Information processing apparatus, information processing method, and program

Abstract

Provided are an information processing apparatus, an information processing method, and a program that enable task design using a chatbot application via natural language and a generative AI that collaborates with storage. **Solution**: An information processing apparatus that performs task design using a chatbot application via natural language and a generative AI, the apparatus including: a storage provided outside or inside the generative AI to collaborate with the generative AI for the task design; and a predetermined task management system in which task information designed by the task design is stored.

Description

Technical Field

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

Background Art

[0002] Before the emergence of generative AI, in the performance of business operations, there have been management issues such as improvement of QCD (Quality, Cost, Delivery) at the organizational level, activation of communication, or visualization of work, and there have been task management or project management methods, as well as information systems and tools to support them (for example, Patent Document 1).

[0003] On the other hand, since ChatGPT, a generative AI service, emerged in 2022 as generative AI, attempts have been made to automatically execute individual specific business operations themselves by generative AI and improve their efficiency (for example, Patent Document 2).

[0004] In the utilization of generative AI, various contracts and various generative AI models have been individually introduced into the organization.

[0005] In order to solve such a conventional complicated situation in a simplified manner, a prior application (Japanese Patent Application No. 2024-210812) has been filed by the applicant of the present application.

Prior Art Documents

Patent Documents

[0006]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0007] However, even in the task management and execution related to the applicant's previous application of this case, further improvement can be considered. That is, not only the execution of tasks but also the design of tasks should be made possible.

[0008] An object of the present invention is to provide an information processing apparatus, an information processing method, and a program that enable task design using a chatbot application via natural language and a generative AI that collaborates with storage.

Means for Solving the Problems

[0009] To achieve the above object, the information processing apparatus according to the present invention A first generative AI for designing a new series of tasks, a second generative AI for executing the designed tasks, a chatbot application for a user to have conversations with the first generative AI and the second generative AI via natural language, storage provided externally or internally to the first generative AI as information storage means, which is a storage location for knowledge information that the first generative AI learns to design the new series of tasks, and a storage location for design information as intermediate or final results handled by the first generative AI, and storage provided externally or internally to the second generative AI as information storage means, which is a storage location for knowledge information that the second generative AI learns to execute the designed tasks, and a storage location for input / output data handled by the second generative AI, a task management system where the final result by the first generative AI is stored, and an adapter functional unit for operating the task management system from the chatbot application, and the first generative AI has pre-learned general common sense knowledge, business / operation knowledge, task design information format knowledge, and adapter function operation knowledge to design the new series of tasks not pre-stored in the task management system, and the first generative AI can reflect the design information as the final result stored in the storage to the task management system via the adapter functional unit, and the second generative AI has pre-learned input / output knowledge, general common sense knowledge, business-specific knowledge, and adapter function operation knowledge to infer the position of input information as information required for the task executed in the task management system and the position of output information as information obtained as the execution result of the task, and the second generative AI is provided separately from the first generative AI or is also served by the first generative AI.

[0012] Also, preferably, the A new series of task of design is made based on the latest situation in the task management system during operation. the

[0013] Also, The information processing method according to the present invention a first generation AI for designing a new series of tasks, a second generation AI for executing the designed tasks, a chatbot application for allowing a user to converse with the first generation AI and the second generation AI via natural language, a storage provided as information storage means either inside or outside the first generation AI, and used as a storage location for knowledge information learned by the first generation AI for designing the new series of tasks and a storage location for design information as an intermediate or final result handled by the first generation AI, and a storage provided as information storage means either inside or outside the second generation AI, and used as a storage location for knowledge information learned by the second generation AI for executing the designed tasks and a storage location for input / output data handled by the second generation AI, a task management system in which the final results by the first generation AI are stored, and an application for operating the task management system from the chatbot application. and an adapter function unit, wherein the second generation AI is provided separately from the first generation AI or serves as the first generation AI as well, the information processing method in an information processing device comprising the steps of: the first generation AI learning common sense knowledge, business / operation knowledge, task design information format knowledge, and adapter function operation knowledge in advance in order to design the new series of tasks not pre-stored in the task management system; the first generation AI reflecting the design information as the final result stored in the storage in the task management system via the adapter function unit; and the second generation AI learning input / output knowledge, common sense knowledge, business-specific knowledge, and adapter function operation knowledge in advance in order to infer the location of input information as information required for a task executed in the task management system, and the location of output information as information obtained as a result of executing the task.

[0015] Also, the program according to the present invention causes a computer to execute each of the above steps.

Effects of the Invention

[0016] According to the present invention, it is possible to enable task design using a chatbot application via natural language and a generative AI that collaborates with storage.

Brief Description of the Drawings

[0017] [Figure 1] It is a diagram showing the generation cycle of a task design and execution system in an information processing apparatus using a generative AI according to an embodiment of the present invention. [Diagram 2] ​(a) is a diagram for explaining the GenTASK device that performs task design, and (b) is a diagram for explaining the DoTASK device that performs task execution and can be combined with the GenTASK device. [Diagram 3] (a) is a diagram showing the GenTASK device that performs task design, and (b) is a diagram for explaining the new DoTASK device that performs task execution and can be combined with the GenTASK device. [Figure 4] It is a comparison diagram of the DoTASK device, the GenTASK device, the GenTASK device + DoTASK device, and the GenTASK device + new DoTASK device. [Diagram 5] It is a diagram showing the actions between the components of the DoTASK device, the new DoTASK device, and the GenTASK device, and also showing the actions between the components in sequential update and batch update. [Figure 6] (a) is an explanatory diagram of a system equipped with an AVCM as a verification means for verifying the validity of the output of the generative AI for the input to the chatbot application before output, and (b) is an explanatory diagram regarding the arrangement of the AVCM. [Figure 7] It is a diagram showing that information terminals such as a PC and a smartphone are connected to a server, a generative AI, a storage, and a predetermined task management system via a network. [Figure 8] It is a diagram showing the types of knowledge and learning methods that the generative AI learns. [Figure 9] It is a diagram showing an example of a large-scale cooking plan for switching the generative AI in the task design process. [Figure 10] It is a diagram regarding a modified example in which the DoTASK device enables task design and task execution of the designed task.

Embodiments for Carrying Out the Invention

[0018] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[0019] (First Embodiment) (Definition) Business processes can be defined in various ways, but one clear way of definition is to consider them as a group of tasks with a certain purpose. Individual tasks can be represented as activities. And a business process can be considered as a collection of those tasks and their relationships. Here, depending on the differences in industries and systems, tasks may be called activities or tickets.

[0020] Also, task design means designing the above-mentioned collection of tasks and their relationships so that the objectives of specific operations, businesses, or projects can be achieved. Task design may sometimes include designing the resources required for task completion and the collection of relationships between tasks and resources. The information resulting from these design operations is called task information.

[0021] This task information is usually input into a task management system or the like and used to give work instructions. Furthermore, actual values resulting from the execution of tasks are input to measure the progress of operations, and when problems such as delays occur, actions for improvement are taken.

[0022] And generally in a task management system, the following attributes are given to tasks.

[0023] · ID (unique number) · Name · Description · Creator · Assignee · Scheduled start date · Scheduled end date · Actual start date · Actual end date · Scheduled man-hours · Actual man-hours Here, in this specification, task design includes task update (including at least partial addition, change, and deletion as changes to task content).

[0024] Situations where task design is carried out include newly generating, modifying for improvement, modifying existing task information and then reusing it for other projects, etc., and converting task information for a different task execution system.

[0025] Such task design has conventional problems such as high costs for task plan input and task plan update, task personalization, inhibition of standardization, and lack of real-time nature of task plans. There is great significance in using generative AI for task design.

[0026] Also, in this specification, task execution is distinguished from task design and does not include task design.

[0027] (Information processing apparatus) The information processing apparatus according to this embodiment basically includes any of the following cases: when it is only a GenTASK apparatus that designs tasks as shown in FIG. 4, when it is a combination of a GenTASK apparatus that designs tasks and a DoTASK apparatus that executes tasks (excluding the execution of the designed tasks), and when it is a combination of a GenTASK apparatus that designs tasks and a new DoTASK apparatus that also executes the designed tasks. Note that the DoTASK apparatus shown in FIGS. 4 and 2(b) is related to the applicant's previous application (Japanese Patent Application No. 2024-210812).

[0028] In Japanese Patent Application No. 2024-210812, the execution of tasks using generative AI was regarded as including the generation of tasks. However, in this specification, the execution of tasks using generative AI also includes operations that are prerequisites for task execution such as the generation, deletion, and further search of tasks themselves. And the task design using generative AI and storage in this specification is distinguished from the above task execution. Here, FIG. 1 shows the generation cycle of a task design / execution system in an information processing apparatus using generative AI according to an embodiment of the present invention.

[0029] Symbol a indicates the input action of a user speaking through natural language to a chatbot application (hereinafter abbreviated as chatbot or also denoted as CBA) to greet or request task execution. Symbol b indicates the action of the chatbot issuing necessary instructions such as task execution to the generative AI. Symbol c indicates the output action from the generative AI, and symbol d indicates the response action to the user.

[0030] And in FIG. 1, symbol e indicates the action of output data generated by the generative AI being stored (saved) in a predetermined task management system in the backyard. Here, as the predetermined task management system, open-source software such as REDMINE can be used.

[0031] And symbol f indicates the action of the generative AI obtaining input data from a predetermined task management system (note that the timing at which the generative AI obtains the input data is, of course, earlier than the timing at which the generative AI stores the output data based on the input data).

[0032] Here, in FIG. 1, between the user who wants to use the generative AI and the generative AI, when the conversation continues for a certain period, the actions of symbols a to d (abcd) repeat. And between the generative AI and a predetermined task management system, when the input and output between the generative AI and the task management system continue for a certain period, the actions of symbols bcef repeat.

[0033] (Knowledge learned by the generative AI) Now, in the present embodiment, as shown in FIG. 8, the generative AI learns general common sense knowledge, business / operation knowledge, task design information format knowledge, and adapter function operation knowledge for task design, and also learns, in addition to general common sense knowledge, business-specific knowledge as α, input / output knowledge as β, and adapter function operation knowledge as γ for task execution.

[0034] As learning methods, in addition to the case of loading AI standard knowledge, there are cases such as fine-tuning and extended prompt methods that learn from data in storage, such as RAG (Retrieval-Augmented Generation).

[0035] As knowledge added to general common sense knowledge, the first knowledge α is knowledge for correctly executing generation using generative AI. Also, the second knowledge β is knowledge for determining the input position and output position on the task management system for task execution, that is, the task to be input or output and its attributes. Further, the third knowledge γ is operational knowledge for acquiring data from the input position on a predetermined task management system and storing the data at the output position on the task management system.

[0036] Note that as an example of the learning method using the extended prompt method, it is possible to use the Assistants API of OpenAI and learn by the tools provided by the API.

[0037] (Task Design by Collaboration between Generative AI and Storage) Now, when the information processing apparatus according to the present embodiment includes the GenTASK apparatus shown in FIG. 2(a) or FIG. 3(a), as indicated by the reference symbol g in FIG. 1, in the backyard, the generative AI collaborates with the storage (which may be inside the generative AI, but in FIG. 1, an external storage provided outside the generative AI is shown) to perform task design.

[0038] Then, the newly task-designed task information is stored (memorized) in a predetermined task management system according to the flows of the reference symbols h, b, c, e shown in FIG. 1. In FIG. 2(a) or FIG. 3(a), the connection notation is shown such that the task design information (final result) is input to the task management system.

[0039] The storage may be inside the generative AI. In the GenTASK device, there are a generative AI, a chatbot application via natural language, a storage provided either outside or inside the generative AI to cooperate with the generative AI for task design, and a predetermined task management system where the task information designed with tasks is stored.

[0040] Note that the external storage provided outside the generative AI mainly refers to storage services on the cloud such as Google Drive, etc., but includes the case of using a task or ticket in task management operable via an adapter function as a temporary data storage location. Here, the GenTASK device shown in FIG. 2(a) or FIG. 3(a) will be described in more detail. In order for the generative AI and the storage to cooperate to perform task design, general common sense knowledge, business and operation knowledge, task design information format knowledge, and adapter function operation knowledge are learned.

[0041] In FIG. 2(a) or FIG. 3(a), task design can be performed based on the latest situation in a predetermined task management system during operation.

[0042] The storage can be connected to other GenTASK devices, DoTASK devices, and an external task management system (external TMS) via a transfer program. And the storage can be connected to a task management system (TMS) via a batch update program.

[0043] Note that the adapter function unit enables operating a predetermined task management system from a chatbot (performing at least input and output via a network), and it may be provided outside each of the chatbot and the predetermined task management system, or may be provided inside one of the chatbot and the predetermined task management system. Also, it may be provided inside both the chatbot and the predetermined task management system.

[0044] Here, regarding the tasks designed by the GenTASK device shown in Fig. 2(a) or Fig. 3(a), task execution can be performed by carrying out input / output operations manually, using only the task management system that stores the task information of the designed tasks.

[0045] However, input / output operations carried out manually are cumbersome, and it is desirable to combine with the new DoTASK device shown in Fig. 3(b) where the input / output positions can be inferred using generative AI.

[0046] Note that the search shown in Fig. 2(a) or Fig. 3(a) indicates the data flow of returning the results retrieved by the task management system in response to a request from the chatbot to the chatbot.

[0047] Also, the update shown in Fig. 2(a) or Fig. 3(a) includes at least partial addition, modification, and deletion as changes to the task content, as described in the initial definition.

[0048] Here, when an information processing device is formed by combining the GenTASK device in Fig. 2(a) and the DoTASK device in Fig. 2(b), it does not include the execution of the tasks designed as shown in Fig. 4 (only the execution of the existing stored tasks is possible).

[0049] In this case, regarding the tasks designed by the GenTASK device, task execution can be performed by carrying out input / output operations manually, using only the task management system that stores the task information of the designed tasks. Note that at the time of task execution, the conversational processing for task execution can be automated by developing a special plugin on the task management system side.

[0050] That is, conversational processing using a chatbot is flexible and can be used for various applications, so there is an advantage in cost reduction. On the other hand, for routine processing, the same conversation is repeated and forced on the user, which is not efficient.

[0051] In such a case, in order to automate without forcing the user to repeat the same conversation, the task management system can develop a special function through a mechanism such as a plug-in to streamline specific processes. However, since dedicated plug-ins need to be developed for different processes, the cost will be high.

[0052] In order to enable both the design of tasks and the execution of the designed tasks using generative AI, it is necessary to combine the GenTASK device shown in Fig. 3(a) and the new DoTASK device shown in Fig. 3(b). The GenTASK device shown in Fig. 3(a) is the same as that described in Fig. 2(a).

[0053] In the new DoTASK device shown in Fig. 3(b), the generative AI collaborates with the storage and learns, in addition to general common sense knowledge, business-specific knowledge, input / output knowledge, and adapter function operation knowledge. If there is storage, by referring to it as the input and the original output destination from the task management system, it becomes easier to handle large-scale data and the like.

[0054] In the new DoTASK device shown in Fig. 3(b), the generative AI and the storage collaborate so that the task execution of the designed task is performed by the task management system, and predetermined knowledge is learned so that the input position where the input data of the generative AI is input and the output position where the output data of the generative AI is stored can be inferred in a predetermined task management system.

[0055] (Input to the task management system as the task execution system for the task information of the designed task) Here, the task information of the designed task must ultimately be input (saved) to the task management system as the task execution system in order to execute that task. In Fig. 2(a) or Fig. 3(a), the connection notation indicates that the task design information (final result) is input to the task management system.

[0056] Here, there are the following two methods for submitting a task to the task execution system. In either case, not only new task information is input, but there are also cases where the input task information is updated or deleted. 1) Sequential method The sequential method is a method of updating task information while appropriately referring to the design information (final result) during the conversation between the user and the AI. In the case of small-scale input, the sequential method is convenient. In the sequential method, not only the design information (final result), but also new task information may be input or the input task information may be updated based on various information. 2) Batch method The batch method is a method of creating a file listing task information and inputting it all at once. For a large amount of input, the batch method is convenient. Next, FIG. 5 shows the operations between the components of the DoTASK device, the new DoTASK device, and the GenTASK device, and also shows the operations between the components in the case of batch update by sequential update and batch update, and an external batch update program not related to the AI.

[0057] In the DoTASK device shown on the left side of FIG. 5, task execution is performed with a set of three points: a chatbot, a generative AI, and a task management system. Also, in the new DoTASK device, the execution of the task designed with a set of four points including a storage can be performed.

[0058] In the GenTASK device shown on the right side of FIG. 5, the case of batch update by sequential update and batch update, and an external batch update program not related to the AI is shown. In either the case of sequential update or batch update, the generative AI is involved with respect to the task management system, while in the case of external batch update, the generative AI is not involved with respect to the task management system.

[0059] (AVCM) As an information processing apparatus according to the present embodiment, as shown in FIG. 6, it is possible to further provide an AVCM as verification means for verifying the validity of the output of the generation AI for the input to the chatbot before output, either between the generation AI and the task management system or between the generation AI and the storage. Here, AVCM means AI Verification and Correction Mechanism and relates to the applicant's previous application (Japanese Patent Application No. 2024-169700).

[0060] In FIG. 6(a), the A subsystem is typically a so-called chatbot, but does not necessarily need to realize a conversation between the generation AI and the user, and may be an AI application that generates information by a simpler operation using a user interface or the like.

[0061] The role of the B subsystem is, after receiving the output from the A subsystem (after batch reception), to verify its correctness, approve the result if it is correct, and if there is inappropriate content such as an error, feedback to the A subsystem to eliminate the inappropriate content (for example, return the reason for the error or return the inappropriate content together with the reason). Also, if the output is not the final one, verification is performed on the content at an intermediate stage (sequential reception), and advice is returned to the A subsystem, that is, to the generation AI.

[0062] (Information Processing Method and Program) Based on the above-described embodiment, the information processing apparatus according to the present invention has been described. The information processing method and program according to the present invention will be described below.

[0063] The information processing method according to the present invention is an information processing method for performing task design using a generation AI via a chatbot application via natural language, having a first step of causing a storage provided outside or inside the generation AI to cooperate with the generation AI for the task design, and a second step of storing the task information designed for the task in a predetermined task management system.

[0064] Further, preferably, in the task management system, a third step is performed in which the generation AI and the storage cooperate so that the designed task is executed, and predetermined knowledge is learned so that the input position where the input data of the generation AI is input and the output position where the output data of the generation AI is stored can be inferred in the predetermined task management system.

[0065] The program according to the present invention causes a computer to execute each of the above steps.

[0066] Here, as shown in FIG. 7, an information terminal such as a personal computer (PC) or a smartphone to which input is made by a user is connected to a server, a generation AI, a storage, and a predetermined task management system via the Internet, and the above program can be executed.

[0067] (Modification example) As described above, the preferred embodiments of the present invention have been described. However, the present invention is not limited to these embodiments, and various modifications and changes can be made within the scope of the gist thereof.

[0068] For example, in the above-described embodiment Regarding the execution of a task designed by the GenTASK device for task design, it has been described that it can be performed through manual input / output operations or in combination with the new DoTASK device shown in Fig. 3(b) (using the first generative AI for task design and the second generative AI for task execution). However, it is also possible to perform task design and the execution of the designed task using only the GenTASK device (the generative AI for task design also serves as the generative AI for task execution). That is, task design can be carried out using Knowledge 1 to 4 shown in Fig. 8 through the cooperation of the generative AI and the storage, and task execution using the generative AI can be performed so that the inference of the input / output positions described in Fig. 3(b) can be made using Knowledge α, β, γ shown in Fig. 8.

[0069] Furthermore, even in the case of a generation AI for task design or a generation AI for task execution, a plurality of generations of AI with different specifications or performances may be switched by a chatbot application or by a user's operation. Regarding this, Figs. 9 and 10 are shown. Fig. 9 shows an example of a large-scale cooking plan in which the generative AI is switched during the task design process. In Fig. 9, time progresses from the upper part to the lower part of the figure. High-performance AI is involved on January 15 and January 25, while popularization version AI is involved at the stage of January 31. After the final determination of the recipe and the quantity, a process such as adjustment with the supplier is developed.

[0071] And in FIG. 10, it shows that task design in the large-scale cooking plan of FIG. 9 is performed and the task execution of the designed task is performed. In addition to the above-described chatbot application, a chatbot application having a function of complementing a conversation in natural language by an operation such as pressing a button and having a function of improving efficiency may be used.

[0072] Also, in the above-described embodiments, the generative AI and the predetermined task management system have been described as existing separately from the PC (Fig. 7), but at least one of the generative AI and the predetermined task management system may exist in the PC.

[0073] Also, in the above-described embodiments, a transfer program has been shown, but the transfer program includes communication by API, and a plurality of devices can be connected to a remote location and distributedly managed.

[0074] Also, the input / output method of the chatbot is not limited to text, and may be voice or the like.

[0075] Also, in the above-described embodiments, a method of using the tool of OpenAI's Assistants API as a method of learning predetermined knowledge has been described. However, in addition to that, there may be methods such as the development of an LLM (Large Language Model), its fine-tuning, or simple learning by a prompt, and further simplifying the method by a prompt with a chatbot application.

Explanation of Reference Numerals

[0076] α ··· Knowledge for correctly executing generation using generative AI, β ··· Input and output positions on the task management system for task execution, that is, knowledge for determining the task to be input or output and its attributes, γ ··· Operational knowledge for acquiring data from the input position on the predetermined task management system and storing the data at the output position on the task management system

Claims

1. A first generative AI for designing a new set of tasks; A second generation AI for performing the designed task; and A chatbot application that allows a user to converse with the first generation AI and the second generation AI via natural language; An information storage means is provided outside or inside the first generation AI, and serves as a storage location for knowledge information that the first generation AI learns in order to design the new series of tasks, and a storage location for design information handled by the first generation AI as an intermediate or final result; A storage is provided outside or inside the second generation AI as an information storage means, and is used as a storage location for knowledge information that the second generation AI learns to execute a designed task, and a storage location for input / output data handled by the second generation AI; a task management system in which the final result by the first generating AI is stored; An adapter function unit for operating the task management system from the chatbot application, The first generation AI has learned common sense knowledge, business / job knowledge, task design information format knowledge, and adapter function operation knowledge in advance in order to design the new series of tasks that are not stored in the task management system in advance, and the first generation AI is capable of reflecting the design information as the final result stored in the storage in the task management system via the adapter function unit, The second generation AI has previously learned input / output knowledge, common sense knowledge, task-specific knowledge, and adapter function operation knowledge in order to infer the location of input information as information required for a task executed in the task management system, and the location of output information as information obtained as a result of executing the task; An information processing device characterized in that the second generation AI is provided separately from the first generation AI or is combined with the first generation AI.

2. 2. The information processing apparatus according to claim 1, wherein the new series of tasks is designed based on a latest status of the task management system in operation.

3. A first generative AI for designing a new set of tasks; A second generation AI for performing the designed task; and A chatbot application that allows a user to converse with the first generation AI and the second generation AI via natural language; An information storage means is provided outside or inside the first generation AI, and serves as a storage location for knowledge information that the first generation AI learns in order to design the new series of tasks, and a storage location for design information handled by the first generation AI as an intermediate or final result; A storage is provided outside or inside the second generation AI as an information storage means, and is used as a storage location for knowledge information that the second generation AI learns to execute a designed task, and a storage location for input / output data handled by the second generation AI; a task management system in which the final result by the first generating AI is stored; An adapter function unit for operating the task management system from the chatbot application, An information processing method in an information processing device in which the second generation AI is provided separately from the first generation AI or the first generation AI also serves as the second generation AI, The first generation AI learns in advance common sense knowledge, business / job knowledge, task design information format knowledge, and adapter function operation knowledge in order to design the new series of tasks that are not stored in the task management system in advance; A step in which the first generation AI reflects the design information as the final result stored in the storage in the task management system via the adapter function unit; An information processing method characterized in that the second generation AI has a step of learning input / output knowledge, general common sense knowledge, business-specific knowledge, and adapter function operation knowledge in advance in order to infer the location of input information as information required for a task executed in the task management system, and the location of output information as information obtained as a result of executing the task.

4. A program for causing a computer to execute the steps according to claim 3.

Images