Proposal support device

Abstract

A proposal support device is provided that supports new proposals for sales activities to customers. [Solution] The proposal support device 1 includes a first prompt creation module 11 that creates a first prompt that instructs a generation AI 3 to generate multiple proposal candidates; a model creation module 13 that acquires past proposal data including multiple past proposals, the past proposal data including a first explanatory variable indicating characteristics of each of the past proposals and a target variable indicating whether or not a contract has been concluded, and creates a model indicating the relationship between the first explanatory variable and the target variable; a second prompt creation module 12 that creates a second prompt that instructs the generation AI 3 to generate a second explanatory variable indicating characteristics of each of the proposal candidates; and a scoring module 14 that inputs the second explanatory variable into the model and calculates a first score for each of the proposal candidates.

Description

[Technical field]

[0001] The present invention relates to a proposal support device that supports new proposals for sales activities to customers. [Background technology]

[0002] In sales activities, it is very important to be able to make attractive proposals to customers. In addition to the content of the proposal, it is also important to use persuasive language to accurately convey its appeal. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2021-120803 A Summary of the Invention [Problem to be solved by the invention]

[0004] Patent Document 1 describes a creation support device that accepts a new inquiry including information on required specifications and supports the creation of a proposal including information on bidding specifications corresponding to the new inquiry. In Patent Document 1, the device calculates the degree of similarity between the contents of the new inquiry and the contents of past inquiries or past proposals, and uses the degree of similarity to extract contents useful for creating a proposal from history data associated with the past inquiries.

[0005] However, since Patent Document 1 extracts content from past history data, it cannot create new proposals. In addition, the proposals in Patent Document 1 are based on the requirements written in a new quotation, and the content of the proposal is bound by the requirements, making it unsuitable for creating new proposals.

[0006] One aspect of the present invention relates to a proposal support device that enables the creation of novel proposal candidates that are not bound by past proposals by having a generation AI generate proposal candidates, and that can eliminate proposal candidates that are unconventional or have a low probability of being concluded by evaluating the proposal candidates made by the generation AI based on the content of other proposals made in the past and whether or not they were concluded. [Means for solving the problem]

[0007] A proposal support device according to one aspect of the present invention comprises: A first prompt creation module that creates a first prompt that instructs a first generation AI to generate a plurality of suggestion candidates; a model creation module that acquires past proposal data including a plurality of past proposals, the past proposal data including a first explanatory variable indicating characteristics of each of the past proposals and a target variable indicating whether or not a contract has been concluded, and creates a model indicating a relationship between the first explanatory variable and the target variable; A second prompt creation module that creates a second prompt that instructs a second generation AI to generate a second explanatory variable indicating each of the characteristics of the suggestion candidates; a scoring module that inputs the second explanatory variable into the model to calculate a first score for each of the proposal candidates; Includes. [Brief description of the drawings]

[0008] [Figure 1] FIG. 1 shows a proposal support device 1 according to the first embodiment and an external device connected to the proposal support device 1. As shown in FIG. [Diagram 2] FIG. 2 shows functions and operations of the proposal support device 1 according to the first embodiment. [Figure 3A] FIG. 3A shows an example of a user input item for creating a first prompt in the first embodiment. [Figure 3B] FIG. 3B shows an example of the first prompt sent to Generation AI3. [Figure 3C]FIG. 3C shows an example of a response from Production AI3 to the first prompt shown in FIG. 3B. [Figure 4] FIG. 4 shows an example of past proposal data for creating a model in the first embodiment. [Diagram 5] FIG. 5 shows a list of explanatory variables generated in the first embodiment. [Figure 6A] FIG. 6A shows an example of a second prompt sent to generation AI3 when variable B1 is sought as the second explanatory variable. [Figure 6B] FIG. 6B shows an example of a second prompt sent to the generation AI3 to determine the variables P1-P3. [Figure 6C] FIG. 6C shows an example of a second prompt sent to generation AI3 for variable R3. [Figure 6D] FIG. 6D shows an example of a second prompt sent to generation AI3 for variable R4. [Figure 6E] FIG. 6E shows an example of a second prompt sent to Generation AI3 for variable R5. [Figure 6F] FIG. 6G shows an example of a second prompt sent to generation AI3 for variable R6. [Figure 7] FIG. 7 shows an example of proposal candidate data generated in the first embodiment. [Figure 8] FIG. 8 shows functions and operations of a proposal support device 1a according to the second embodiment. [Figure 9] FIG. 9 is a flowchart showing a process in which the latent score calculation module 16 calculates the latent score. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. Each embodiment described below shows an example of the present invention, and does not limit the contents of the present invention. Furthermore, not all of the configurations and operations described in each embodiment are necessarily essential as the configurations and operations of the present invention. Note that the same components are given the same reference symbols, and duplicated explanations will be omitted.

[0010] <1. First embodiment> <1-1. Configuration> 1 shows a proposal support device 1 according to the first embodiment and an external device connected to the proposal support device 1. The proposal support device 1 is a computer system including a CPU, a memory, and the like (not shown). The proposal support device 1 may be configured with one computer, or may be configured with multiple computers connected via a network. The proposal support device 1 is connected to external devices such as a database 2 and a generation AI 3.

[0011] The database 2 stores past proposal data and other data. The database 2 is not limited to one in which data is stored in one storage device, and may be one in which data is distributed and stored in multiple storage devices. The proposal support device 1 acquires various data from the database 2 and performs processes such as prompt creation, model creation, and scoring.

[0012] The generation AI 3 includes a large-scale language model (LLM). The large-scale language model is a language model constructed using a large amount of text data and deep learning technology, and processes tasks such as text generation, translation, answering questions, text summarization, and sentiment analysis in response to prompts sent from the proposal support device 1. It is desirable for the large-scale language model to have an attention mechanism. The attention mechanism is a mechanism for extracting important parts from the input, and greatly contributes to improving the processing speed and accuracy of the large-scale language model.

[0013] The generation AI 3 may further include an image generation AI. The image generation AI is a system that generates an image from a prompt written in text, and for example, a diffusion model is used. Note that the "image" in the following description may be a video. The generation AI 3 may further include a presentation material generation AI.

[0014] The proposal support device 1 sends a prompt to the generation AI 3 and obtains an output from the generation AI 3.

[0015] <1-2. Functions and operations> 2 shows functions and operations of the proposal support device 1 according to the first embodiment. The proposal support device 1 includes a first prompt creating module 11, a second prompt creating module 12, a model creating module 13, a scoring module 14, and a proposal selecting module 15. These modules are realized by loading programs into a memory included in the proposal support device 1 and executing them by a CPU.

[0016] The past proposal data stored in the database 2 includes, for each of a plurality of past proposals, a text, an actual proposal document or the like, a first explanatory variable, and a target variable. Details of the past proposal data will be described later with reference to FIG.

[0017] The first prompt creation module 11 receives user input such as information about the proposal recipient and creates a first prompt that instructs the creation AI 3 to generate multiple proposal candidates. A specific example of creating the first prompt will be described later with reference to Figures 3A to 3C.

[0018] The second prompt creation module 12 acquires a plurality of proposal candidates generated by the generation AI 3, and creates a second prompt that instructs the generation AI 3 to generate a second explanatory variable from the proposal candidates. The second explanatory variable is not necessarily created based only on the text of the proposal candidate, but may be created by referring to the text of a past proposal, by referring to the actual proposal document of a past proposal, or by referring to the first explanatory variable of a past proposal. A specific example of creating a second prompt will be described later with reference to Figures 6A to 6F.

[0019] The model creation module 13 acquires the first explanatory variables and the dependent variables of each of the multiple past proposals to be learned from the database 2, and performs machine learning from the combination of the first explanatory variables and the dependent variables to create a model showing the relationship between the first explanatory variables and the dependent variables.

[0020] The scoring module 14 acquires a second explanatory variable for each of the multiple proposal candidates to be predicted from the generation AI 3, inputs the second explanatory variable into the model, and calculates a first score for each of the multiple proposal candidates.

[0021] The proposal selection module 15 selects one or more proposals from the plurality of proposal candidates based on the first score.

[0022] In the present application, the generation AI3 that generates proposal candidates in response to the first prompt may be referred to as the first generation AI, and the generation AI3 that generates the second explanatory variables in response to the second prompt may be referred to as the second generation AI. Furthermore, the generation AI3 (see FIG. 8) that generates presentation materials in response to the third prompt may be referred to as the third generation AI. The first to third generation AIs may be the same generation AI or may be different generation AIs.

[0023] <1-3. Details of the first prompt> FIG. 3A shows an example of user input items for creating a first prompt in the first embodiment. The user can specify a proposal recipient (e.g., a specialized company) and a proposal object (e.g., creating an advertisement page for a product) from an input device (not shown) of the proposal support device 1 or an input device of a computer (not shown) connected to the proposal support device 1 via a network. If the user wants to limit the proposal recipient or proposal object to a specific or narrow range, the user can input specific information. Alternatively, if the user does not want to limit the proposal recipient or proposal object, the user does not need to input specific information such as the attributes of the proposal recipient, company name, and reference information.

[0024] FIG. 3B shows an example of a first prompt sent to the generation AI 3. The first prompt creation module 11 creates a first prompt including an explanatory sentence for explaining the proposed recipient and the proposed object based on the information input by the user (e.g., there is a specialized company that creates advertising pages for products), and a template for the generation AI 3 to generate proposal candidates (e.g., please provide 10 proposal ideas to get a job from this company). When creating the first prompt, the user may provide the input information to the generation AI 3 to have the generation AI 3 create the explanatory sentence. The template may be stored in advance. The number of proposal candidates generated by the generation AI 3 is not limited to 10, and may be about 100.

[0025] Figure 3C shows an example of a response from the generation AI3 to the first prompt shown in Figure 3B. The generation AI3 generates multiple suggestions using the information provided by the first prompt and by filling in the missing information with the most likely content. The suggestions generated by the generation AI3 are reproduced in Figure 7 as "Topic" and "Summary".

[0026] <1-4. Details of model creation> 4 shows an example of past proposal data for creating a model in the first embodiment. The past proposal data includes, for each of a plurality of past proposals, an ID, text, a first explanatory variable, and a target variable, and further includes an actual proposal document (not shown). Note that the second score is not necessary information in the first embodiment, and will be described later with reference to FIG. 8.

[0027] The text includes a theme, an outline, and a field, and the content manually input when it was actually proposed in the past can be used as is.

[0028] The first explanatory variables include variables B0 to B11, variables P1 to P3, and variables R1 to R6, and the specific contents of these will be described later with reference to Fig. 5. The first explanatory variables can be generated by the proposal support device 1 or can be generated by the generation AI 3, and the method thereof will be described together with the method of generating the second explanatory variables.

[0029] The objective variable is expressed as a value of 1 or 0 to indicate whether a trade was successfully concluded when it was actually proposed in the past.

[0030] The model creation module 13 creates a function indicating a relationship between the first explanatory variable and the dependent variable as a model, based on the first explanatory variable and the dependent variable for a plurality of past proposals.

[0031] <1-5. Details of variable generation> 5 shows a list of explanatory variables generated in the first embodiment. Since the second explanatory variables are input to a model created using the first explanatory variables, the items of the explanatory variables are common to the first explanatory variables and the second explanatory variables. However, the first explanatory variables indicating the characteristics of past proposals and the second explanatory variables indicating the characteristics of new proposal candidates differ in the presence or absence of performance data of the proposals, so the methods of generating the explanatory variables may differ.

[0032] <1-5-1. Variables B0~B11> The variables B0 to B11 are variables derived from the contents of the proposal data text or the actual proposal document.

[0033] The variable B0 is obtained by converting the fields included in the text of the proposal data into codes such as industrial classifications.

[0034] When the variables B1 to B11 are obtained as the first explanatory variables, they are mainly obtained from the contents of the actual proposal document, etc., or from the text of the proposal data, or other data. The generation AI 3 may be given the contents of the actual proposal document, etc., the text of the proposal data, or other data to generate the variables B1 to B11. As variables obtained from other data, attribute information on the proposal act such as the proposal history (e.g., value 1 for "introduction" and value 0 for "public recruitment") and the implementation form (e.g., value 1 for "collaboration" and value 0 for "contract") may be added, or the number of proposers or attribute information of the proposers may be added. On the other hand, when the variables B1 to B11 are obtained as the second explanatory variables, since there is no actual proposal, a generation method different from that for the first explanatory variables is used as follows.

[0035] FIG. 6A shows an example of a second prompt sent to the generation AI3 when the variable B1 is required as the second explanatory variable. The second prompt creation module 12 gives the generation AI3 the text of a plurality of past proposals (see FIG. 4) and the text of the proposal candidate (see FIG. 7), extracts the top five similarities of themes with the proposal candidate among the past proposals, and weights the variables B1 of the top five past proposals by contract success or failure. By weighting the average by contract success or failure (for example, weight 1 for success and weight 0.5 for failure), the data of success is given more importance. In this way, although there is no performance data for the proposal candidate, the variable B1 of the proposal candidate can be generated by using the variable B1 of a similar past proposal. The variables B2 to B11 of the proposal candidate can be generated in the same manner. Here, the case of extracting the top five similarities has been described, but it is not necessary to use five. Also, the case of weighting the average by contract success or failure has been described, but weighting by similarity may be further performed.

[0036] The proposal candidate variables B1 to B3 can also be used as the estimated amount, estimated number of personnel, and estimated construction period when actually proposing the contents of this proposal candidate. Since each of the proposal candidate variables B1 to B3 is a weighted average of the success or failure of past proposals, it is possible to present an estimate that is more likely to produce a successful result.

[0037] <1-5-2. Variables P1~P3> 5 again, the variables P1 to P3 are variables derived only from the text of the proposed data. The variables P1 to P3 are merely examples, and the variables may be further increased by, for example, decomposing the viewpoint.

[0038] 6B shows an example of a second prompt sent to the generation AI3 to obtain variables P1 to P3. The second prompt creation module 12 provides the generation AI3 with the text of a proposal candidate (see FIG. 7) and requests it to evaluate the difficulty, profitability, and cost. Although FIG. 6B shows a second prompt for obtaining variables P1 to P3 as second explanatory variables, the same may be applied to the first explanatory variables.

[0039] <1-5-3. Variables R1 to R6> 5 again, the variables R1 to R6 are variables derived by referring to past proposal data. The variables R1 to R6 are merely examples, and the variables may be further increased by, for example, decomposing the viewpoint.

[0040] Variable R1 is the number of hits in a search of past proposal data using the theme keyword, and the more proposals that contain the same keyword, the larger the value. Variable R2 is the number of types of actual past proposal documents (proposals, estimates, specifications, minutes, daily reports, etc.) that are hit in a search using the theme keyword, and the larger the value, the more proposals that have reached a higher proposal phase. Variables R1 and R2 can be obtained by performing a process of extracting keywords from the theme and a search process. These processes may be performed by the proposal support device 1, or may be performed by the generation AI 3.

[0041] The variable R3 (evaluation of the task), the variable R4 (evaluation of the implementation method), the variable R5 (evaluation of the variables P1, P2, P3), and the variable R6 (evaluation of the variables B1, B2, B3) can be generated by the generation AI3 using a technique called retrieval-augmented generation (RAG). Examples of the second prompt sent to the generation AI3 to obtain the variables R3 to R6 are shown in Figs. 6C to 6F, respectively. These figures show the second prompt for obtaining the variables R3 to R6 as the second explanatory variables, but the same can be applied to the first explanatory variables. However, when there is no past proposal data to be referred to, or when there is not enough of it, it is not necessary to refer to the past proposal data when obtaining the variables R3 to R6 as the first explanatory variables.

[0042] <1-6. Scoring details> 7 shows an example of proposal candidate data generated in the first embodiment. The proposal candidate data includes an ID, text, and a second explanatory variable for each of a plurality of proposal candidates.

[0043] The text includes a theme, an outline, and a field, where the theme and outline are directly taken from the text of the proposed candidate generated in FIG. 3C, and the field is taken from the proposed object entered by the user in FIG. 3A.

[0044] Figure 7 also shows the first score obtained by inputting the second explanatory variable into the model. The first score is a value corresponding to the objective variable in past proposal data, that is, a value indicating the probability that each proposal candidate will result in a trade. The larger the first score, the higher the probability that the proposal will be accepted by the customer. Therefore, by selecting a proposal candidate with a large first score, it is possible to make a proposal with a high expectation of trade.

[0045] <1-7.Effects> According to the first embodiment, the proposal support device 1 includes: a first prompt generation module 11 that generates a first prompt that instructs the generation AI 3 to generate a plurality of suggestion candidates; a model creation module 13 that acquires past proposal data including a plurality of past proposals, the past proposal data including a first explanatory variable indicating each characteristic of the past proposal and a target variable indicating whether or not the contract has been concluded, and creates a model indicating a relationship between the first explanatory variable and the target variable; A second prompt generation module 12 that generates a second prompt that instructs the generation AI 3 to generate a second explanatory variable indicating each characteristic of the proposal candidate; a scoring module 14 that inputs the second explanatory variables into the model to calculate a first score for each of the proposal candidates; Includes.

[0046] According to this, by having the generation AI3 generate new proposal candidates and calculating the first score based on past proposal data, it is possible to obtain free ideas that go beyond rehashing past proposals, and then evaluate the expectation of successful execution using the first score. The generation AI3 can have a phenomenon called illusion, which indicates a response that deviates from the prompt request, but since unusual proposals can be eliminated using the first score, this phenomenon is rather convenient from the perspective of generating new proposals, and the capabilities of the generation AI3 can be utilized without requiring perfection.

[0047] According to the first embodiment, the second prompt instructs the generation AI3 to generate second explanatory variables based on first explanatory variables of past suggestions similar to each of the suggestion candidates.

[0048] According to this, since the second explanatory variable can be generated even for a proposal candidate with no track record, it is possible to fill in the gaps in the second explanatory variable. In addition, by using a second explanatory variable generated based on the first explanatory variable together with a second explanatory variable generated without being based on the first explanatory variable, it is possible to perform a highly accurate evaluation of the proposal candidate.

[0049] <2. Second embodiment> <2-1. Functions and operations> Fig. 8 shows functions and operations of a proposal support device 1a according to the second embodiment. In addition to the various modules included in the proposal support device 1 described with reference to Fig. 2, the proposal support device 1a further includes a latent score calculation module 16, a content information acquisition module 17, and a third prompt creation module 18. These modules are realized by loading programs into a memory included in the proposal support device 1a and executing them by a CPU.

[0050] The past suggestion data stored in the database 2 further includes a second score for each of the past suggestions. The second score is obtained by the scoring module 14 inputting the first explanatory variables into the model. An example of the second score is shown in FIG. 4.

[0051] The potential score calculation module 16 calculates a potential score for a keyword included in the text of a proposal (see FIG. 7) selected by the proposal selection module 15, using the method shown in FIG. 9. The potential score indicates the potential of the keyword itself that can contribute to a contract.

[0052] FIG. 9 is a flowchart showing a process in which the latent score calculation module 16 calculates a latent score. In 161, the latent score calculation module 16 extracts a keyword from the text of the selected proposal. In 162, the latent score calculation module 16 searches past proposals with the extracted keyword. The search in 162 is not limited to a match search, but may be a similarity search. In 163, the latent score calculation module 16 weights the second scores of the searched past proposals based on whether or not they have been concluded (for example, weight 1 for agreement and weight 0.5 for no agreement). The higher the second score of the past proposal containing the keyword is and the more often it has actually been concluded, the higher the latent score obtained. By using a keyword with a high latent score in presentation materials, the possibility of concluding a deal can be increased.

[0053] Referring again to FIG. 8, the content information acquisition module 17 searches a database (not shown) of the company (the proposer company) or searches the web to acquire content information that will form the basis of presentation materials, such as the company's product information and web pages.

[0054] The third prompt creation module 18 creates a third prompt that instructs the generation AI 3 to create presentation materials based on the proposal data (text, variables B1 to B3, etc.) selected by the proposal selection module 15 and content information.

[0055] The third prompt includes a prompt that instructs the generation AI 3 to generate a presentation text with a high appeal. The third prompt preferably instructs the generation AI 3 to always include a keyword with a high potential score in the presentation text, or to use a keyword with a high potential score multiple times in the presentation text.

[0056] The third prompt includes a prompt that instructs the generation AI 3 to generate an image for a highly appealing presentation. The third prompt that instructs the generation AI 3 to generate an image preferably includes keywords with a high potential score, and the third prompt may be generated by the generation AI 3.

[0057] The third prompt includes a prompt that instructs the Generation AI3 to create a presentation using the generated text and images.

[0058] <2-2.Effects> According to the second embodiment, the proposal support device 1a A latent score calculation module 16 extracts keywords from the proposal candidates selected based on the first score and calculates latent scores of the keywords; A third prompt generation module 18 that generates a third prompt that instructs the generation AI 3 to generate a presentation material for the selected proposal candidate; Further comprising: The scoring module 14 further inputs the first explanatory variable into the model to calculate a second score for each of the past suggestions; The latent score calculation module 16 searches past proposal data by the keyword, and calculates a latent score based on a second score of the searched past proposal data; A third prompt creation module 18 creates a third prompt to include in the presentation material the keywords selected based on the potential scores.

[0059] This allows you to create appealing presentation materials by using keywords selected based on the potential score. [Explanation of symbols]

[0060] 1, 1a... proposal support device, 2... database, 3... generation AI, 11... first prompt creation module, 12... second prompt creation module, 13... model creation module, 14... scoring module, 15... proposal selection module, 16... potential score calculation module, 17... content information acquisition module, 18... third prompt creation module

Claims

1. A first prompt creation module that creates a first prompt that instructs a first generation AI to generate a plurality of suggestion candidates; a model creation module that acquires past proposal data including a plurality of past proposals, the past proposal data including a first explanatory variable indicating characteristics of each of the past proposals and a target variable indicating whether or not a contract has been concluded, and creates a model indicating a relationship between the first explanatory variable and the target variable; A second prompt creation module that creates a second prompt to instruct a second generation AI to generate a second explanatory variable indicating each characteristic of the suggestion candidate; a scoring module that inputs the second explanatory variables into the model to calculate a first score for each of the proposal candidates; A proposal support device comprising:

2. The proposal support device according to claim 1, The second prompt instructs the second generation AI to generate the second explanatory variables based on the first explanatory variables of the past suggestions similar to each of the candidate suggestions. Proposal support device.

3. The proposal support device according to claim 1, a latent score calculation module that extracts keywords from the proposal candidates selected based on the first score and calculates latent scores of the keywords; A third prompt creation module that creates a third prompt to instruct a third generation AI to generate a presentation material of the selected proposal candidate; Further comprising: The scoring module further inputs the first explanatory variable into the model to calculate a second score for each of the past suggestions; the potential score calculation module searches the past proposal data by the keyword, and calculates the potential score based on the second score of the searched past proposal data; the third prompt creation module creates the third prompt to include in the presentation material a keyword selected based on the potential score. Proposal support device.

Images