Information processing system, information processing method, and program

Abstract

We provide an information processing system, information processing method, and program that enable the streamlining of the planning and development process for new businesses. [Solution] The processor of the support server 10 in the information processing system causes the large-scale language model in the LLM server 2 to generate business ideas corresponding to the business theme of the new business. The processor causes the large-scale language model to score the business ideas and performs a judgment process to determine whether the scoring result of the business ideas meets the judgment criteria. If the scoring result meets the judgment criteria, the processor causes the large-scale language model to generate planning data based on the business ideas. If the scoring result does not meet the judgment criteria, the processor causes the large-scale language model to improve the business ideas and then performs the judgment process again.

Description

Technical Field

[0001] The present invention relates to an information processing system, an information processing method, and a program for supporting the planning and establishment of a new business.

Background Art

[0002] In the work of planning and establishing a new business, technologies for improving the efficiency of such work by using a computer system have been studied. For example, Patent Document 1 describes a business strategy evaluation system that performs business strategy evaluation using a computer system. In the technology described in Patent Document 1, the probability of success of a business strategy is evaluated by using a learned model corresponding to a plurality of strategy frameworks.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In the planning and establishment of a new business, further improvement in work efficiency is required to cope with rapid changes in the market and diversification of consumer needs.

Means for Solving the Problems

[0005] An information processing system that solves the above problems is an information processing system equipped with a processor, wherein the processor performs an idea generation process to cause a large-scale language model to generate business ideas corresponding to a business theme related to a new business, causes the large-scale language model to score the business ideas, and performs a judgment process to determine whether the scoring result of the business ideas meets the judgment criteria, if the scoring result in the judgment process meets the judgment criteria, it performs a planning data generation process to cause the large-scale language model to generate planning data based on the business ideas, and if the scoring result in the judgment process does not meet the judgment criteria, it performs an improvement process to cause the large-scale language model to improve the business ideas, and then performs the judgment process again.

[0006] An information processing method for solving the above problems is an information processing method using an information processing system equipped with a processor, wherein the processor executes an idea generation process to cause a large-scale language model to generate business ideas corresponding to a business theme related to a new business, causes the large-scale language model to score the business ideas, and executes a judgment process to determine whether the scoring result of the business ideas meets the judgment criteria, if the scoring result in the judgment process meets the judgment criteria, executes a planning data generation process to cause the large-scale language model to generate planning data based on the business ideas, and if the scoring result in the judgment process does not meet the judgment criteria, executes an improvement process to cause the large-scale language model to improve the business ideas, and then executes the judgment process again. A program for solving the above problems makes the processor equipped with the information processing system function as a means for executing the above information processing method.

[0007] According to the above information processing system, information processing method, and program, the planning and development of new businesses can be streamlined by generating and improving business ideas aligned with the business theme into a large-scale language model. Furthermore, if the scoring results do not meet the evaluation criteria, the quality of business ideas obtained from the large-scale language model can be ensured by repeatedly improving the business ideas through the improvement process.

[0008] In the above-described information processing system, if the scoring result in the judgment process does not meet the judgment criteria, the processor may divide the content of the business idea into elements of a framework for examining the business content in a manner consistent with the framework, identify at least one of the elements in the framework as the pivot target element, have the large-scale language model generate multiple pivot ideas by gradually changing the content of the pivot target element from the base content based on the content of the business idea divided into elements of the framework, have the large-scale language model score each pivot idea, and then execute a pivot process to identify the pivot idea with the best scoring result among the multiple pivot ideas as the pivoted business idea, before executing the improvement process. With the above configuration, the direction of a business idea that does not meet the judgment criteria can be changed. In this case, by generating multiple pivot ideas by gradually changing the content of a specific pivot target element in the business idea, it is possible to comprehensively identify proposed changes to the pivot target element. Then, by identifying the pivot idea with the highest score from among the multiple pivot ideas comprehensively identified, the highest quality pivot idea can be selected as the business idea for the pivoted business.

[0009] In the above-described information processing system, in the process of generating the multiple pivot ideas in the large-scale language model, the processor may be configured to generate the multiple pivot ideas in the large-scale language model by changing the content of the pivot target element to content obtained by rotating the semantic vector of the sentence representing the base content at fixed angles between 0 and 360 degrees. With the above configuration, it is possible to more comprehensively identify proposed changes to the pivot target element.

[0010] In the above-described information processing system, the processor may be configured such that, if the number of times the judgment process has been executed is less than a set number, and the scoring result in the judgment process does not meet the judgment criteria, it executes the improvement process without performing the pivot process, and then executes the judgment process again. If the number of times the judgment process has been executed is equal to or greater than the set number, and the latest scoring result in the latest judgment process does not meet the judgment criteria, and the increase in the scoring result when comparing the scoring result of a standard judgment process performed before the latest judgment process with the latest scoring result is less than a threshold, it executes the pivot process. With the above configuration, for business ideas that can be improved without performing a pivot process, the improvement of the business idea can be advanced while making use of the elements of the base business idea. Furthermore, by performing a pivot process when an increase in the scoring result of the business idea can no longer be expected through improvements without a pivot process, the business idea can be improved more effectively.

[0011] In the above-described information processing system, the processor may be configured to have the large-scale language model score the business idea from the perspective of multiple judges, each with a different personality profile. With this configuration, the large-scale language model can score the business idea to be scored from multiple perspectives.

[0012] In the above-described information processing system, the multiple judges may include a judge whose personality profile has been set as the person in charge of the corporate entity implementing the new business. With this configuration, the compatibility between the new business and the corporate entity can be appropriately evaluated.

[0013] In the above-described information processing system, the processor may, in the improvement process, cause the large-scale language model to generate a personality profile of a virtual expert in the business field of the business idea, cause the large-scale language model to generate responses from an interview conducted with the virtual expert having the personality profile to obtain an evaluation of the business idea, and cause the large-scale language model to improve the business idea based on the responses from the interview. With the above configuration, the business idea can be improved based on an evaluation from the perspective of a virtual expert corresponding to the business idea.

[0014] In the above-described information processing system, the processor may, in the improvement process, cause the large-scale language model to generate a personality profile of the target user of the business idea, cause the large-scale language model to generate responses from an interview conducted with the target user having the personality profile to obtain an evaluation of the business idea, and cause the large-scale language model to improve the business idea based on the interview responses. With the above configuration, the business idea can be improved based on an evaluation from the perspective of a hypothetical user corresponding to the business idea. [Effects of the Invention]

[0015] According to this invention, the planning and development of new businesses can be made more efficient. [Brief explanation of the drawing]

[0016] [Figure 1] Figure 1 is a diagram showing the configuration of an information processing system. [Figure 2] Figure 2 shows the hardware configuration of the support server. [Figure 3] Figure 3 shows the configuration of the processor and memory unit of the support server. [Figure 4] Figure 4 is a flowchart showing the steps of the support process. [Figure 5]FIG. 5 is a flowchart showing the procedure of the reconsideration process in the support process.

MODE FOR CARRYING OUT THE INVENTION

[0017] Hereinafter, an embodiment of an information processing system, an information processing method, and a program will be described with reference to the drawings. The information processing system of this embodiment is a computer system for supporting the planning and establishment of a new business. The new business to be planned and established may be a business related to any trading partner. The trading partner includes goods and services.

[0018] <Configuration of Information Processing System> As shown in FIG. 1, the information processing system of this embodiment includes a support server 10. The support server 10 supports the planning and establishment of a new business. The support server 10 is communicably connected to the user terminal 1, the LLM server 2, and the external server 3 via a network line.

[0019] The user terminal 1 is a computer terminal managed by a user who uses the information processing system. The user terminal 1 is, for example, a mobile terminal such as a tablet or a smartphone corresponding to a mobile communication system. Note that the user terminal 1 may be a desktop or laptop personal computer.

[0020] The LLM server 2 includes a large language model (LLM). The large language model is a model learned from a large text corpus. The large language model is used to perform natural language understanding tasks. The large language model has a function of interpreting the sentence given to the prompt and generating an appropriate response in that context. The LLM server 2 may be a server managed by an enterprise body or the like that manages the information processing system, or may be an external server. Note that the enterprise body is composed of one enterprise or a set of enterprises having economic and organizational relationships.

[0021] External server 3 is a server to be searched on the external network when support server 10 searches for various information. External server 3 may be any server that includes external information necessary when a large language model supports the planning and establishment of a new business. Support server 10 inputs the external information obtained from external server 3 into the large language model together with the prompt or as part of the prompt. External server 3 includes, for example, information on the technical field (such as academic papers) related to the new business that support server 10 targets for planning, trends in the field related to the new business, market size, data on laws or guidelines.

[0022] <Hardware Configuration> Next, referring to FIG. 2, an example of the hardware configuration of support server 10 will be described. Note that the hardware configuration of support server 10 is not limited to the following example, and it may have other hardware configurations. Also, user terminal 1 may have a configuration similar to the following hardware configuration.

[0023] As shown in FIG. 2, support server 10 includes a processor 11, a memory 12, a storage unit 13, an input / output IF 14, and a communication IF 15. Processor 11 is hardware for executing a set of instructions described in a program. Processor 11 is composed of an arithmetic unit, registers, peripheral circuits, etc.

[0024] Memory 12 temporarily stores programs and data processed by programs and the like. Memory 12 is, for example, a volatile memory such as DRAM (Dynamic Random Access Memory). Storage unit 13 is a storage for storing data. Storage unit 13 is, for example, a flash memory, HDD (Hard Disk Drive), etc.

[0025] The input / output IF14 is an interface that includes an input device for receiving operation input from the administrator of the support server 10 and an output device for presenting information to the administrator. The input device is, for example, a pointing device such as a mouse or a keyboard. The input device may also be a microphone for voice input. The output device is, for example, a display panel such as an LCD display panel or an OLED panel. The input / output IF14 may also be a touch panel that combines, for example, a display panel as a display unit for displaying images and a touch pad as an input unit for receiving operations from the user. The communication IF15 is an interface for inputting and outputting signals for communication with external devices.

[0026] <Processor 11> As shown in Figure 3, the processor 11 of the support server 10 inputs prompts to the large-scale language model provided by the LLM server 2, and obtains answers corresponding to the input prompts from the large-scale language model. Specifically, the processor 11 functions as an idea acquisition unit 11A, an improvement processing unit 11B, a judgment processing unit 11C, a handover item acquisition unit 11D, a pivot processing unit 11E, a planning data acquisition unit 11F, etc., by executing the support program. The support program is stored in the storage unit 13.

[0027] <Idea Acquisition Unit 11A> The idea acquisition unit 11A executes an idea generation process that generates business ideas for new businesses in accordance with the business theme input by the user, using a large-scale language model. For example, the business theme may include information about the trading targets of the new business. The information about the trading targets may include information about the technology field of the trading targets, or information such as functions, features, and price range. For example, the business theme may include information about the target audience of the new business. For example, the business theme may include information about the purpose of the new business, or the goals and vision to be achieved through the new business.

[0028] Business ideas generated by large-scale language models include, for example, information about the concepts and characteristics of the transactions to be handled in the new business, and information about the target audience of the new business. For example, business ideas include information about the social significance and problems to be solved in the new business. For example, business ideas include information about the risk assessment of the new business.

[0029] <Improvement Processing Unit 11B> The improvement processing unit 11B executes an improvement process to refine the business idea into a large-scale language model. The improvement process focuses on four main aspects: originality, usefulness, feasibility, and impact on social issues.

[0030] Specifically, the improvement process involves refining the business idea into a large-scale language model through improvement tasks that include the following seven items: The items related to originality consist of two items: expert interviews and market research. The items related to usefulness consist of interviews with potential users. The items related to feasibility consist of two items: technology research and risk assessment. The items related to impact on social issues consist of two items: social impact analysis and stakeholder analysis.

[0031] In expert interviews, responses from experts in the relevant business field are output to a large-scale language model to evaluate the originality of the business idea and identify areas for improvement. Originality, for example, includes three aspects: the originality of the idea itself, technological advancements, and the new value it brings to the market.

[0032] Specifically, in the expert interview, the improvement processing unit 11B first generates a personality profile of a hypothetical expert corresponding to the business idea in a large-scale language model. For example, the expert's personality profile includes information such as job title, skill set, job responsibilities, professionalism, and the direction of the desired answers. Next, the improvement processing unit 11B outputs the responses of the hypothetical expert with the generated personality profile to the large-scale language model, based on an interview about the originality of the business idea. Subsequently, the improvement processing unit 11B improves the business idea in the large-scale language model based on the interview responses.

[0033] In the market research, the improvement processing unit 11B first conducts a competitive analysis of the business field of the business idea and market trend research by searching the external server 3 using a search API (Application Programming Interface). At this time, the improvement processing unit 11B may have the large-scale language model create the research results data that summarizes the research results. Next, the improvement processing unit 11B inputs the research results as external information into the large-scale language model and has the large-scale language model evaluate the originality of the business idea based on the research results. After that, the improvement processing unit 11B improves the business idea in the large-scale language model based on the evaluation results of the market research.

[0034] In the prospective user interviews, the responses of prospective users of the business idea's target product are output to a large-scale language model, based on interviews about their pains, gains, and the acceptability of the product. Pains refer to problems or issues that customers want to eliminate. Gains refer to added value for customers. Acceptability of the product is information used to evaluate whether the product will be accepted by prospective users. Responses regarding acceptability of the product include, for example, the prospective users' impressions of using the product.

[0035] Specifically, in the prospective user interview, the improvement processing unit 11B first generates a personality profile of the prospective user of the transaction target handled by the business idea in a large-scale language model. For example, the prospective user's personality profile includes information such as the prospective user's age, gender, personality, values, and problems they are facing. Next, the improvement processing unit 11B outputs the responses from the prospective user with the generated personality profile to the large-scale language model if an interview is conducted regarding pains and gains and the acceptability of the transaction target. Subsequently, the improvement processing unit 11B improves the business idea in the large-scale language model based on the interview responses.

[0036] In the technical investigation, the improvement processing unit 11B first collects technical information related to the business field of the business idea by searching the external server 3 using a search API. At this time, the improvement processing unit 11B may also collect technical information from academic papers using a paper search API for searching academic papers. Alternatively, the improvement processing unit 11B may have a large-scale language model create a technical summary data that compiles the collected technical information. Next, the improvement processing unit 11B inputs the collected technical information as external information into the large-scale language model and improves the technical aspects of the business idea in the large-scale language model based on the technical information.

[0037] In the risk assessment, the improvement processing unit 11B first collects normative information, such as laws and guidelines related to the business field of the business idea, by searching the external server 3 using a search API. The type of normative information is not limited as long as it includes information for conducting a risk assessment from technical, legal, and ethical perspectives regarding the transaction targets handled by the business idea. Next, the improvement processing unit 11B inputs the collected normative information as external information into the large-scale language model and uses the normative information to identify and improve the potential risks of the business idea in the large-scale language model.

[0038] In the social impact analysis, the improvement processing unit 11B first generates personality profiles of experts in the business field of the business idea in a large-scale language model, similar to expert interviews. Next, the improvement processing unit 11B outputs to the large-scale language model the responses of experts with the generated personality profiles, who quantitatively and qualitatively evaluate the impact that the proposed transaction target of the business idea will have on society. Subsequently, the improvement processing unit 11B improves the business idea in the large-scale language model based on the evaluation results.

[0039] In stakeholder analysis, the improvement processing unit 11B outputs evaluation results of the positive and negative impacts that the transaction will have on key stakeholders in the business area addressed by the business idea to a large-scale language model. Subsequently, the improvement processing unit 11B improves the business idea in the large-scale language model based on the evaluation results.

[0040] Furthermore, the improvement process is not limited to improving the business idea item by item in the improvement task. For example, one could first output all the information needed to improve the business idea obtained from each item of the improvement task (including interview responses and evaluation results), and then improve the business idea based on all of that output.

[0041] <Determination Processing Unit 11C> The judgment processing unit 11C has a large-scale language model score the business idea from the perspective of multiple virtual judges, and performs a judgment process to determine whether the scoring result of the business idea meets the judgment criteria. For example, the multiple judges who score the business idea are set up with different personality profiles so that the business idea can be evaluated from multiple perspectives. The judgment process includes a scoring process and a pass / fail judgment process.

[0042] In the scoring process, the judgment processing unit 11C causes a large-scale language model to generate personality profiles of the judges who will score the business ideas. The personality profiles set for the judges are based on the following types of individuals: for example, the head of a company implementing a new business, an engineer, a manager, a government official in charge of new business development, or a venture capitalist.

[0043] Preferably, the multiple judges who score the business ideas include at least one judge whose personality profile has been established for a person in charge of the company. Persons in charge of the company include, for example, the CEO, Chairman of the Board, COO, Directors, CTO, CMO, and business unit heads.

[0044] The personality profile of the person in charge of a corporate entity is set to match the corporate information, such as the corporate field, size, corporate philosophy, mission and value in society, the corporate vision, values, behavioral guidelines, corporate strategy, and management policies. The judges for the person in charge of the corporate entity are assigned the role of reviewing business ideas from the perspective of whether they have synergy with the corporate entity's business and whether they fit the corporate entity's values ​​and culture.

[0045] The engineer's personality profile will define their character as an expert in the technical field that will be addressed in the new business. The engineer judges will be tasked with evaluating business ideas from the perspective of whether they are technically feasible.

[0046] Executives serving as judges are tasked with reviewing business ideas from a financial perspective, considering factors such as feasibility, risk, and whether the investment can be recouped. Government officials in charge of new business development serve as judges, reviewing business ideas from a government standpoint. Venture capitalists serve as judges, as venture capital investment managers, comprehensively reviewing the future potential of business ideas.

[0047] The personality profiles of the judges who score the business ideas may be generated by a large-scale language model according to the business idea, or they may be predefined personality profiles. For example, system prompts containing predefined personality profiles may be stored in the memory unit 13. In this case, the process of generating the judges' personality profiles in the large-scale language model may be omitted during the scoring process. Alternatively, the personality profiles of some of the judges may be generated by the large-scale language model, while predefined personality profiles may be applied to the other judges.

[0048] Next, in the scoring process, the judgment processing unit 11C generates scoring results in a large-scale language model when a business idea is scored from the perspective of multiple judges, each with a different personality profile. The evaluation items in the scoring are broadly classified into four perspectives: originality, usefulness, feasibility, and impact on social issues. From the perspective of originality, three items are evaluated: the innovativeness of the business idea, the advancedness of the technology, and the novelty of the market. From the perspective of usefulness, three items are evaluated: problem-solving ability, ease of use, and practicality. From the perspective of feasibility, three items are evaluated: technical feasibility, financial feasibility, and the planning for implementation. From the perspective of impact on social issues, three items are evaluated: impact on social issues, sustainability, and fairness. The judgment processing unit 11C generates scoring results in a large-scale language model when a business idea is scored by each of the multiple judges based on the above 12 items.

[0049] In the scoring process, the judgment processing unit 11C inputs the scoring criteria for each evaluation item into a large-scale language model. For example, in the scoring process, each evaluation item is scored on a scale from 0 to 1 point. At this time, the judgment processing unit 11C inputs examples of 0 points and 1 point as scoring criteria for each evaluation item into the large-scale language model. As an example, an example of a 0 point for the innovativeness of a business idea is "an idea that merely slightly modifies or improves upon existing ideas or products and lacks fundamental novelty." An example of a 1 point for the innovativeness of a business idea is "an innovative idea that has the potential to overturn industry norms and redefine the rules of the market."

[0050] In the scoring process, the large-scale language model generates scores for each evaluation item based on the evaluation criteria described above, assuming that each of the multiple evaluators scores that item. In addition, during the scoring process, the evaluation processing unit 11C causes the large-scale language model to generate feedback on each evaluation item that serves as the basis for each evaluator's scoring result, along with the scoring result for each evaluation item.

[0051] In the pass / fail judgment process, the judgment processing unit 11C determines whether the scoring result of the business idea, which was scored in the scoring process, meets the judgment criteria. The judgment criteria are stored in the storage unit 13. For example, the judgment criteria include a threshold value for the average score of all judges for all evaluation items. For example, the judgment processing unit 11C determines that the scoring result of the business idea is above the passing score if the average score of all judges for all evaluation items is above the threshold value.

[0052] The evaluation criteria may include a threshold for the average score, as well as a lower limit that must be met for at least one of the multiple evaluation items. In this case, the evaluation processing unit 11C determines that the business idea has passed or is above the passing score if the scores given by each evaluator for the item for which a lower limit has been set all exceed the lower limit, and the average score of all evaluators for all evaluation items is equal to or greater than the threshold.

[0053] <Transfer Information Acquisition Unit 11D> The handover item acquisition unit 11D executes a handover item generation process for business ideas that have been determined to meet the judgment criteria by the judgment processing unit 11C. This process generates handover items in a large-scale language model that will be passed on to the user as matters for future consideration. For example, the handover item acquisition unit 11D generates handover items in a large-scale language model based on feedback for each evaluation item obtained in the scoring process.

[0054] The handover information acquisition unit 11D may generate information regarding the market size of the new business in addition to the handover information of the business idea using a large-scale language model. For example, the market size information may include TAM (Total Addressable Market), SAM (Serviceable Addressable Market), and SOM (Serviceable Obtainable Market) values. TAM represents the market size if the trading target captures 100% of the market. SAM represents the market size that can actually be targeted from within TAM, based on factors such as geographical location and demographics. SOM represents the market size within SAM, taking into account actual competition with other companies and resource constraints.

[0055] <Pivot Processing Unit 11E> The pivot processing unit 11E performs a pivot process to change the direction of the business content of a business idea that has been determined by the judgment processing unit 11C to not meet the judgment criteria. The pivot process includes a target element identification process, an identification process, and a selection process.

[0056] In the target element identification process, the pivot processing unit 11E causes the large-scale language model to perform a process of dividing the content of the business idea into multiple elements, in accordance with a framework for examining the business content by dividing it into multiple elements. The framework here is a business framework for visualizing the business model of a startup. The following explanation will describe the case where the framework is the Lean Canvas, but other frameworks such as the Business Model Canvas may also be used.

[0057] In the Lean Canvas, a new business is typically broken down into the following nine elements, which are then examined in order. The nine elements of the Lean Canvas are: 1. Customer Problem, 2. Customer Segment, 3. Value Proposition, 4. Solution, 5. Channels, 6. Revenue Streams, 7. Cost Structure, 8. Key Metrics, and 9. Outstanding Advantage. The numbers preceding each element indicate the order in which they should be examined in the Lean Canvas.

[0058] In the target element identification process, the pivot processing unit 11E instructs the large-scale language model to divide the content of the business idea into the nine elements mentioned above. It also identifies the pivot target element from among the multiple elements in the Lean Canvas. The specific method for identifying the pivot target element will be described later.

[0059] The pivot processing unit 11E performs a target element identification process, followed by an element identification process. In the element identification process, the pivot processing unit 11E uses the content obtained by dividing the business idea into each element of the Lean Canvas as a base, and generates multiple pivot ideas in a large-scale language model by gradually changing the content of the pivot target element from the base content. Note that in the pivot ideas, the content of each element of the Lean Canvas other than the pivot target element remains unchanged from the base content.

[0060] For example, a pivot idea is data that includes the base content of each element of the Lean Canvas other than the pivot element, and the content of the pivot element as it has been modified from its base content. Note that a pivot idea may be data consisting of the content of each element of the Lean Canvas, as in the example above, or it may be data with an amount of information equivalent to a business idea, including information other than the content of each element of the Lean Canvas.

[0061] As a concrete example of the brainstorming process, the pivot processing unit 11E causes the large-scale language model to generate multiple pivot ideas, in which the content of the pivot target element is changed by rotating it at fixed angles between 0 and 360 degrees relative to the semantic vector of the base content.

[0062] For example, if the rotation angle is 45 degrees, the large-scale language model will output seven pivot ideas in 45-degree increments from 45 degrees to 315 degrees. In this case, the closer the rotation angle is to 180 degrees relative to the semantic vector of the base content, the greater the change in the direction of the business content will be in the pivot idea. Note that rotation angles of 0 degrees and 360 degrees correspond to the base content. When the pivot processing unit 11E causes the large-scale language model to generate pivot ideas, it adds constraints to ensure that the content of the pivot ideas does not deviate from the concept set as the business theme.

[0063] The following is a concrete example of the brainstorming process. For example, let's consider a business idea with the theme of "nutritional management services and food sales promotion for elderly people living alone." In this business idea, let's assume that the content of "2. Customer Segment" in the Lean Canvas is "elderly people." In this case, we will rotate the meaning vector of the word "elderly people" in 45-degree increments between 0 and 360 degrees. In this case, using "elderly people" at rotation angles of 0 and 360 degrees as a base, seven pivot ideas will be output in which the content of "2. Customer Segment" is changed, such as "elderly people's families" at 45 degrees, "nursing care facilities" at 90 degrees, "medical institutions" at 135 degrees, "fitness centers" at 180 degrees, "local community centers" at 225 degrees, "home delivery service providers for the elderly" at 270 degrees, and "administrators of online communities for the elderly" at 315 degrees.

[0064] In this pivot idea, the content of "2. Customer Segment" shifts from targeting the elderly themselves to targeting those connected to the elderly, or facilities and services used by the elderly. Furthermore, while it is generally believed that the elderly tend to be less active, as the rotation angle approaches 180 degrees, the content shifts to targeting facilities and services used by more active elderly people. Thus, in the brainstorming process, multiple pivot ideas with different directions are output to the large-scale language model as options when changing the pivot target element.

[0065] Furthermore, the rotation angle used to generate pivot ideas is not limited to 45-degree increments; it can be any angle, and the rotation angle does not have to be constant. The rotation range used to generate pivot ideas is not limited to 0 to 360 degrees; for example, it could be 0 to 180 degrees. In addition, two or more elements in the framework may be identified as pivot targets. In this case, the semantic vector of the text that collectively describes the content of the multiple elements identified as pivot targets is rotated.

[0066] The pivot processing unit 11E performs a selection process after performing an idea generation process. In the selection process, the pivot processing unit 11E has a large-scale language model score each pivot idea output in the idea generation process, and identifies the pivot idea with the best score as the business idea after the pivot.

[0067] Furthermore, the scoring method used in the selection process can be the same as that used by the judgment processing unit 11C. That is, in the selection process, each pivot idea is scored by a large-scale language model from the perspective of a judge with the same personality profile as the judge processing unit 11C. The pivot idea with the best scoring result is, for example, the pivot idea with the highest average score across all evaluation items by all judges.

[0068] The pivot processing unit 11E may execute a pivot necessity determination process to determine whether or not to perform pivot processing for business ideas that have been determined by the determination processing by the judgment processing unit 11C to not meet the judgment criteria. Details of the pivot necessity determination process will be described later.

[0069] <Planning Data Acquisition Department, 11th Floor> The planning data acquisition unit 11F executes a planning data generation process that generates planning data in a large-scale language model based on business ideas that have been determined to meet the judgment criteria by the judgment processing unit 11C. The planning data is a data file equivalent to a business plan for a new business that has been considered as a business idea. The form of the planning data may be a document file containing text data or image data, but it may also be an audio file or a video file. The planning data acquisition unit 11F outputs the planning data acquired from the large-scale language model to the user terminal 1 or another device.

[0070] The planning data includes at least the content of the business idea. Preferably, the planning data also includes data on handover items and market size obtained in the handover item generation process. The planning data may also include the scoring results of the scoring process and the history of considerations, such as whether or not pivot processing was performed. In the improvement task, the planning data may also include various information output from the large-scale language model.

[0071] <Storage section 13> As shown in Figure 3, the storage unit 13 of the support server 10 includes a prompt database 13A and a business planning database 13B.

[0072] The prompt database 13A stores prompt templates that the processor 11 inputs to the large-scale language model. The prompt templates include system prompts that describe commands and other information for the large-scale language model. The processor 11 reads the prompt templates necessary for each process from the prompt database 13A and incorporates various types of information that change according to the business theme (e.g., business ideas) into the templates as user prompts. This generates prompts to be input to the large-scale language model. The business planning database 13B stores the planning data generated by the planning data acquisition unit 11F in the large-scale language model.

[0073] <Support Processing> Next, we will explain the support processing using the information processing system with reference to Figures 4 and 5. As shown in Figure 4, in the support process, first, the processor 11 receives input of a business theme from the user (step S1). For example, the user inputs the business theme using the user terminal 1.

[0074] Next, the idea acquisition unit 11A executes an idea generation process (step S2) which generates business ideas for new businesses in a large-scale language model according to the business theme input by the user.

[0075] Next, the improvement processing unit 11B executes an improvement process to improve the business idea into a large-scale language model using an improvement task (step S3). In step S3, the improvement process performed on the business idea before the improvement process is called the first improvement process. The first improvement process is an example of a pre-judgment improvement process performed before the judgment process.

[0076] Next, the judgment processing unit 11C executes a scoring process in which the large-scale language model scores the business idea from the perspective of multiple virtual judges (step S4). In step S4, the judgment processing unit 11C may have the large-scale language model generate personality profiles of judges corresponding to the business idea, or it may apply predefined personality profiles.

[0077] Next, the judgment processing unit 11C performs a pass / fail judgment process (step S5) to determine whether the scoring result of the business idea scored in step S4 meets the judgment criteria. If, in step S5, the business idea is judged to meet the criteria (step S5: YES), the handover information acquisition unit 11D executes a handover information generation process to generate handover information for the business idea in a large-scale language model (step S6). Furthermore, in step S6, the handover information acquisition unit 11D may also generate information regarding the market size of the new business in the large-scale language model in addition to the handover information.

[0078] Next, the planning data acquisition unit 11F executes a planning data generation process (step S7) to generate planning data based on the business idea in a large-scale language model. After that, the planning data acquisition unit 11F outputs the planning data acquired from the large-scale language model to the user terminal 1 or another device. With the above process completed, the support process is finished.

[0079] If, in step S5, the business idea is determined not to meet the criteria (step S5: NO), the processor 11 performs a re-examination process to further review and improve the content of the business idea (step S8). After the re-examination process in step S8 is completed, the processor 11 performs the judgment process again, which includes the scoring process in step S4 and the pass / fail judgment process in step S5. In other words, the processor 11 repeats the re-examination process in step S8 and the judgment processes in steps S4 and S5 until the scoring result of the business idea meets the criteria.

[0080] Figure 5 shows the procedure for the re-examination process in step S8. As shown in Figure 5, in the re-examination process in step S8, first the pivot processing unit 11E performs a pivot necessity determination process to determine whether or not to perform pivot processing (step S8-1).

[0081] In the pivot necessity determination process, the pivot processing unit 11E determines whether both the following first pivot condition and second pivot condition are met. The first pivot condition is that the number of times the judgment processes in steps S4 and S5 have been executed is equal to or greater than a set number. The set number is stored in the storage unit 13 in advance.

[0082] The second pivot condition is that the increase in the score result when comparing the latest score result from the most recent judgment process with the score result from a standard judgment process performed before the most recent judgment process is less than a threshold. The standard judgment process refers to, for example, a judgment process performed a specific number of times before the most recent judgment process (e.g., two times before). In other words, the second pivot condition is that the improvement in the latest score result when compared with the score result from the standard judgment process is less than a predetermined threshold. Note that the specific number of times is less than the number set for the first pivot condition. For example, if the set number of times is 3, the specific number of times is 2 or less.

[0083] If both the first and second pivot conditions are met, it means that even if the re-examination process in step S8 and the determination processes in steps S4 and S5 are repeated multiple times, no effective improvement to the business idea is observed. In this embodiment, if both the first and second pivot conditions are met, the pivot process is executed. If either the first or second pivot condition is not met, the pivot process is not executed.

[0084] If it is determined in step S8-1 that both the first pivot condition and the second pivot condition are met (step S8-1: YES), the pivot processing unit 11E performs the target element identification process (step S8-2). In step S8-2, the pivot processing unit 11E causes the large-scale language model to divide the content of the business idea into each element of the Lean Canvas.

[0085] At this time, the pivot processing unit 11E specifies the pivot target element according to the order of consideration in the Lean Canvas. For example, if it is the first pivot processing, the element "1. Customer Issues" is specified as the pivot target element. If it is the second pivot processing, the element "2. Customer Segments" is specified as the pivot target element.

[0086] Similarly, when using business frameworks other than the Lean Canvas, you may specify the pivot elements according to the recommended order of consideration within that framework. Furthermore, the method for determining the pivot elements is not limited to the examples above. For example, the number of pivot processes may be associated with the pivot elements, regardless of the recommended order of consideration within the framework.

[0087] Next, the pivot processing unit 11E performs an analysis process to output multiple pivot ideas to a large-scale language model, each of which the direction of the content of the pivot target element is changed step by step from the base content (step S8-3).

[0088] Next, the pivot processing unit 11E has a large-scale language model score each pivot idea output in the brainstorming process, and performs a selection process to identify the pivot idea with the best score as the business idea after the pivot (step S8-4).

[0089] Next, the improvement processing unit 11B executes an improvement process to improve the pivoted business idea into a large-scale language model using the improvement task (step S8-5). In step S8-5, the improvement process performed on the pivoted business idea is referred to as the second improvement process. The second improvement process is an example of a post-judgment improvement process performed after the judgment process. After the completion of the second improvement process in step S8-5, the judgment processes in steps S4 and S5 are executed again.

[0090] If it is determined in step S8-1 that neither the first pivot condition nor the second pivot condition is met (step S8-1: NO), the pivot processing unit 11E performs an improvement process instead of performing a pivot process (step S8-6). In step S8-6, the improvement process performed on the business idea is referred to as the third improvement process. The third improvement process is an example of a post-determination improvement process performed after the determination process. After the completion of the third improvement process in step S8-6, the determination processes in steps S4 and S5 are performed again.

[0091] In the third improvement process of step S8-6, instead of implementing all seven improvement tasks, the large-scale language model may be instructed to implement only the necessary items from the seven, based on the scoring results and feedback obtained in the scoring process. More specifically, the four perspectives in the scoring process—originality, usefulness, feasibility, and impact on social issues—are the same as the four perspectives used when improving the business theme in the improvement tasks. Therefore, in the third improvement process, only the improvement tasks corresponding to the perspectives that received low scores in the scoring process may be implemented.

[0092] Based on the above, in the re-examination process of step S8, if the number of times the judgment process has been executed is less than the set number, and the scoring result in the judgment process does not meet the judgment criteria, the third improvement process is executed without performing the pivot process, and then the judgment process is executed again. Also, if the number of times the judgment process has been executed is equal to or greater than the set number, and the latest scoring result in the latest judgment process does not meet the judgment criteria, and the increase in the scoring result when comparing the scoring result of the standard judgment process with the latest scoring result is less than the threshold, then the pivot process is executed. After the pivot process, the second improvement process is executed, and then the judgment process is executed again. In other words, in the re-examination process of step S8, regardless of whether the pivot process is performed or not, either the second improvement process or the third improvement process (post-judgment improvement process) is executed.

[0093] <Effects of the Embodiment> (1) According to the information processing system of this embodiment, the planning work for new businesses can be made more efficient by generating and improving business ideas in line with the business theme using a large-scale language model. Furthermore, if the scoring result does not meet the judgment criteria, the quality of the business ideas obtained from the large-scale language model can be ensured by repeatedly improving the business ideas through a post-judgment improvement process.

[0094] (2) In the re-examination process of step S8, the direction of business ideas that do not meet the evaluation criteria can be changed by performing a pivot process. At this time, by generating multiple pivot ideas in which the content of a specific pivot target element in the business idea is changed step by step, all possible changes to the pivot target element can be comprehensively identified. Then, by identifying the pivot idea with the highest scoring result from among the multiple pivot ideas comprehensively identified as the business idea after the pivot, the highest quality pivot idea can be identified as the business idea after the pivot. In addition, by generating multiple pivot ideas in which the content of the pivot target element is changed by rotating the semantic vector of the base content at fixed angles between 0 and 360 degrees, all possible changes to the pivot target element can be comprehensively identified.

[0095] (3) By performing the pivot necessity determination process in step S8-1, business ideas that can be improved without pivoting can be improved while retaining the elements of the base business idea. Then, when improvements without pivoting can no longer be expected to increase the score of the business idea, pivoting can be performed to improve the business idea more effectively.

[0096] (4) In the scoring process, a large-scale language model is used to score business ideas from the perspective of multiple judges, each with a different personality profile. This allows the large-scale language model to score the business ideas being evaluated from multiple viewpoints.

[0097] (5) It is preferable that the multiple judges in the scoring process include judges who have set up personality profiles of the person in charge of the corporate entity implementing the new business. In this case, the compatibility between the new business and the corporate entity can be appropriately evaluated.

[0098] (6) In the scoring process, by having a large-scale language model score business ideas from the perspective of experts in the technology fields covered by the new business, it is possible to obtain scoring results from a technical perspective on the business ideas. By having a large-scale language model score business ideas from the perspective of managers, it is possible to obtain scoring results from a financial perspective on the business ideas. By having a large-scale language model score business ideas from the perspective of government officials in charge of new business development, it is possible to obtain scoring results from an administrative perspective on the business ideas. By having a large-scale language model score business ideas from the perspective of venture capitalists, it is possible to obtain scoring results that represent a comprehensive evaluation of the future potential of the business ideas.

[0099] (7) In the expert interviews and social impact analyses of the improvement task, a large-scale language model is used to generate personality profiles of hypothetical experts in the business field of the business idea. Next, the large-scale language model is used to generate responses from interviews conducted with these hypothetical experts to obtain their evaluations of the business idea. The large-scale language model is then used to improve the business idea based on the responses from these interviews. This improvement method allows for the improvement of the business idea based on evaluations from the perspective of hypothetical experts relevant to the business idea.

[0100] (8) In the improvement task, the interview with the hypothetical user generates a personality profile of the hypothetical user who will be the target of the business idea using a large-scale language model. Next, the interview with the hypothetical user who has the generated personality profile generates the responses to an interview to obtain their evaluation of the business idea using the large-scale language model. Then, the business idea is improved using the large-scale language model based on the responses to the interview. With this improvement method, the business idea can be improved based on an evaluation from the perspective of a hypothetical user who will be the target of the business idea.

[0101] <Example of changes> This embodiment can be implemented with the following modifications. This embodiment and the following modifications can be combined with each other to the extent that they do not contradict each other technically.

[0102] • In the scoring process, the personality profiles assigned to multiple judges may be modified as appropriate depending on the project theme. The scoring method in the scoring process is not limited to having a large-scale language model score business ideas from the perspectives of multiple judges, each with a different personality profile, as long as it is a method that yields appropriate scoring results for the business theme. For example, the business idea could be scored from the perspective of a single judge with an arbitrary personality profile. Alternatively, instead of having judges, the necessary perspectives for the scoring process could simply be input into the large-scale language model.

[0103] The conditions for determining whether a pivot is necessary in step S8-1 may be changed as appropriate. Also, the pivot determination process may be omitted in the re-examination process in step S8. In this case, if the scoring result in the determination process does not meet the criteria, the pivot process must be performed before the post-determination improvement process.

[0104] • In the review process of step S8, the pivot process may be omitted. In this case, if the scoring result in the judgment process does not meet the judgment criteria, the post-judgment improvement process is executed without performing the pivot process.

[0105] The first improvement process in step S3, which is performed on business ideas before the scoring process, may be omitted. In this case, the business ideas obtained from the large-scale language model in the idea generation process in step S2 are scored by the large-scale language model in the scoring process in step S4, without going through the first improvement process in step S3.

[0106] In any of the improvement processes from the first to the third improvement process, it is acceptable to implement at least one of the seven improvement tasks, rather than all of them. The content of each improvement task may be changed as appropriate. Furthermore, in any of the improvement processes from the first to the third improvement process, it is also acceptable to improve the business idea using methods other than the seven improvement tasks mentioned above.

[0107] The process of generating handover information in step S6 may be omitted. Also, in step S6, the process by which the handover information acquisition unit 11D generates information regarding the market size of the new business in a large-scale language model may be omitted.

[0108] The content of interviews with hypothetical experts or hypothetical users who have personality profiles generated by the large-scale language model is not limited. In other words, processor 11 may have the large-scale language model generate responses when conducting interviews with hypothetical experts or hypothetical users to obtain their evaluations of any content related to a business idea.

[0109] The process in step S1 is not limited to receiving business theme input from user terminal 1, but may also be, for example, to generate business themes in a specific field or of arbitrary content in a large-scale language model in response to operational input from user terminal 1. In this case, information such as the technology field and target audience may be input to the large-scale language model as conditions for generating business themes.

[0110] The information processing system may be implemented as a single device, such as the support server 10, or it may be distributed across multiple devices or subsystems that collaborate to execute programs. In the information processing system, the support server 10 and other devices may be implemented as a single device. In the information processing system, a cloud server may handle at least a part of the configuration of the support server 10. [Explanation of Symbols]

[0111] 1…User terminal 2…LLM Server 3…External server 10…Support Server 11… Processor 11A... Idea Acquisition Department 11B…Improvement Processing Unit 11C...Determination Processing Unit 11D... Department for acquiring handover information 11E... Pivot Processing Unit 11F…Project Data Acquisition Department 13...Storage section 13A…Prompt Database 13B... Business Planning Database

Claims

1. An information processing system equipped with a processor, The aforementioned processor, We execute an idea generation process that generates business ideas corresponding to the business theme of a new business using a large-scale language model. The large-scale language model is used to score the business idea, and a judgment process is performed to determine whether the scoring result of the business idea meets the judgment criteria. If the scoring result in the judgment process satisfies the judgment criteria, a planning data generation process is executed to generate planning data based on the business idea using the large-scale language model. If the scoring result in the judgment process does not meet the judgment criteria, an improvement process is performed on the large-scale language model to improve the business idea, and then the judgment process is performed again. Information processing system.

2. The aforementioned processor, If the scoring result in the aforementioned determination process does not meet the aforementioned determination criteria, In accordance with a framework for examining business content by dividing it into multiple elements, the content of the business idea is divided into each element of the framework, and at least one of the multiple elements in the framework is identified as the pivot element. Based on the content obtained by dividing the aforementioned business idea into each element of the framework, the large-scale language model generates multiple pivot ideas by gradually changing the content of the pivot target element from the base content. The large-scale language model scores each pivot idea, and after performing a pivot process to identify the pivot idea with the best score among the multiple pivot ideas as the post-pivot business idea, the improvement process is executed. The information processing system according to claim 1.

3. In the process of generating the aforementioned multiple pivot ideas using the large-scale language model, The processor causes the large-scale language model to generate a plurality of pivot ideas, which are obtained by changing the content of the pivot target element by rotating the semantic vector of the sentence representing the base content at fixed angles between 0 and 360 degrees. The information processing system according to claim 2.

4. The aforementioned processor, If the number of times the judgment process has been executed is less than the set number, and the scoring result in the judgment process does not meet the judgment criteria, the improvement process is executed without executing the pivot process, and then the judgment process is executed again. If the number of times the judgment process has been executed is greater than or equal to the set number, and the latest scoring result in the latest judgment process does not meet the judgment criteria, and the increase in the scoring result when comparing the scoring result of the standard judgment process performed before the latest judgment process with the latest scoring result is less than the threshold, then the pivot process is executed. The information processing system according to claim 2 or 3.

5. In the judgment process, the processor causes the large-scale language model to score the business idea from the perspective of multiple judges, each with a different personality profile. An information processing system according to any one of claims 1 to 3.

6. The aforementioned multiple assessors include assessors whose personality profiles have been set as those responsible for the corporate entity implementing the new business. The information processing system according to claim 5.

7. In the improvement process, the processor The large-scale language model generates a personality profile of a hypothetical expert in the business field of the aforementioned business idea. The large-scale language model generates responses from an interview conducted with the aforementioned hypothetical expert, who possesses the aforementioned personality profile, in order to obtain an evaluation of the business idea. The large-scale language model is used to improve the business idea based on the answers to the interview. An information processing system according to any one of claims 1 to 3.

8. In the improvement process, the processor The large-scale language model generates personality profiles of the target users of the business idea, The large-scale language model generates responses from the assumed users having the aforementioned personality profile, when an interview is conducted to obtain their evaluation of the business idea. The large-scale language model is used to improve the business idea based on the answers to the interview. An information processing system according to any one of claims 1 to 3.

9. An information processing method using an information processing system equipped with a processor, The aforementioned processor, We execute an idea generation process that generates business ideas corresponding to the business theme of a new business using a large-scale language model. The large-scale language model is used to score the business idea, and a judgment process is performed to determine whether the scoring result of the business idea meets the judgment criteria. If the scoring result in the judgment process satisfies the judgment criteria, a planning data generation process is executed to generate planning data based on the business idea using the large-scale language model. If the scoring result in the judgment process does not meet the judgment criteria, an improvement process is performed on the large-scale language model to improve the business idea, and then the judgment process is performed again. Information processing methods.

10. The processor in the information processing system is made to function as a means for executing the information processing method described in claim 9. program.

Images