AI agent 1C can be used. Customer 2 specifies an AI agent 1C and requests a task. In response to the request, the specified AI agent 1C generates the tasks necessary to perform the task. An AI agent 1C is assigned to each generated task. The tasks, along with their execution order and the assignment results of AI agents 1C, are presented to customer 2 via a step confirmation screen 1D. On the confirmation screen 1D, the assignment of AI agents 1C to each task can be changed. Finally, each task is executed by the AI ​​agent 1C specified on the confirmation screen 1D.

Description

Technical Field

[0001] The present invention relates to an information processing apparatus and a program.

Background Art

[0002] With the remarkable progress of large language models (LLMs), AI (Artificial Intelligence) agents using LLMs have been put into practical use. This AI agent is a system or program that autonomously determines tasks to be executed for achieving a specific purpose and executes the determined tasks. For this purpose, in addition to autonomy, the AI agent has features such as acquiring information from the environment (perception), making decisions based on the acquired information (inference / planning), and taking appropriate actions (execution).

[0003] When enabling one AI agent to handle various tasks, the accuracy of each task is likely to decrease. The accuracy can be increased by specializing in a specific task. However, when specializing in a specific task, the generality becomes low, and the possibility of requiring expansion or modification according to the assumed work (business) is higher. This becomes an obstacle in the practical use of AI agents. For these reasons, when it is difficult to obtain the required accuracy with one AI agent, a multi-agent system may be adopted (see, for example, Non-Patent Document 1).

[0004] This multi-agent system is a system in which a plurality of AI agents cooperate or compete with each other to achieve a specific purpose. Each AI agent operates independently while exchanging information or making adjustments with other AI agents to perform a high-level task as a whole. In the multi-agent system, by executing a plurality of AI agents, high generality can be realized while enabling effective use of AI agents specialized in specific tasks.

Prior Art Documents

[0005] [Non-Patent Document 1] Masafumi Mitsuka, "LLM (Large-Scale Language Models) Seeing the Potential for Full-Scale Application in Business Operations - AI Agents and Multi-Agents," December 2, 2024, Fujisoft Inc. [Retrieved February 7, 2025], Internet<URL: https: / / www.fsi.co.jp / blog / 11929 / > [Overview of the project] [Problems that the invention aims to solve]

[0006] Even in a multi-agent system, autonomously determined tasks are executed according to autonomously determined procedures (processes). Each AI agent used in a multi-agent system possesses its own expertise. Therefore, each AI agent is autonomously assigned tasks to perform.

[0007] Even within the same specialized field, AI agents possess different characteristics depending on the data used for training. These characteristics can lead to variations in the results obtained by different AI agents, even within the same specialized field. In a multi-agent system, the execution results of a task by one AI agent may influence the execution results of tasks by other AI agents.

[0008] The present invention aims to provide a technology for realizing a multi-agent system that enables more appropriate assignment of AI agents to each task. [Means for solving the problem]

[0009] An information processing device according to one aspect of the present invention includes: an input processing unit that inputs work content information representing the content of the work to be performed; a selection unit that selects an artificial intelligence agent to be assigned to each task from among a plurality of artificial intelligence agents for each task to perform the work represented by the work content information; and an execution instruction unit that, according to the selection result by the selection unit, instructs the artificial intelligence agent to which the task has been assigned to execute the task. [Effects of the Invention]

[0010] The present invention makes it possible to realize a multi-agent system that can more appropriately assign AI agents to each task. [Brief explanation of the drawing]

[0011] [Figure 1] This figure illustrates an overview of an example of a platform service provided by an information processing device according to one embodiment of the present invention. [Figure 2] This figure illustrates an example configuration of an information processing system constructed using an information processing device according to one embodiment of the present invention, and an example of its application. [Figure 3] This figure shows an example of the hardware configuration of an AP server, which is an information processing device according to one embodiment of the present invention. [Figure 4] This figure shows an example of a functional configuration implemented on an AP server, which is an information processing device according to one embodiment of the present invention. [Figure 5] This flowchart shows an example of the processing flow executed by the employee terminal, as well as the CPU and GPU on the AP server, when a task is requested. [Modes for carrying out the invention]

[0012] Embodiments of the present invention will be described below with reference to the drawings. Figure 1 is a diagram illustrating an overview of an example of a platform service provided by an information processing device according to one embodiment of the present invention.

[0013] Service provider 1 is a company that provides platform services, or PaaS (Platform as a Service). In this embodiment, it provides platform 1B, which has an execution environment for artificial intelligence (AI) agents built on it. To make it easier to run any AI agent on platform 1B, service provider 1 operates a sales site 1A that sells AI agents 1C. AI agents 1C purchased by customer 2 from sales site 1A are added (copied) to platform 1B used by customer 2. Customer 2 is mainly an individual or a legal entity such as a company.

[0014] AI Agent 1C is a program that enables a processor such as a CPU (Central Processing Unit) or GPU (Graphics Processing Unit) on an information processing device to function as an AI agent. Therefore, unless otherwise specified, hereafter, AI Agent 1C will refer to the program for use as an AI agent. The program for use as an AI agent will henceforth be referred to as the "AI Application."

[0015] Note that Figure 1 shows that the system for sales site 1A and the system for providing platform 1B are built within service provider company 1, but this is for the sake of ease of understanding. These systems may be built outside of service provider company 1 using cloud services or the like. These systems may also be built separately. For these reasons, the location and configuration of these systems are not particularly limited.

[0016] Platform 1B is designed for multi-agent systems (multi-type AI agent systems). Even if AI agents 1C are specialized in the same field, they will have different characteristics depending on the data used for training. As a result, it is common for multiple AI agents 1C specializing in the same field to exist. In a multi-agent system, the execution result of a task by one AI agent 1C may affect the execution result of tasks by other AI agents 1C. Therefore, even if the same task is requested, the final result may differ significantly depending on which AI agent 1C performs the task.

[0017] In a multi-agent system, it is possible to autonomously assign an AI agent 1C to each task. However, this autonomous assignment may not be optimal for customer 2. There may be other AI agents 1C that would yield more desirable results for customer 2. For this reason, in this embodiment, while making it possible to easily customize the assignment of AI agents 1C to each task, customer 2 can obtain the AI ​​agent 1C that is more desirable for them through the sales site 1A.

[0018] By making it possible for Customer 2 to obtain a more desirable AI agent 1C, Customer 2 can assign a more desirable AI agent 1C to each task. This allows Customer 2 to more easily optimize, or at least make more appropriate, the assignment of AI agent 1C to each task for the work they request. Customer 2 can also easily compare the results for each assignment of AI agent 1C to each task. In this way, Customer 2 can more easily obtain optimal or near-optimal results.

[0019] Platform 1B assumes the existence of a plurality of AI agents 1C. Therefore, in this embodiment, a plurality of predetermined AI agents 1C are provided to customers 2 who use Platform 1B. Thereby, while enabling customers 2 to use the minimum necessary AI agents 1C, customers 2 can purchase the necessary AI agents 1C from sales site 1A.

[0020] Many of the AI agents 1C available to customer 2 are equipped with the function of autonomously decomposing the requested work into a plurality of tasks and determining the execution order of each task. Even if the same work is requested, the decomposition into tasks or the execution order of each task may differ depending on the AI agent 1C. Therefore, in this embodiment, customer 2 is allowed to select the AI agent 1C that requests the work. In FIG. 1, "business" is used in the sense of work. Therefore, hereinafter, unless otherwise specified, "business" is used in the same sense as work. Also, the AI agent 1C selected by customer 2 for the request of business is denoted as "business request agent 1C" for the purpose of distinguishing it from other AI agents 1C.

[0021] The business request agent 1C determines a plurality of tasks required for the execution of the requested business and the execution order of each task. As shown in FIG. 1, platform 1B generates a step confirmation screen 1D in which the task creation result obtained by such determination is arranged and presents it to customer 2.

[0022] The step confirmation screen 1D shown as an example in FIG. 1 is for the case where the creation of a proposal is requested as a business, and is presented to customer 2 as, for example, a pop-up screen. The strings 1D1 on screen 1D, "1. Requirements Confirmation," "2. Proposal Template Preparation (labeled "Proposal Template Preparation" in Figure 1)," "3. Content Creation," and "4. Overall Structure & Output," respectively, represent the content of each created task and the execution order of that task. The input boxes 1D2 located below each string 1D1 allow you to specify the AI ​​agent 1C to be assigned to the corresponding task. The "Partner AI" displayed in each input box 1D2 represents the name of the AI ​​agent 1C automatically assigned by platform 1B.

[0023] Each input box 1D2 has a menu button 1D21. This menu button 1D21 is for displaying the selectable AI agent 1C. Therefore, if customer 2 has a task for which they want to change the AI ​​agent 1C, they can, for example, click the menu button 1D21 corresponding to that task and select the desired AI agent 1C from the list that appears as a result of that click. The confirmation screen 1D also contains an "Add Step" button 1D3, a "Cancel" button 1D4, and an "Execute" button 1D5. Each of the buttons 1D3 to 1D5 is assigned the following functions:

[0024] The "Add Step" button 1D3 allows users to add or delete tasks to be executed. When button 1D3 is clicked, platform 1B generates and presents another screen (not shown) where tasks can be added, deleted, or the execution order of tasks can be changed. This allows customer 2 to add, delete, or change the execution order of tasks as needed on that screen. This screen will hereafter be referred to as the "edit screen".

[0025] Step confirmation screen 1D shows the breakdown of the requested work into tasks, the execution order of the tasks, and the assignment of AI agent 1C to each task. Each task is executed by the assigned AI agent 1C in the order shown on step confirmation screen 1D. Therefore, if customer 2 makes any edits on the editing screen, such as adding tasks, deleting tasks, or changing the execution order of tasks, the results of those edits will be reflected on step confirmation screen 1D.

[0026] The "Cancel" button 1D4 is used to instruct the system to cancel the execution of each task shown on the step confirmation screen 1D. The "Execute" button 1D5 is used to instruct the system to execute each task. If customer 2 wishes to add, delete, or change the execution order of tasks, they should click the "Add Step" button 1D3, perform the necessary editing on the editing screen, and then click the "Execute" button 1D5. The confirmation screen 1D is updated and redisplayed upon confirmation that the editing process is complete. Clicking the "Execute" button 1D5 causes platform 1B to have the AI ​​agent 1C assigned to each task execute the tasks shown on the confirmation screen 1D.

[0027] By allowing the addition, deletion, and modification of task execution order, Customer 2 can execute the tasks they deem necessary in the order they deem optimal. This makes it easier for Customer 2 to obtain more desirable results from their work requests. Each AI agent 1C executing a task utilizes web searches and services provided by provider 3 as needed. Provider 3, where AI agents are available, requests subtasks to perform the necessary processing. The tasks shown on the step confirmation screen 1D are tasks broken down for business execution and are distinct from the tasks (subtasks) autonomously created by each AI agent 1C assigned to that task.

[0028] If Customer 2 is allowed to arbitrarily select the AI ​​agent to assign to each task, Customer 2 can simply prepare the AI ​​agent 1C they deem necessary, much like selecting team members to address the problem, issue, or project. By assigning such AI agent 1C to the appropriate task, Customer 2 can more easily obtain more desirable results. If the tasks to be executed and their execution order can be changed, Customer 2 can execute the tasks in a more desirable order depending on the project. As a result, Customer 2 can obtain more desirable results. Thus, both changing the AI ​​agent assigned to each task and changing the order in which tasks are executed are customizations that enable Customer 2 to obtain better results. These customizations, in turn, enable more appropriate assignment of AI agents to each task. Note that purchasing AI agent 1C to address a project is similar to hiring a team member.

[0029] Figure 2 illustrates an example configuration of an information processing system constructed using an information processing device according to one embodiment of the present invention, and an example of its application. Note that the example configuration and application of the information processing system are merely examples and are not particularly limited.

[0030] The information processing system shown as an example in Figure 2 was built by service provider 1 for providing platform services. The information processing device in this embodiment is implemented as AP server 12. For this reason, the information processing device in this embodiment will also be denoted as "12" from now on. Although the information processing system is built within service provider 1, it may also be built using cloud services as described above.

[0031] As shown in Figure 2, the information processing system includes an AP server 12, a Web server 11, and a DB (Database) server 13, all connected to a network 14. Network 14 is, for example, a LAN (Local Area Network). The DB server 13 is used to manage the AI ​​applications to be sold. In addition to network 14, the Web server 11 is connected to an external network 20. This network 20 is, for example, a composite network including the internet.

[0032] This network 20 is connected to, or can be connected to, employee terminals 21 used by customer 2, and web servers 31 installed by each provider 3. Employee terminals 21 are information processing devices equipped with communication functions, and are used, for example, by employees working for a company (corporation) that is customer 2. Therefore, unless otherwise specified, "employee" will be used as another term for customer 2. Specifically, employee terminals 21 are, for example, PCs (Personal Computers), smartphones, or tablet devices.

[0033] An AP server 32 is connected to a web server 31 installed by provider 3. This AP server 32 is capable of running AI applications and functioning as an AI agent. Subtasks requested by service provider 1 are processed by the AP server 32.

[0034] Figure 3 shows an example of the hardware configuration of an AP server, which is an information processing device according to one embodiment of the present invention. This hardware configuration example is just one example and is not particularly limited. For example, only one CPU 121 and one GPU 124 are shown, but multiple CPUs and GPUs may be installed. A configuration without a GPU 124 is also possible.

[0035] As shown in Figure 3, the AP server 12 has a configuration in which the CPU 121, ROM (Read Only Memory) 122, RAM (Random Access Memory) 123, GPU 124, NIC (Network Interface Card) 125, auxiliary storage device 126, media drive 127, and I / FC (Interface Controller) group 128 are connected to the bus 129. The GPU 124 is connected to VRAM (video RAM) 124A.

[0036] NIC125 enables communication via network 14. Communication via network 14 may be either wireless or wired. It may also support multiple communication standards. Although Figure 3 shows only one NIC125, multiple NIC125s supporting different communication standards may be installed.

[0037] The auxiliary storage device 126 is a device capable of permanently storing data, such as a hard disk drive or an SSD (Solid State Drive). The media drive 127 is a device on which the recording medium 127A can be attached and detached. The media 127A is such as a CD (Compact Disc)-ROM, DVD-ROM, DVD-RAM, etc.

[0038] The I / FC group 128 includes various I / FCs that enable communication with various peripheral devices, including the input device 128A and the display device 128B, or with external devices. The input device 128A and the display device 128B are temporarily connected to the I / FC group 128 as needed. The auxiliary storage device 126 stores the OS (Operating System) and various application programs that run on that OS as programs. These various application programs include various AI applications, applications for sales site 1A, and applications for platform 1B. Hereafter, applications for sales site 1A and applications for platform 1B will be referred to as "sales applications" and "platform applications," respectively.

[0039] ROM122 is also a device capable of permanently storing data, such as firmware and various other data. The CPU121 reads the firmware stored in ROM122 into RAM123 and executes it. Subsequently, the firmware reads the OS stored in auxiliary storage device 126 into RAM123 and executes it. Sales applications and platform applications are read into RAM123 by the OS and executed. The GPU124 can execute various AI applications stored in auxiliary storage device 126 and read into VRAM124A.

[0040] The platform application may be stored on media 127A and distributed. If network 20 is a complex network including the internet, it may also be distributed via network 20. When distributed via network 20, the platform application should be stored on a recording medium that can be directly or indirectly accessed by the information processing device distributing it. In other words, the storage medium may be directly or indirectly accessible by another information processing device that can communicate with the information processing device distributing it.

[0041] Figure 4 shows an example of a functional configuration implemented on an AP server, which is an information processing device according to one embodiment of the present invention. The example of a functional configuration on CPU 121 is mainly implemented by CPU 121 executing sales applications and platform applications. The example of a functional configuration on GPU 124 is implemented by executing the corresponding AI application. Note that the functional configuration is not particularly limited, and various modifications are possible.

[0042] As shown in Figure 4, the CPU 121 of the AP server 12 is functionally configured to include a transmit / receive processing unit 1211, a sales processing unit 1212, and a platform processing unit 1213. The platform processing unit 1213 is for providing platform 1B to each customer, for example. To this end, a virtual machine is built on the AP server 12 for each customer, and the platform processing unit 1213 is implemented in each virtual machine.

[0043] In Figure 4, only one platform processing unit 1213 is shown, assuming only one customer 2. Accordingly, the example of the functional configuration on the GPU 124 and the example of the data storage area allocated on the auxiliary storage device 126 are also assumed to be for only one customer 2. The first agent processing unit 1241, the second agent processing unit 1242, the third agent processing unit 1243, and the fourth agent processing unit 1244, all implemented on the GPU 124, are components realized by the GPU 124 executing different AI applications, and each functions as an AI agent 1C with different characteristics. For this reason, when it is not necessary to distinguish between each agent processing unit 1241 to 1244, or when they are collectively referred to as "AI agent 1C," the type and number of agent processing units implemented on the GPU 124 will vary depending on the AI ​​agent 1C (AI application) that the customer purchases from the sales site 1A.

[0044] While this functional configuration is realized on the CPU 121 and the GPU 124, the auxiliary storage device 126 is reserved as a data storage area for the agent management information storage unit 1261, the provider management information storage unit 1262, and the AI ​​application storage unit 1263.

[0045] The information stored in each of the memory units 1261 to 1263 is, for example, as follows: The agent management information stored in the agent management information storage unit 1261 is for managing the AI ​​agent 1C implemented by the AI ​​application. This management information includes various types of information such as the application ID (IDentification) assigned to the AI ​​application, the URL (Uniform Resource Locator) indicating the storage location of the AI ​​application, the scope of support, operational characteristics, execution flag, task completion rate, evaluation, and validity period. The AI ​​application is stored in the AI ​​application storage unit 1263. Therefore, the URL in the agent management information indicates the storage location within the AI ​​application storage unit 1263.

[0046] The "Scope of Capabilities" section in the agent management information indicates, for example, the types of tasks the agent can handle. "Operational Characteristics" includes information representing various features such as learning ability, accuracy, available time, ability to handle multiple tasks, and average response time. The "Running" flag indicates whether there are tasks currently running. The "Task Completion Rate" indicates the progress of currently running tasks. The "Evaluation" represents the employee's overall assessment of the task's execution results. The "Validity Period" indicates the period during which the AI ​​application can be used. This validity period allows AI agent 1C to be treated as a limited-time rental product. A rental product is like a member dispatched for a limited time.

[0047] The types and combinations of information that constitute the agent management information are not particularly limited. This agent management information only needs to enable the appropriate selection of AI agent 1C to be assigned to each task. The provider management information storage unit 1262 stores provider management information for each provider 3. This provider management information is about provider 3 that is recommended for access or should be the recipient of a specific subtask. The main provider 3s are those that make AI agents equipped with LLM or generative AI available. By providing such provider management information, each AI agent 1C can request a subtask from the appropriate provider 3.

[0048] Information stored in either memory unit 1261 or 1262 is actually read into RAM 123 and processed by CPU 121. Each AI application stored in AI application memory unit 1263 is read into VRAM 124A and executed by GPU 124. CPU 121 communicates with Web server 11 via NIC 125. Data input and output between CPU 121 and GPU 124 is performed via RAM 123. These are ignored in Figure 4 for convenience. This will also be the case in subsequent explanations.

[0049] The parts 1211 to 1213 implemented on the CPU 121 have, for example, the following functions. The transmission / reception processing unit 1211 performs processing for sending and receiving various data, including requests, with the Web server 11. Receiving work requests from employee terminals 21, displaying the results of the requested work execution, and displaying various screens, including the step confirmation screen 1D, to the employee terminals 21 are all done via the Web server 11. This transmission / reception processing unit 1211 corresponds to the input unit in this embodiment.

[0050] As described above, the sales processing unit 1212 enables the information processing system to function as a sales site 1A. To this end, the sales processing unit 1212 includes a presentation processing unit 1221, a payment processing unit 1222, and an additional processing unit 1223, as shown in Figure 1. The presentation processing unit 1221 enables customer 2 to provide various information, including the presentation of purchasable AI agents 1C. The payment processing unit 1222 enables payment associated with the purchase of the AI ​​agent. The addition processing unit 1223 adds the AI ​​agent 1C purchased by customer 2 to the platform 1B used by customer 2. This addition includes, for example, storing the AI ​​application of the corresponding AI agent 1C in the AI ​​application storage unit 1263, and generating agent management information and storing it in the agent management information storage unit 1261.

[0051] The platform processing unit 1213 enables the provision of an available platform 1B to customer 2. To this end, as shown in Figure 1, the platform processing unit 1213 includes a business content information processing unit 1231, a task assignment unit 1232, a task execution instruction unit 1233, and an external request processing unit 1234.

[0052] While a detailed explanation is omitted, Customer 2 selects AI Agent 1C and inputs information deemed necessary for the task request as task content information, and then requests the task. Upon receiving such a request, a request containing the task content information and the selection result of AI Agent 1C is sent from employee terminal 21 to web server 11. This request is then sent, for example, to AP server 12 via web server 11, and passed to task content information processing unit 1231 by transmission / reception processing unit 1211. The task content information processing unit 1231 outputs the task content information directly to the task request agent 1C, which is the AI ​​Agent 1C indicated by the selection result, and instructs it to process the information. The instructed task request agent 1C then determines the tasks necessary to execute the task detailed in the task content information, as well as their execution order. The results of this processing are output to task assignment unit 1232.

[0053] The task assignment unit 1232 refers to agent management information and selects an AI agent 1C to be assigned to each task. This selection may be performed, for example, by identifying the type of task for each task, identifying AI agents 1C that are within the range of capabilities for the identified task type, and extracting the AI ​​agent 1C with the highest evaluation among the identified AI agents 1C. After selecting an AI agent 1C for each task in this manner, the task assignment unit 1232 passes the selection result to the transmission / reception processing unit 1211 and instructs it to send the step confirmation screen 1D. In response to such an instruction, the transmission / reception processing unit 1211 sends a request to the Web server 11 to display the step confirmation screen 1D, which includes the selection result. As a result, the step confirmation screen 1D is sent from the Web server 11 to the employee terminal 21 that sent the work request.

[0054] Customer 2 can purchase AI Agent 1C as they deem necessary. As they purchase AI Agent 1C, the AI ​​Agent 1C that is best assigned to each task also changes. Therefore, even if customers are allowed to select which AI Agent 1C to assign to each task from a set of options, the assignment of AI Agent 1C to each task can be made more appropriate.

[0055] As described above, the step confirmation screen 1D is a screen that allows the addition of tasks, deletion of tasks, and editing of the task execution order. The contents of the step confirmation screen 1D are changed according to the editing results. The task assignment unit 1232 responds to updating the step confirmation screen 1D according to the editing results. When the "Execute" button 1D5 on the step confirmation screen 1D is clicked, a task execution request is sent from the employee terminal 21 to the web server 11. This execution request stores task assignment information representing each task shown on the step confirmation screen 1D, the execution order of each task, and the assignment result of the AI ​​agent 1C to each task.

[0056] This execution request is sent, for example, to the AP server 12 via the Web server 11, and is passed to the task execution instruction unit 1233 by the transmission / reception processing unit 1211. The task execution instruction unit 1233 instructs the AI ​​agent 1C that should execute the task to execute the task, according to the task assignment information in the execution request.

[0057] To instruct AI agent 1C to perform such tasks, the task execution instruction unit 1233 receives the task execution results of each AI agent 1C from the GPU 124. Therefore, if the task execution result received from the GPU 124 is that of the last task, the task execution instruction unit 1233 performs processing to display the execution results of the requested task on the employee terminal 21. This display of execution results is performed, for example, by instructing the transmission / reception processing unit 1211 to display the execution results of each task. In response to this instruction, the transmission / reception processing unit 1211 sends a display request, including the execution results of each task, to the web server 11. In response to this display request, the web server 11 generates a screen to present the execution results and sends the generated screen to the employee terminal 21.

[0058] Employee evaluations of the AI ​​agent 1C that performed each task may be made on the screen. This evaluation will influence the selection of the AI ​​agent 1C to be assigned to the task. For this reason, the evaluation content entered by the employee on the screen may be processed by the task assignment unit 1232, which will then update the agent management information of the corresponding AI agent 1C. The submission of the evaluation content may be done at the employee's instruction.

[0059] AI agent 1C, instructed to execute a task, creates and executes tasks (subtasks) as needed to carry out the instructed task. In doing so, AI agent 1C may perform web searches or request subtasks from AI agents available on provider 3. The external request processing unit 1234 handles such external requests. In this embodiment, the business content information processing unit 1231 corresponds to the processing instruction unit. The task assignment unit 1232 corresponds to the selection unit. The task execution instruction unit 1233 corresponds to the execution instruction unit.

[0060] Figure 5 is a flowchart illustrating an example of the processing flow executed by the employee terminal, the CPU, and the GPU on the AP server when a task is requested. Next, referring to Figure 5, we will further explain the processing executed by the employee terminal 21, the CPU 121, and the GPU 124 on the AP server 12 when a task is requested.

[0061] When an employee requests a task, they select AI agent 1C, input the task details, and issue a request. The employee terminal 21 responds to the employee's work and, upon receiving the request, sends a request to the web server 11 containing the selection result of AI agent 1C and the task details (step SA1).

[0062] This request is sent to the AP server 12 via the Web server 1 and input to the CPU 121. The CPU 121 processes this request and outputs business content information to the AI ​​agent 1C indicated by the selection result, instructing it to generate a task including the execution order (step SB1). The example in Figure 5 shows the case where the business request agent 1C is instructed to the first agent processing unit 1241.

[0063] The first agent processing unit 1241, in response to the instructions, breaks down the business content information into tasks, determines the execution order of the tasks, and outputs them to the CPU 121 (step SC1). The CPU 121 receives these processing results, refers to the agent management information, selects an AI agent 1C to which each task should be assigned, and sends the selection result to the employee terminal 21.

[0064] The tasks and the assignment results of AI agents 1C to each task are sent from the web server 11 as a step confirmation screen 1D, as described above. After receiving the step confirmation screen 1D, the employee terminal 21 displays the confirmation screen 1D as a pop-up (step SA2). Subsequently, if the employee clicks the "Add Step" button 1D3 on the displayed confirmation screen 1D, the employee terminal 21 performs editing processing to add tasks, delete tasks, or change the execution order of tasks, in accordance with the employee's operation (step SA3). After performing the editing processing, if the employee clicks the "Execute" button 15D on the step confirmation screen 1D, the employee terminal 21 sends a task execution request to the web server 11 containing information representing the contents of the step confirmation screen 1D (step SA4). This information represents each task, the execution order of each task, and the AI ​​agent 1C assigned to each task, as described above.

[0065] This execution request is sent to the AP server 12 via the Web server 11 and input to the CPU 121. The CPU 121 instructs the CPU to execute the task according to the information in this execution request (step SB3). The example in Figure 5 shows the case where the second agent processing unit 1242 is first instructed to execute the task.

[0066] Subsequently, CPU 121 instructs the assigned AI agent 1C to execute the remaining tasks according to the information in the execution request. After inputting the execution results of the last task, CPU 121 sends a display request containing the execution results of each task to Web server 11 (step SB4). As a result, the web server 11 generates a screen showing the execution result of each task and sends it to the employee terminal 21. After receiving this screen, the employee terminal 21 displays it (step SA5).

[0067] In this embodiment, the AI ​​agent 1C selected by customer 2 (user) is made to generate tasks for executing the work, but task generation may be omitted. This is because customer 2 may have knowledge of the tasks and their execution order that they consider optimal, depending on the content of the work. Assuming such a customer 2, multiple flows representing tasks and their execution order may be prepared, and the customer may be made to select the one they consider optimal from among the prepared flows. Customization of the prepared flows may also be made possible. If multiple such flows are prepared, customer 2, who has the necessary knowledge, will not have to perform editing work to add, delete, or change the execution order of tasks, or if they do perform such editing work, the amount of work will be reduced.

[0068] Furthermore, while this embodiment allows customization on a task-by-task basis, it is also possible to allow customization of tasks (subtasks) that each AI agent 1C performs autonomously. In other words, it is possible to allow the addition of subtasks, the deletion of subtasks, or the change in the execution order of subtasks.

[0069] In this embodiment, the platform 1B provided as a service enables customization on a task-by-task basis, but the platform 1B may also be made available on an information processing device specified by customer 2. This can be achieved by preparing a platform application that anticipates the information processing device that customer 2 is expected to specify, and having the customer install the prepared application.

[0070] In this embodiment, the platform application is installed on AP server 12, but the installation destination is not limited to a fixedly installed information processing device. Installation may be performed on an information processing device mounted on, for example, an autonomous vehicle or a humanoid robot. In that case, for example, an AI agent 1C that is considered optimal for the assumed environment may be selected and made to perform each task. For these reasons, the application field of this embodiment, the type (characteristics) of AI agent 1C, and the method of using AI agent 1C are not particularly limited. [Explanation of symbols]

[0071] 1 Service provider, 1A Sales site, 1B Platform, 1C AI agent, 1D Step confirmation screen, 2 Customer, 3 Provider, 1211 Send / receive processing unit, 1212 Sales processing unit, 1213 Platform processing unit, 1231 Business content information processing unit, 1232 Task assignment unit, 1233 Task execution instruction unit, 1234 External request processing unit.

Claims

1. An input processing unit that inputs work content information representing the content of the work to be performed, A selection unit selects an artificial intelligence agent from among multiple artificial intelligence agents to be assigned to each task for performing the work described in the aforementioned work content information, According to the selection results by the selection unit, for each task, an execution instruction unit instructs the artificial intelligence agent to which the task is assigned to execute the task, An information processing device equipped with the following features.

2. The input processing unit can receive, along with the work content information, selection information representing an artificial intelligence agent selected from among multiple artificial intelligence agents. The input processing unit, upon receiving the selection information, further comprises a processing instruction unit that instructs the artificial intelligence agent represented by the selection information to process the work content information and determine the tasks necessary for executing the work. The selection unit selects, for each task determined by the artificial intelligence agent represented by the selection information, an artificial intelligence agent to be assigned to that task from among multiple artificial intelligence agents. The information processing apparatus according to claim 1.

3. The selection results of the artificial intelligence agent for each task by the selection unit are presented to the user who has entered the work content information in a way that allows them to change the selection. If the user changes the selection result, the execution instruction unit instructs the artificial intelligence agent assigned to the task to execute the task according to the changed selection result. The information processing apparatus according to claim 1 or 2.

4. The tasks determined by the artificial intelligence agent, represented by the aforementioned selection information, are presented to the user who has inputted the work content information, allowing them to choose at least one of the following: add, delete, or change the execution order. The execution instruction unit, when the user performs at least one of the following actions regarding the addition, deletion, and modification of the execution order of the tasks, instructs the artificial intelligence agent assigned to the task to execute the tasks that exist after the modification. The information processing apparatus according to claim 2.

5. In an information processing device, The user is asked to input work content information that describes the tasks to be performed. For each task to perform the work described in the aforementioned work content information, an artificial intelligence agent to be assigned to the task is selected from among multiple artificial intelligence agents. In accordance with the results of the above selection, for each task, the artificial intelligence agent to which the task is assigned is instructed to perform the task. A program that executes a process.

Images