Recruitment support server, recruitment support system, and recruitment support method

Abstract

To provide technology that streamlines and optimizes recruitment operations. [Solution] The present invention relates to a recruitment support server that automates and streamlines recruitment operations using generation AI. The recruitment support server of the present invention comprises a prompt generation unit that generates prompts for recruitment decisions from a company's recruitment information, a recruitment decision unit that generates recruitment decision results from applicant information and recruitment decision prompts, a primary scout message generation unit that generates scout messages from applicant information and recruitment decision results, and a final scout message generation unit that generates a final scout message from transmission criteria and a draft scout message.

Description

Technical Field

[0001] The present invention relates to a recruitment support system.

Background Art

[0002] In recent years, in the recruitment activities of companies, with the increase and diversification of applicants, the burden on recruitment staff has been increasing. In particular, candidate selection, setting of recruitment criteria, and communication with candidates require a great deal of time and effort. In addition, due to judgments based on the experience and subjectivity of recruitment staff, there are variations in recruitment criteria, and there is a risk of overlooking excellent talents.

[0003] Furthermore, in communication with candidates, a uniform message cannot attract the interest of candidates, making it difficult to secure excellent talents. On the other hand, careful communication tailored to individual candidates places a heavy burden on recruitment staff and hinders efficient recruitment activities.

[0004] The document creation support system of Patent Document 1 extracts evaluation items from a source document including profile information of an organization or an individual, and creates an approach document after receiving evaluations for them.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0006] The document creation support system described in Patent Document 1 is intended for creating approach documents for organizations and individuals, but it is not specifically designed for recruitment. Therefore, it does not adequately address challenges specific to recruitment, such as the increased burden on recruiters due to the growing number and diversity of applicants, the inconsistency in recruitment criteria and the risk of overlooking talented individuals, and the optimization of communication with applicants.

[0007] Therefore, the present invention aims to solve these problems and provide a technology that realizes the efficiency and optimization of recruitment operations. [Means for solving the problem]

[0008] According to the present invention, A prompt generation unit that, by inputting recruitment information from a company, causes a first generation AI to generate prompts for hiring decisions for that company, A recruitment determination unit that, by inputting applicant information and the aforementioned recruitment determination prompt, causes a second generating AI to generate a recruitment determination result for the applicant, A recruitment determination correction unit that corrects the recruitment determination result by inputting the recruitment determination result, the inclusion criteria for allowing the applicant to be hired, and the exclusion criteria for excluding the applicant from being hired, the third generating AI A primary scouting message generation unit that, by inputting the aforementioned applicant information and the corrected hiring decision result, causes a fourth generating AI to generate a primary scouting message for the applicant, The system includes a final scout message generation unit that, by inputting the transmission criteria and the scout message draft, causes a fifth generation AI to generate the final scout message, A recruitment support server is available. [Effects of the Invention]

[0009] According to the present invention, by automatically generating prompts for hiring decisions from a company's recruitment information, the burden on hiring managers in setting hiring criteria can be significantly reduced. Furthermore, by using prompts, variations in hiring criteria can be suppressed, enabling consistent hiring decisions.

[0010] By automatically generating hiring decisions from applicant information and hiring decision prompts, the time and effort required for applicant selection can be significantly reduced. Furthermore, using AI-generated results allows for a multifaceted and objective evaluation of applicant characteristics, reducing the risk of overlooking talented individuals.

[0011] By automatically generating recruitment messages from applicant information and hiring decisions, the burden on recruiters in communicating with applicants can be significantly reduced. Furthermore, using AI-generated messages allows for the creation of personalized recruitment messages tailored to each applicant's individual needs and characteristics, thereby attracting applicant interest and securing top talent.

[0012] By automatically generating the final recruitment message from the sending criteria and draft message, the quality of the recruitment message can be ensured while reducing the effort required of recruiters to send them. Furthermore, by setting sending criteria, the content and tone of recruitment messages can be standardized to align with recruitment policies.

[0013] Furthermore, by using different generation AIs for recruitment decision-making, scout message generation, and final scout message generation, it becomes possible to utilize the most suitable AI for the purpose of each process, enabling more advanced automation and efficiency. In addition, combining multiple AIs ensures flexibility and versatility that cannot be achieved with a single AI. [Brief explanation of the drawing]

[0014] [Figure 1] This figure shows an example configuration of a recruitment support system according to an embodiment of the present invention. [Figure 2] This figure shows an example configuration of a recruitment support server according to an embodiment of the present invention. [Figure 3] This flowchart shows the processing flow of the recruitment support server according to an embodiment of the present invention. [Modes for carrying out the invention]

[0015] An embodiment of the present invention will be described by listing its contents. The present invention is configured as follows. [Item 1] A prompt generation unit that generates a prompt for employment determination of the company in the generation AI by inputting recruitment information of the company, An employment determination unit that causes the generation AI to generate an employment determination result of the applicant by inputting applicant information and the employment determination prompt, A primary scout sentence generation unit that causes the generation AI to generate a primary scout sentence for the applicant by inputting the applicant information and the employment determination result, A final scout sentence generation unit that causes the generation AI to generate a final scout sentence by inputting a transmission criterion and the scout sentence text, comprising: An employment support server. [Item 2] The employment support server according to Item 1, further comprising an employment determination correction unit that causes the generation AI to correct the employment determination result by inputting the employment determination result, an inclusion criterion for allowing employment of the applicant, and an exclusion criterion for excluding employment of the applicant, The primary scout sentence generation unit generates the next scout sentence by inputting the applicant information and the corrected employment determination result. An employment support server. [Item 3] The employment support server according to Item 1, The primary scout sentence generation unit generates the primary scout sentence that refers to the content of the applicant information. An employment support server. [Item 4] The employment support server according to Item 1, The final scout sentence generation unit evaluates the primary scout sentence syntactically and semantically to generate a final scout sentence. An employment support server. [Item 5] A prompt generation unit that generates a prompt for employment determination of the company in the generation AI by inputting recruitment information of the company, A recruitment determination unit that, by inputting applicant information and the recruitment determination prompt, causes the generating AI to generate a recruitment determination result for the applicant, A primary scouting message generation unit that, by inputting the applicant information and the hiring decision result, causes the generating AI to generate a primary scouting message for the applicant, The system includes a final scout message generation unit that, upon input of the transmission criteria and the draft scout message, causes the generating AI to generate the final scout message. Recruitment support system. [Item 6] The process involves inputting company recruitment information to generate prompts for hiring decisions for that company using a generation AI, and The steps include: inputting applicant information and the aforementioned hiring decision prompt to cause the generating AI to generate a hiring decision result for the applicant; The steps include: inputting the applicant information and the hiring decision result to cause the generating AI to generate a first-round recruitment message for the applicant; The process includes the step of causing the generating AI to generate a final scout message by inputting the transmission criteria and the scout message draft, Recruitment support methods.

[0016] <Details of the embodiment> Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[0017] <Overview> As shown in Figure 1, the recruitment support system SY of the present invention includes a user terminal UT and a recruitment support server RSS. Users access the recruitment support server via the internet using the user terminal (UT). The user terminal can be any device that can connect to the internet, such as a PC, smartphone, or tablet. The recruitment support server is connected to the internet and accepts requests from the user terminal.

[0018] As shown in Figure 2, the present invention relates to a recruitment support server that automates and streamlines recruitment operations using generation AI. The recruitment support server of the present invention comprises a prompt generation unit that generates prompts for recruitment decisions from a company's recruitment information, a recruitment decision unit that generates recruitment decision results from applicant information and recruitment decision prompts, a recruitment decision correction unit that corrects the recruitment decision results from the recruitment decision unit, a primary scout message generation unit that generates a scout message from applicant information and recruitment decision results, and a final scout message generation unit that generates a final scout message from transmission criteria and a draft scout message.

[0019] The prompt generation unit receives company recruitment information as input and generates a prompt for hiring decisions using the first generation AI. The hiring decision unit receives applicant information and the hiring decision prompt as input and generates a hiring decision result using the second generation AI. The hiring decision correction unit has the third generation AI correct the hiring decision result. The first scout message generation unit receives applicant information and the hiring decision result as input and generates a first scout message using the third generation AI. The final scout message generation unit receives the transmission criteria and the first scout message as input and generates a final scout message using the fourth generation AI.

[0020] According to the recruitment support server of the present invention, by automating and streamlining each stage of the recruitment process using generation AI, the burden on recruiters can be significantly reduced, and the quality and speed of recruitment can be improved. Furthermore, by combining multiple generation AIs, it is possible to utilize the optimal AI tailored to the purpose of each stage, enabling more advanced automation and efficiency.

[0021] According to the present invention, in particular, the enormous amount of time and effort spent on selecting applicants has increased significantly with the rise in the number of applicants. This includes reviewing resumes and application forms, conducting interviews, and so on. The present invention significantly reduces the time and effort spent on the selection process by automatically generating hiring decision results from applicant information and hiring decision prompts.

[0022] <Description of each functional part> The following explanation of each functional part will be given with reference to Figure 3.

[0023] <Prompt generation unit: First generation AI> The prompt generation unit is responsible for automatically generating prompts for hiring decisions (prompts that generate hiring criteria) from companies' recruitment information. The prompt generation unit uses the first generation AI to analyze information such as job title, work content, required skills, and experience included in the recruitment information, and generates prompts optimized for hiring decisions based on this information.

[0024] The recruitment information input to the prompt generation unit is provided in text or template format and structured using natural language processing techniques as needed. The first generating AI is based on Large Language Models (LLMs) and may acquire knowledge and know-how related to recruitment operations through fine-tuning (the "generating AI" described below is similar). This enables the generation of prompts that accurately capture the key points of the recruitment information and include all the necessary information for hiring decisions.

[0025] In addition, RAG (Retrieval-Augmented Generation) can be used to dynamically incorporate external knowledge such as company-specific recruitment policies, past recruitment performance data, and industry trends (the same applies to the "Generative AI" described below).

[0026] Specifically, when recruitment information is entered, the first generation AI uses RAG to search for and acquire relevant external knowledge. The acquired external knowledge is dynamically incorporated into the generation AI's input and used to generate prompts. This makes it possible to generate highly customized prompts that reflect company-specific requirements and the latest industry trends. By introducing RAG, the prompt generation unit can flexibly respond not only to recruitment information but also to changes in the company's recruitment strategy and market environment. Furthermore, the external knowledge incorporated by RAG can be continuously updated and expanded through interaction with recruiters.

[0027] The prompts generated in this way may be made available for review and modification by recruiters. The prompt generation unit also has a function to continuously improve the quality of the prompts through interaction with recruiters. Furthermore, by learning each company's recruitment policies and past recruitment performance data, the prompt generation unit can also generate prompts customized to the specific recruitment criteria of each company.

[0028] The prompts generated by the prompt generation unit are not simply summaries of recruitment information, but include questions and indicators for a multifaceted evaluation of the applicant's abilities and suitability. This allows the subsequent hiring decision unit to make a more accurate decision on whether or not to hire an applicant based on their information. The prompt generation unit plays a crucial role in improving the accuracy and ensuring fairness of the hiring decision.

[0029] Examples of possible inputs, such as recruitment information, and the generated prompts are shown. <Example of recruitment information> [Recruitment Information] • Job Title: Software Engineer • Job description: Development, testing, and maintenance of in-house products Required skills: Java, Python, AWS • Experience: 3+ years of development experience • Educational background: Bachelor's degree or higher [Company-specific recruitment policies] • Emphasis on understanding and enthusiasm for our own products • Evaluate teamwork and communication skills. • Expect continuous learning and growth [Past hiring data] • Candidates with 5 or more years of practical Java experience have a higher retention rate. • Candidates with AWS knowledge tend to perform better after joining the company. [Industry Trends] • Demand for personnel with knowledge in the fields of AI and machine learning is increasing. • There is a need for people who can adapt to remote work. <Example of generated prompt> Please evaluate applicants' experience, skills, and mindset from multiple perspectives, focusing on the following points, and determine their alignment with our company's recruitment policies and product development direction. In particular, we place importance on practical Java experience, AWS knowledge, understanding of and enthusiasm for our products, and a continuous learning attitude. We also consider knowledge of AI and machine learning, as well as adaptability to remote work. 1. Practical experience and proficiency in Java, Python, and AWS. We particularly favor candidates with 5 or more years of experience in Java. We value knowledge and experience using AWS. 2. Understanding of and enthusiasm for our own products Understanding of the functions and features of our own products Motivation and concrete ideas for developing and improving our own products 3. Teamwork and communication skills Team roles and contributions in past projects Teamwork and communication style with other members 4. A commitment to continuous learning and growth Interest in new technologies and trends, and a desire to learn. Past self-development experiences and future career plans 5. Knowledge and experience in the field of AI and machine learning Understanding the fundamentals and application examples of AI and machine learning. Experience in developing and implementing machine learning models 6. Adaptability to remote work Remote work experience and results The ability to perform tasks autonomously and the ingenuity of communication methods.

[0030] <Recruitment Judgment Unit: Second Generation AI> The hiring decision unit is responsible for automatically determining whether an applicant is suitable for employment as a first-stage hiring decision based on applicant information and hiring decision prompts. The hiring decision unit uses a second generation AI to analyze information such as the applicant's career history, skills, experience, and educational background, and comprehensively evaluates the applicant's suitability based on the hiring decision prompts generated by the prompt generation unit.

[0031] Applicant information entered into the hiring decision-making unit is extracted from documents submitted by applicants, such as resumes, work histories, cover letters, and portfolios. This information is structured using natural language processing technology and used as input data for a second generative AI. The second generative AI is based on a large-scale language model and has acquired knowledge and know-how related to hiring operations, as well as applicant evaluation methods and logic through fine-tuning.

[0032] In the recruitment evaluation unit, a second generation AI takes applicant information and prompts as input and analyzes the applicant's suitability from multiple perspectives. The prompts reflect the skills and experience required for the advertised position, the company's recruitment policies, and past recruitment performance data, allowing the second generation AI to evaluate applicants from these viewpoints. For example, it comprehensively determines whether the applicant's skills and experience match the recruitment requirements, whether the applicant's mindset and values ​​are compatible with the company culture, and whether the applicant's past achievements meet the company's expectations.

[0033] The second generation AI performs an initial hiring decision on each applicant's information using the hiring decision prompts generated above. The decision may include individual evaluations of individual skills and mindset items, as well as an overall evaluation based on the scores of the individual evaluations. It may also include setting a cutoff score or adding prompts for specific reasons that should lead to exclusion from hiring.

[0034] The hiring decision unit can evaluate an applicant's suitability on a five-point scale. For example, it can set five levels such as "5: Excellent," "4: Good," "3: Average," "2: Somewhat Poor," and "1: Poor." The second generation AI assigns a five-point score to each evaluation item based on the applicant information and hiring decision prompts. This allows for a relative understanding and comparison of the applicant's strengths and weaknesses.

[0035] The hiring decision unit can also evaluate applicants' suitability on a three-point scale: A, B, and C. For example, a simple classification can be set, such as "A: Excellent," "B: Some ability," and "C: Insufficient ability." The second generation AI then determines whether each evaluation item is A, B, or C based on the applicant information and hiring decision prompts. This method has the advantage of having clear evaluation criteria and making it easy to decide whether or not to hire an applicant.

[0036] The hiring decision unit can also quantify an applicant's suitability on a scale of 0 to 100 points. The second generation AI assigns a score from 0 to 100 points to each evaluation item based on the applicant information and hiring decision prompts. For example, evaluation criteria can be set according to the score, such as "90 points or higher: Excellent," "70-89 points: Good," "50-69 points: Average," "30-49 points: Slightly below average," and "29 points or lower: Below average." This method has the advantage of allowing for a quantitative understanding of the applicant's suitability and enabling objective comparison.

[0037] The recruitment support server can also combine and switch between these evaluation formats. For example, it can use a 5-point scale for the first selection round and A, B, C for the second selection round. It is also possible to convert and display the evaluation results from the recruitment decision unit using other evaluation scales.

[0038] The second generation AI evaluates applicants, providing detailed feedback that includes not only a decision on whether or not to hire the applicant, but also their strengths, weaknesses, and future potential. This feedback supports the hiring manager's decision-making and is also used in communication with applicants. The hiring decision unit ensures transparency and fairness in the hiring process by presenting the applicant evaluation results in an explainable format.

[0039] Furthermore, the hiring decision unit has the ability to accumulate applicant evaluation results and improve itself through machine learning. The accumulated data is used to fine-tune the second generation AI, continuously improving the accuracy and consistency of evaluations. In addition, by incorporating feedback from recruiters, the hiring decision unit achieves a more sophisticated applicant evaluation by combining human experience and machine learning capabilities.

[0040] The second generation AI evaluation could also detect and determine potential candidates by using positive prompts such as "possesses / has a strong startup mindset" rather than using negative prompts such as "lacks the mindset of someone with a traditional corporate background."

[0041] <Recruitment Judgment Correction Unit: Third Generation AI> The hiring decision correction unit is responsible for correcting the hiring decision result using inclusion and exclusion criteria in order to bring the hiring decision result as close as possible to the user's hiring decision criteria. The hiring decision correction unit uses a third generation AI to receive the hiring decision result generated by the hiring decision unit and the inclusion and exclusion criteria defined by the user as input, and adjusts the hiring decision result by comprehensively considering them.

[0042] Inclusion criteria are requirements that actively encourage the hiring of candidates, and refer to abstract evaluation items that are easy for users to understand during the initial setup. For example, sales experience or marketing experience would fall into this category. Candidates who meet the inclusion criteria will have their hiring decisions adjusted to receive a higher evaluation.

[0043] On the other hand, exclusion criteria are requirements that make candidates less likely to be hired, and refer to specific evaluation items that tend to become apparent during the operational phase. For example, this includes conditions such as excluding inside sales experience from sales experience. Candidates who violate the exclusion criteria will have their hiring decisions adjusted to receive a lower evaluation.

[0044] The third generation AI is based on a large-scale language model and may acquire knowledge and know-how related to recruitment operations, as well as fine-tuned methods for interpreting and applying inclusion and exclusion criteria. The third generation AI analyzes candidate resumes and recruitment decisions, and calculates correction values ​​for the recruitment decisions based on the inclusion and exclusion criteria. These correction values ​​are reflected in the scores and evaluation comments of the recruitment decisions, generating more detailed decisions that align with the user's criteria.

[0045] One of the key features of the hiring decision correction unit is the flexible definition and updating of inclusion and exclusion criteria. Users define inclusion and exclusion criteria during initial setup, but can modify and add to them at any time based on insights gained through system use and evolving hiring policies. The third generation AI learns from user-updated criteria and constantly corrects hiring decisions based on the latest criteria. This allows the hiring decision correction unit to continuously adapt to the user's hiring criteria.

[0046] Furthermore, the hiring decision correction unit also has the function of verifying the validity and effectiveness of hiring decisions by accumulating the application history of inclusion and exclusion criteria and analyzing this data. For example, it can track the post-hiring performance of candidates who meet the inclusion criteria and the hiring status of candidates who violate the exclusion criteria, and evaluate the effectiveness of the criteria. The results of this analysis are provided as feedback to the user and used to continuously improve the hiring decisions criteria.

[0047] In the embodiment described above, the inclusion criteria and exclusion criteria are input simultaneously to the third generating AI in order to improve the accuracy of the selection decision. However, the inclusion criteria may be input to the second generating AI, and the exclusion criteria may be input to the third generating AI as an exception.

[0048] Similar to the evaluation by the second generation AI, the evaluation by the third generation AI may also detect and determine candidates for employment based on positive prompts rather than negative prompts.

[0049] <Primary Scout Message Generation Unit: 4th Generation AI> The Primary Recruitment Message Generation Unit is responsible for automatically generating primary recruitment messages to send to applicants based on applicant information and hiring decision results. The Primary Recruitment Message Generation Unit uses a fourth generation AI to analyze information such as the applicant's background, skills, experience, and evaluation results from the hiring decision unit, and generates personalized recruitment messages that highlight the applicant's characteristics and strengths.

[0050] The applicant information entered into the primary scouting message generation unit includes not only the resumes and other documents submitted by the applicants, but also the evaluation results of the applicants generated by the hiring decision unit. This information is used as input data for the fourth generation AI. The fourth generation AI is based on a large-scale language model and has acquired knowledge and know-how related to recruitment operations, as well as effective scouting message creation techniques and templates through fine tuning.

[0051] In the first stage of the recruitment message generation process, the fourth generation AI takes applicant information and hiring decision results as input and generates a recruitment message tailored to the applicant. The recruitment message is generated based on a recruitment format. The recruitment message includes specific and attractive expressions highlighting the applicant's strengths in skills and experience, their suitability for the company's hiring policies and work environment, and their future potential. The recruitment message also includes information on the next steps, such as interviews and the selection process, as well as the company's contact information.

[0052] The fourth generation of AI-generated recruitment messages aims not merely to list information, but to attract applicants' interest and increase their engagement with the company. Therefore, the messages are crafted with a narrative structure, incorporating messages tailored to the applicant's personality and career plans, highlighting the company's appeal and growth opportunities, and highlighting common ground and connections with the applicant. Furthermore, the tone and style of the messages are optimized to match the applicant's attributes and the company's culture.

[0053] The recruitment messages generated by the first-stage recruitment message generation unit are reviewed and revised by the final recruitment message generation unit before being sent to applicants. Applicant responses to the sent recruitment messages are tracked and analyzed, and used as training data for the fourth generation AI. This allows the first-stage recruitment message generation unit to continuously improve the effectiveness of its recruitment messages and enhance engagement with applicants.

[0054] Furthermore, the initial scout message generation unit also includes a function to customize and refine the scout message through interaction with recruiters. Recruiters can review the generated scout message and make corrections or additions as needed.

[0055] In a recruitment support server according to one embodiment of the present invention, the primary scout message generation unit is configured to generate a primary scout message that refers to the content of the applicant information.

[0056] Specifically, the initial recruitment message generation unit analyzes the applicant's background, skills, qualifications, experience, and motivations included in the applicant information, and then generates an initial recruitment message based on this analysis. This allows for the creation of personalized recruitment messages that accurately capture the individual characteristics and strengths of each applicant.

[0057] For example, if the applicant information includes specific details such as "5 years of experience in web application development" or "a master's degree in machine learning," the initial recruitment message generation unit can incorporate this information into the recruitment message as follows. "Your five years of experience in front-end and back-end development for web applications will be invaluable in our service development. Furthermore, your master's degree in machine learning means you can contribute to a wide range of areas, including collaboration with our data science team."

[0058] By directly referring to applicant information in this way, companies can fairly evaluate applicants' experience and skills and clearly communicate their expectations. Furthermore, applicants can feel that their background has been thoroughly considered before being scouted, which can increase their positive feelings towards the company and their desire to join.

[0059] For the primary scout message generation unit to achieve this functionality, the fourth generation AI needs to be able to accurately understand the content of applicant information and weave it into natural-sounding sentences. The fourth generation AI addresses this challenge by combining advanced natural language processing technology based on a large-scale language model with expertise in recruitment operations.

[0060] <Final Scout Message Generation Unit: 5th Generation AI> The final recruitment message generation unit uses a fifth generation AI to evaluate the stylistic aspects of the generated text, such as whether it contains inappropriate words or is of appropriate length, by referring to the initial recruitment message generated by the first recruitment message generation unit and the format check information. It is responsible for generating the final recruitment message to be sent to applicants. The final recruitment message generation unit uses a fifth generation AI to check and revise the content and expression of the recruitment message draft to ensure that it is linguistically appropriate according to the sending standards, and to produce a recruitment message that is free from legal and ethical issues and in line with the company's recruitment policy.

[0061] The input to the final recruitment message generation unit consists of the recruitment message draft generated by the primary recruitment message generation unit and the sending criteria set by the company. The sending criteria define the information that should be included in the recruitment message (company information, job openings, selection process, etc.) and the expressions that should be avoided (discriminatory language, exaggerated language, information that violates confidentiality, etc.). The fifth generation AI is based on a large-scale language model and may acquire the ability to check recruitment messages from the perspective of compliance and risk management, in addition to knowledge and know-how related to recruitment operations, through fine tuning.

[0062] In the final scout message generation section, a fifth generation AI scrutinizes the draft scout message against the transmission criteria and makes any necessary corrections or additions. Specifically, it checks whether the draft scout message includes all the essential information defined in the transmission criteria, whether there are any inappropriate expressions or misleading phrases, and whether personal information is handled appropriately. The overall structure and expression of the scout message are also reviewed and corrected from the perspectives of clarity, politeness, and persuasiveness.

[0063] The final recruitment message generation unit incorporates not only automatic correction of recruitment messages by the fifth generation AI, but also a review and approval process by recruiters. The generated recruitment messages are presented to recruiters for final checks of content and expression. Recruiters revise the messages as needed and grant permission for submission. By incorporating this human oversight process, a high level of quality and appropriateness of recruitment messages can be guaranteed.

[0064] The final recruitment message generation unit sends the generated recruitment message to the applicant. The sent recruitment messages are tracked and analyzed along with the applicant's response and the progress of the selection process, and this data is used as training data for the fifth generation AI. This allows the final recruitment message generation unit to continuously improve the quality of the recruitment messages and their impact on the recruitment process.

[0065] Furthermore, the final recruitment message generation unit also has the functionality to flexibly respond to changes in corporate recruitment policies and legal / social requirements through setting and updating sending criteria and communicating with recruiters. Sending criteria are regularly reviewed and updated as needed in cooperation with recruiters and the HR department. In addition, the final recruitment message generation unit receives feedback from recruiters and improves the logic and criteria for refining recruitment messages.

[0066] In a recruitment support server according to one embodiment of the present invention, the final scout message generation unit is configured to generate a final scout message by syntactically and semantically evaluating the initial scout message.

[0067] Specifically, the final scout message generation unit receives the primary scout message generated by the primary scout message generation unit as input and analyzes the structure and semantic content of the message in detail. This analysis is performed by a fifth generation AI, and advanced natural language processing techniques are applied. Syntactic evaluation checks the grammatical structure, punctuation, and word choice of the initial recruitment letter to ensure accuracy and readability. For example, the correspondence between subject and predicate, the use of particles, and the presence of complex sentence structures are checked, and revisions are made as needed. This improves the quality of the recruitment letter as a piece of writing and enhances the impression it makes on applicants.

[0068] In semantic evaluation, the content of the initial recruitment message is carefully examined to determine whether it contains all the necessary information, whether it is logically consistent, and whether it aligns with the company's recruitment policies. For example, it is checked whether the assessment of the applicant's experience and skills is appropriate, whether the description of the company's business and work environment is attractive and persuasive, and whether the next steps are clearly indicated. This improves the overall quality of the recruitment message and helps to gain the applicant's understanding and empathy.

[0069] The final scout message generation unit, based on the results of these syntactic and semantic evaluations, makes necessary corrections and additions to the initial scout message to generate the final message. In this process, the fifth generation AI utilizes a large-scale language model to automatically refine and polish the text. It also checks against the submission standards, performing checks from a compliance and risk management perspective.

[0070] For the final recruitment message generation unit to achieve this functionality, the fifth generation AI needs to have a deep understanding of the structure and meaning of natural language and the ability to evaluate and improve the quality of the text from multiple perspectives. The fifth generation AI addresses this challenge by acquiring knowledge of linguistics and writing through deep learning and applying it to the context of recruitment work.

[0071] In other words, the quality assessment of scout messages by the fifth generation AI (hereinafter referred to as "CheckerAI") is carried out in the following procedure and is continuously improved. First, the scout messages that CheckerAI has quality-checked are also evaluated in parallel by a professional reviewer (human). The reviewer judges the appropriateness of the content and expression of the scout messages based on their own experience, common sense, and ethical values. Then, the CheckerAI judgment result and the human judgment result are compared and the degree of agreement is measured. Next, the F-measure is used to evaluate the validity of the CheckerAI judgment result. The F-measure is the harmonic mean of precision and recall, and is an indicator that shows the balance between the two. In this case, the F0.5 value is used to give emphasis to precision. This is an indicator that considers precision with twice the weight of recall. The higher the F0.5 value, the more scout messages that CheckerAI has judged to be "sendable" that are actually safe to send.

[0072] Next, we extract the scout messages that CheckerAI judged as "sendable" but that a human judged as "not sendable," and perform error analysis. These scout messages are likely to contain problems that CheckerAI missed. We analyze the causes of the errors and identify problems with CheckerAI's judgment logic and dataset. For example, we might discover that CheckerAI is failing to detect inappropriate expressions in certain patterns, or that it is misjudging the usage of certain industry terms. Then, we improve CheckerAI to solve the problems identified in the error analysis. This includes correcting the judgment logic, adding or modifying the dataset, and retraining the machine learning model. Finally, we perform another quality check of the scout messages using the improved CheckerAI.

[0073] Finally, the improved CheckerAI judgment results are compared again with human judgment results, and the F0.5 value is calculated. If the F0.5 value has improved, it is determined that the improvement of CheckerAI has been successful. On the other hand, if the F0.5 value has not improved sufficiently, error analysis is performed again to aim for further improvement. By continuously implementing this cycle, the accuracy of CheckerAI's judgments is gradually increased.

[0074] Through the process described above, CheckerAI continuously improves itself by incorporating human judgment, thereby continuously enhancing the accuracy of quality control for recruitment messages. This enables recruiters to automatically generate consistently high-quality recruitment messages, achieving both operational efficiency and quality assurance. CheckerAI's improvement cycle is one of the key features of this invention and can be said to be an indispensable function for realizing the automation and optimization of recruitment communication through collaboration between AI and humans.

[0075] <Processing flow> Referring to Figure 3, the sequence of processes in the recruitment support system SY of the present invention will be explained. First, the recruitment criteria (CRY) are passed to the first generating AI, which generates an AJP for recruitment decision based on the recruitment criteria. Next, the applicant information (AST) and the AJP for recruitment decision are passed to the second generating AI, which generates an AJR for initial recruitment decision from the applicant information (AST) and the AJP for recruitment decision.

[0076] Next, the initial selection result (AJR) is passed to a third generating AI, which then outputs the final selection result (FJR) based on the initial selection result (AJR). Subsequently, a fourth generating AI generates the initial scout message (SMS) based on the selection result (FJR), the scout format, and the format check information. The initial scout message is then further processed into the final scout message (SMC) by a fifth generating AI.

[0077] The final scouting message may be generated along with a report on the suitability assessment results.

[0078] <Variation> A modified example according to the embodiment of the present invention will be described. <Example 1> The recruitment support server of the present invention can be expanded to support multiple job types and recruitment positions. In this case, the prompt generation unit generates recruitment decision prompts specific to each job type and position. Furthermore, the first-stage scout message generation unit and the final scout message generation unit generate scout messages that incorporate different appeal points to applicants for each job type and position. This enables detailed recruitment activities that meet diverse talent needs.

[0079] <Modification 2> The recruitment support server can be enhanced with interactive communication features for applicants. For example, a chatbot or messaging system can be implemented to automatically respond to applicant inquiries and guide them through the selection process. To achieve this functionality, a fifth-generation generative AI is introduced to understand applicants' statements and generate appropriate responses. This enhances engagement with applicants and improves the applicant experience.

[0080] <Variation 3> By linking the recruitment support server with external data sources, the accuracy of recruitment decisions can be improved. For example, information from social media and job search websites can be acquired and used to supplementarily analyze applicants' work history, skills, and reputation. By incorporating this information into the recruitment decision-making unit, applicant suitability can be evaluated from a more multifaceted perspective. Furthermore, utilizing external data such as industry trends and the recruitment status of competitors can help optimize recruitment strategies.

[0081] <Modification 4> By adding collaboration features with recruiters to the recruitment support server, human expertise and AI processing power can be effectively combined. For example, recruiters can revise and approve generated recruitment messages, and comment on AI-generated hiring decisions. Furthermore, features for information sharing and task management among recruiters will be provided to streamline the entire recruitment process. This will enable AI and humans to collaborate and advance recruitment activities while leveraging each other's strengths.

[0082] <Modification 5> The recruitment support server can be enhanced with a function to analyze data generated at each stage of the recruitment process and create reports. For example, it can visualize and provide information such as applicant attributes, evaluation statistics, the status of recruitment message transmissions and applicant responses, and the progress of the recruitment process to recruiters and HR departments. By utilizing these reports, it is possible to measure the effectiveness of recruitment activities, identify challenges, and implement a PDCA cycle. Furthermore, it can be used for continuous improvement of recruitment operations through comparative analysis over time.

[0083] <Generation AI> In this embodiment, if the first to fourth generating AIs are different, each generating AI is configured as an individual module with its own unique functions and characteristics. The first generating AI is responsible for prompt generation, the second for hiring decisions, the third for generating initial scout messages, and the fourth generating AI (CheckerAI) for generating final scout messages and quality checks. Each generating AI implements specialized functions using its own architecture, dataset, and learning algorithm. For example, the first generating AI could use a language model specialized in knowledge of hiring criteria, and the second generating AI could incorporate a machine learning model to evaluate applicant characteristics. The advantage of this configuration is that each process can utilize a generating AI optimized for that process. By independently developing and improving each generating AI, the accuracy and efficiency of each process can be increased. On the other hand, a disadvantage is that ensuring coordination and consistency between the generating AIs becomes a challenge.

[0084] On the other hand, if the first to fourth generative AIs are the same, a single general-purpose generative AI will handle all stages of the recruitment process. This generative AI must be based on a large-scale language model and possess knowledge and skills related to all aspects of recruitment. Because the same generative AI is used in all stages, coordination between stages is smooth, and consistency of the generated products is easier to maintain. Furthermore, development and operational costs can be reduced. However, it is difficult to fully specialize for the specific requirements of each stage, and the accuracy of each function may be slightly inferior compared to using individual generative AIs.

[0085] The choice of configuration must be determined by considering factors such as the scale of the recruitment support system (SY), the required accuracy, and development resources. A configuration using individual generative AIs is suitable for large-scale recruitment support systems (SY) or when a high level of expertise is required. On the other hand, a configuration using a single generative AI is suitable for small to medium-sized recruitment support systems (SY) or when development costs need to be kept down.

[0086] Furthermore, as an intermediate configuration between the two, a hybrid configuration is also conceivable, in which the same generation AI is used in some processes and separate generation AIs are used in other processes. For example, this configuration could use two generation AIs: a "recruitment judgment AI" that combines the functions of the first and second generation AIs, and a "scout message generation AI" that combines the functions of the third and fourth generation AIs.

[0087] <Example Hardware Configuration> Each of the functional blocks described above can be configured using hardware, a DSP (Digital Signal Processor), or software, for example, provided in a server device (terminal device). For example, when configured using software, each functional block is actually configured using a computer's CPU, RAM, ROM, etc., and is realized by the operation of programs stored on recording media such as RAM, ROM, hard disk, or semiconductor memory.

[0088] The embodiments described above are merely illustrative to facilitate understanding of the present invention and are not intended to limit its interpretation. The present invention can be modified and improved without departing from its spirit, and it goes without saying that the present invention includes equivalents thereof. [Explanation of symbols]

[0089] UT User Terminal RSS Recruitment Support Server

Claims

1. A prompt generation unit that, by inputting recruitment information from a company, causes the AI ​​to generate prompts for the company's hiring decision, A recruitment determination unit that, by inputting applicant information and the recruitment determination prompt, causes the generating AI to generate a recruitment determination result for the applicant, A primary scouting message generation unit that, by inputting the applicant information and the hiring decision result, causes the generating AI to generate a primary scouting message for the applicant, The system includes a final scout message generation unit that, upon inputting the transmission criteria and the draft scout message, causes the generating AI to generate the final scout message. Recruitment support server.

2. The system further includes a recruitment determination correction unit that corrects the recruitment determination result by inputting the recruitment determination result, inclusion criteria for allowing the applicant to be hired, and exclusion criteria for excluding the applicant from being hired. The first-stage scouting message generation unit generates the second-stage scouting message by inputting the applicant information and the corrected hiring decision result. Recruitment support server.

3. The recruitment support server according to claim 1, The aforementioned primary scouting unit generates the primary scouting message that refers to the content of the applicant information. Recruitment support server.

4. The recruitment support server according to claim 1, The final scout sentence generation unit generates the final scout sentence by syntactically and semantically evaluating the primary scout sentence. Recruitment support server.

5. A prompt generation unit that, by inputting recruitment information from a company, causes the AI ​​to generate prompts for the company's hiring decision, A recruitment determination unit that, by inputting applicant information and the recruitment determination prompt, causes the generating AI to generate a recruitment determination result for the applicant, A primary scouting message generation unit that, by inputting the applicant information and the hiring decision result, causes the generating AI to generate a primary scouting message for the applicant, The system includes a final scout message generation unit that, upon inputting the transmission criteria and the draft scout message, causes the generating AI to generate the final scout message. Recruitment support system.

6. The process involves inputting company recruitment information to generate prompts for hiring decisions for that company using a generation AI, and The process involves inputting applicant information and the aforementioned hiring decision prompt to cause the generating AI to generate a hiring decision result for the applicant, The steps include: inputting the applicant information and the hiring decision result to cause the generating AI to generate a first-round recruitment message for the applicant; The process includes the step of causing the generating AI to generate a final scout message by inputting the transmission criteria and the scout message draft, Recruitment support methods.

Images