Program, information processing method, and information processing apparatus.

Abstract

We provide programs and other tools that can effectively support the approval of expense claims. [Solution] The program obtains expense application information related to the expense application submitted by the expense applicant, and inputs a prompt to the language model 40 that includes the expense application information and expense regulation information indicating the approval conditions for expenses in the organization to which the expense applicant belongs, thereby causing the computer to execute a process to generate a check result of whether or not to approve the expense application and to output the check result.

Description

Technical Field

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

Background Art

[0002] There is a technology for assisting the approval of expense applications (for example, Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In one aspect, an object is to provide a program or the like that can suitably assist the approval of expense applications.

Means for Solving the Problems

[0005] The program obtains expense application information regarding an expense application applied for by an expense applicant, and inputs a prompt including the expense application information and expense regulation information indicating the approval conditions of expenses in the organization to which the expense applicant belongs into a language model, thereby generating a check result as to whether to approve the expense application, The expense application information and the check results are output to an external operator who will re-check the expense application, and the operator will input the re-check results of the expense application, and the expense application information, the check result , and the results of the recheck mentioned above and addressed to the accounting staff within the organization causes a computer to execute a process of outputting.

Effects of the Invention

[0006] In one aspect, it is possible to suitably assist the approval of expense applications.

Brief Description of the Drawings

[0007] [Figure 1] It is a diagram showing a configuration example of an expense approval system. [Figure 2] This is a block diagram showing an example server configuration. [Figure 3] This figure shows an example of the record layout for the organization database, user database, and application database. [Figure 4] This is a diagram illustrating the outline of the embodiment. [Figure 5] This figure shows an example of a report display screen. [Figure 6] This figure shows an example of a timeline screen. [Figure 7] This flowchart shows an example of the processing steps performed by the server. [Modes for carrying out the invention]

[0008] The present invention will be described in detail below with reference to the drawings illustrating its embodiments. (Embodiment) Figure 1 shows an example of the configuration of an expense approval system. In this embodiment, an expense approval system is described in which an LLM (Large Language Model) 40 functions as an AI (Artificial Intelligence) agent to check whether or not to approve expense applications, and the LLM 40 automatically approves expense applications from applicants. The expense approval system includes an information processing device 1, an applicant terminal 2, an accounting staff terminal 3, a generation server 4, and an operator terminal 5. Each device is connected to the others via a network N such as the Internet.

[0009] Information processing device 1 is an information processing device capable of various information processing and information transmission / reception, such as a server computer or a personal computer. In this embodiment, information processing device 1 is assumed to be a server computer, and for simplicity, it will be referred to as server 1 below. Server 1 functions as the management device of this system and causes LLM40 to check whether or not to approve expense applications from expense applicants.

[0010] Applicant terminal 2 is a terminal device used by the expense applicant, such as a smartphone, tablet, or personal computer. For example, server 1 obtains expense application information related to the expense application submitted by the expense applicant from applicant terminal 2 and creates a prompt containing the obtained expense application information. Server 1 inputs the created prompt into LLM40 to generate a check result indicating whether or not to approve the expense application, and outputs the generated check result to applicant terminal 2, etc.

[0011] The accounting staff terminal 3 is a terminal device used by the accounting staff of the organization (e.g., company) to which the expense applicant belongs, and is, for example, a personal computer. Server 1 outputs the results of the expense application check by LLM40 to the accounting staff terminal 3 (see Figures 5 and 6), and accepts final approval from the accounting staff terminal 3 whether or not to approve the expense application.

[0012] The generation server 4 is a server computer that generates text data using LLM40. LLM40 is a machine learning model that generates text when it receives prompt input, and is a large-scale language model such as GPT (Generative Pretrained Transformer: registered trademark). Note that LLM40 may be a model other than GPT (e.g., BERT (Bidirectional Encoder Representations from Transformer), T5, etc.). The generation server 4 inputs the prompt received from the server into LLM40 and generates text representing the result of whether or not to approve the expense claim.

[0013] Operator terminal 5 is a terminal device used by an operator outside the organization to which the expense claimant belongs (for example, an employee of the company operating this system), such as a personal computer. In this system, as an option for the expense claim checking service, in addition to automatic checking by LLM40, manual checking by an operator is also performed. If the option for manual checking by an operator is set, server 1 outputs the expense claim information and the LLM40 check results to operator terminal 5 before outputting the check results to accounting staff terminal 3, and accepts input of re-check results from operator terminal 5. Server 1 outputs the LLM40 check results and the operator's re-check results to accounting staff terminal 3.

[0014] As described above, Server 1 generates a check result of whether or not to approve the expense application by inputting a prompt containing expense application information related to the expense application submitted by the expense applicant to LLM40. In particular, in this embodiment, Server 1 provides LLM40 with expense regulation information indicating the approval conditions for expenses in the organization to which the expense applicant belongs, i.e., internal regulations regarding expenses, as a prompt, causing LLM40 to determine whether or not the expense application conforms to the internal regulations. This makes it possible to check whether or not to approve the expense application appropriately, taking into account the different internal regulations of each organization (company).

[0015] Figure 2 is a block diagram showing an example configuration of Server 1. Server 1 comprises a control unit 11, a main memory unit 12, a communication unit 13, and an auxiliary memory unit 14. The control unit 11 is a processor such as one or more CPUs (Central Processing Units), MPUs (Micro-Processing Units), GPUs (Graphics Processing Units), etc., and performs various information processes by reading and executing the program P stored in the auxiliary storage unit 14. The main memory unit 12 is a temporary storage area such as SRAM (Static Random Access Memory), DRAM (Dynamic Random Access Memory), etc., and temporarily stores the data necessary for the control unit 11 to execute arithmetic processing. The communication unit 13 is a communication module for performing communication-related processes, and transmits and receives information to and from the outside.

[0016] The auxiliary storage unit 14 is a non-volatile storage area such as a hard disk, a large-capacity memory, etc., and stores the program P (program product) and other data necessary for the control unit 11 to execute processing. Further, the auxiliary storage unit 14 stores an expense regulation file 100, an organization DB 141, a user DB 142, and an application DB 143.

[0017] The expense regulation file 100 is a file that describes expense regulation information (company regulations) indicating the approval conditions of expenses in the organization to which the expense applicant belongs, and is a tabular file such as a CSV (Comma-Separated Values) file, for example. The expense regulation file 100 is prepared for each organization (customer company) using this system, for example, or for each organization and each expense item.

[0018] The organization DB 141 is a database that stores information on each organization (customer company) using this system. The user DB 142 is a database that stores information on each individual user (employee of the customer company) using this system. The application DB 143 is a database that stores information related to expense applications received from expense applicants (users).

[0019] The auxiliary storage unit 14 may be an external storage device connected to the server 1. Furthermore, the server 1 may be a multi-computer system consisting of multiple computers, or it may be a virtual machine virtually constructed by software.

[0020] Furthermore, in this embodiment, the server 1 is not limited to the above configuration and may include, for example, an input unit for receiving operation input, a display unit for displaying images, etc. The server 1 may also include a read unit for reading portable storage media 1a such as a CD (Compact Disc)-ROM or DVD (Digital Versatile Disc)-ROM, and may read and execute the program P from the portable storage media 1a. The server 1 may also download and execute the program P from an external source via a network N.

[0021] Figure 3 shows an example of the record layout for Organization DB141, User DB142, and Application DB143.

[0022] Organization DB141 includes columns for Organization ID, Organization Name, Accounting Staff ID, Expense Regulation File, and Operator Options. The Organization ID column stores the Organization ID, which is the identifier for each organization. The Organization Name, Accounting Staff ID, Expense Regulation File, and Operator Options columns each store, in association with the Organization ID, the organization name (company name), the user ID (identifier of the accounting staff), the file name of Expense Regulation File 100 (which contains the organization's expense regulation information), and whether or not the operator has the option to perform a manual check.

[0023] User DB142 includes columns for User ID, Organization, Account Information, and User Information. The User ID column stores the User ID, which is the identifier for each user. The Organization, Account Information, and User Information columns, associated with the User ID, store the Organization ID of the organization to which the user belongs, the user's account information (account name, password, etc.), and user information (user name, department, job title, etc.), respectively.

[0024] Application DB143 includes an Application ID column, a Status column, and an Expense Application Information column. The Application ID column stores the Application ID, which is the identifier for each expense application. The Status column stores the current status of the expense application in association with the Application ID. In this embodiment, expense applications are classified into three statuses: "Approved," "Requires Confirmation," and "Returned." The Expense Application Information column stores various information representing the details of the expense application (e.g., the applicant's User ID, the application date of the expense application, the transaction date of the expense, the amount of the expense, the payee, etc.) in association with the Application ID.

[0025] Figure 4 is a diagram illustrating the overview of the embodiment. The overview of this embodiment will be described based on Figure 4.

[0026] As described above, Server 1 has LLM40 function as an AI agent to check whether or not to approve expense claims, and has LLM40 perform the check (initial approval) of expense claims. Figure 4 conceptually illustrates the expense claim check flow. Depending on the check results from LLM40, Server 1 returns the expense claim received from the expense claimer to the expense claimer, or if it determines that the expense claim is "approved" or "requires confirmation", it outputs it to the accounting staff (via the operator). The following describes the processing performed by Server 1 in accordance with Figure 4.

[0027] Server 1 first receives expense applications from applicant terminal 2. For example, applicant terminal 2 accepts input of information such as the transaction date, expense amount, payee, whether a qualified invoice exists, qualified invoice issuer registration number, purpose of the expense, account (expense item), tax amount (consumption tax amount), non-taxable amount, name of the internal project that required the expense, whether it was an advance payment, destination, start date, end date, and daily allowance if the expense application is for travel expenses, participant names, cost per participant, and expense allocation if the expense application is for meeting expenses, departure point, arrival point, destination, and comments optionally entered by the expense applicant. Applicant terminal 2 sends this information to server 1 along with images of supporting documents (e.g., receipts, invoices, etc.). Server 1 may also read necessary information from the images of supporting documents using OCR (Optical Character Recognition).

[0028] Server 1 adds information about the expense applicant (user ID, username, etc.) and information about the application itself (application date, application ID, etc.) to the information obtained from the applicant terminal 2. As a result, Server 1 obtains expense application information related to the expense application submitted by the expense applicant.

[0029] Server 1 creates a prompt that includes the acquired expense claim information and instructs LLM40 to check whether to approve the expense claim. For example, Server 1 creates a prompt that includes the query "Please make a decision on the following expense claims based on the approval conditions," the expense claim information listed above, and the output format for the check result. The output format for the check result specifies, for example, the output format of the check result (e.g., JSON format), that the result should be one of the following: "Approved," "Returned," or "Requires Confirmation," and that a reason for the decision (text) within a specified number of characters should be generated.

[0030] In this embodiment, Server 1 provides LLM40 with expense regulations information indicating the approval conditions for expenses in each organization, as a prompt, to be used as a criterion when LLM40 checks whether or not to approve the expense request.

[0031] Expense policy information is data that describes the expense regulations within an organization (company). For example, if the company regulations stipulate a per-person limit of 10,000 yen for meeting expenses, the expense policy information can be defined as "Meeting expenses / Number of participants ≤ 10,000 yen". Expense policy information is created as an expense policy file 100 (e.g., a CSV file), with each of these regulations written on a separate line, and stored on server 1. An expense policy file 100 is prepared for each organization (customer company) using this system.

[0032] Server 1 creates a prompt that adds the company regulations described in expense regulations file 100 as approval conditions, and inputs this prompt into LLM40. LLM40 checks (determines) whether to approve the expense request according to the prompt. LLM40 reads each approval condition (company regulation) described in expense regulations file 100 line by line and determines whether the expense request satisfies each approval condition. LLM40 generates a check result that describes the determination result and the reason for the determination for each approval condition. For example, as shown in Figure 4, LLM40 generates a determination result for each approval condition such as "attachment of supporting documents," "participant information," and "meeting expenses," and if it determines that the approval condition is not met, it also generates the reason for the determination. Finally, LLM40 generates a determination result (status) such as "approved," "returned," or "requires confirmation" as an overall determination result, and also generates the reason for the determination.

[0033] If the judgment result is "return," Server 1 outputs the check result to Applicant Terminal 2 and returns the expense claim. In this case, the expense claim applicant can either fill in the missing information and resubmit the claim, or withdraw the expense claim for the time being.

[0034] If the judgment result is "approved" or "requires confirmation," Server 1 outputs the expense application information and the check result of the expense application to the accounting staff terminal 3. The accounting staff, having received the check result from LLM40, check the expense application information and either give final approval or return it to the expense applicant.

[0035] Figure 5 shows an example of a report display screen. Figure 6 shows an example of a timeline screen. Figures 5 and 6 illustrate examples of screens presented to accounting personnel.

[0036] For example, as shown in Figure 5, Server 1 generates a report summarizing the check results of multiple expense applications received from each expense applicant at predetermined intervals (for example, daily) and outputs (notifies) it to the accounting staff terminal 3. For example, Server 1 calculates the number of expense applications and the total application amount for each judgment result (status) such as "Approved," "Returned," and "Requires Confirmation," and outputs them as a report. If "Download CSV file" is selected on the screen in Figure 5, Server 1 outputs a CSV file (tabular file) showing the details of each expense application in a list to the accounting staff terminal 3.

[0037] For example, if the screen in Figure 5 accepts an operation to click the link "Click here for details on the application requiring confirmation," Server 1 outputs the timeline screen shown in Figure 6 to the accounting staff terminal 3. This screen displays the timeline from when the expense application is received until the AI ​​agent (LLM40) checks it and the accounting staff gives final approval. For example, as shown in Figure 6, Server 1 displays the date and time the expense application was received, the applicant's comments, the fact that the AI ​​agent checked the expense application, and the results of the AI ​​agent's check on the screen in chronological order. In the example screen in Figure 6, the expense staff is at the stage of final approval, so the message "Approval from the expense approver is required" is displayed. For example, the accounting staff gives final approval to the expense application by entering a comment indicating whether or not they approve in response to this message.

[0038] Let's return to Figure 4 and continue the explanation. As mentioned above, Server 1 ultimately outputs the LLM40 check results to the expense manager, but this system (service) also allows for manual re-checking by an external operator (for example, an employee of the company operating this system) as an option. Specifically, if the organization (customer company) has set the option for manual checks by an operator, Server 1 outputs the expense application information and the LLM40 check results to Operator Terminal 5. Operator Terminal 5 displays the expense application information and check results and accepts input from the operator regarding the re-check results of the expense application. If Server 1 receives input of the re-check results from the operator, Server 1 outputs the expense application information, the LLM40 check results, and the operator's re-check results to Accounting Staff Terminal 3. In this way, this system can also perform a double check with an AI agent check and a re-check by an operator.

[0039] In this way, Server 1 provides LLM40 with expense policy information (expense policy file 100) that shows different expense approval conditions for each organization, allowing LLM40 to check whether the expense application conforms to the regulations of each organization.

[0040] By the way, in the example above, it was explained that expense regulation file 100 is prepared for each organization, but in reality, internal expense regulations are often very complex and have a wide range of approval conditions, and it is inefficient for LLM40 to read and judge all of these internal regulations. For this reason, server 1 may prepare (store) expense regulation file 100 divided by organization and by expense category. In this case, server 1 classifies the expense application into one of several expense categories (business trip expenses, transportation expenses, meeting expenses, etc.) from the account title and other information included in the expense application information, and reads the expense regulation file 100 that corresponds to the organization to which the expense applicant belongs and the classified expense category. Server 1 then adds the contents of the read expense regulation file 100 to the prompt and causes LLM40 to generate a check result of whether or not to approve the expense application. In this way, processing efficiency can be improved by subdividing the expense regulation file 100 (expense regulation information) by organization and by expense category.

[0041] Figure 7 is a flowchart illustrating an example of the processing procedure performed by Server 1. Based on Figure 7, the processing details performed by Server 1 will be explained. The control unit 11 of server 1 acquires expense application information related to the expense application submitted by the expense applicant (step S11). The control unit 11 also acquires expense regulation information (e.g., expense regulation file 100) that indicates the approval conditions for expenses in the organization to which the expense applicant belongs (step S12).

[0042] The control unit 11 creates a prompt that instructs the LLM 40 to check whether to approve the expense application, taking into account the expense application information obtained in step S11 and the expense regulation information obtained in step S12 (step S13). Specifically, the control unit 11 creates a prompt that instructs the LLM 40 to determine one of the following statuses for the expense application: "Approved," "Requires Confirmation," or "Returned," and to generate a reason for the determination. The control unit 11 inputs the created prompt into the LLM 40 to generate the result of whether to approve the expense application (step S14).

[0043] The control unit 11 determines whether the check result is "return" (step S15). If it determines that it is "return" (S15: YES), the control unit 11 outputs the check result to the expense applicant (step S16) and terminates the series of processes. In this case, the expense applicant will need to fill in the necessary information and submit the expense application again.

[0044] If the check result is determined not to be "returned" (S15: NO), the control unit 11 determines whether the option for manual check by an operator is set for the organization to which the expense applicant belongs (step S17). If it is determined that the option is set (S17: YES), the control unit 11 outputs the expense application information and the check result to an operator outside the organization (step S18). The control unit 11 accepts input of the re-check result of the expense application from the operator (step S19).

[0045] After the process in step S19 is executed, or if the answer in step S17 is NO, the control unit 11 outputs the expense claim information, the check results from LLM40, and optionally the re-check results by the operator to the accounting staff within the organization (step S20). The accounting staff, upon receiving the check results, decide whether or not to give final approval to the expense claim and input the decision into this system. The control unit 11 then terminates the series of processes.

[0046] Based on the above, this embodiment can suitably support the approval of expense claims.

[0047] The embodiments disclosed herein should be considered in all respects to be illustrative and not restrictive. The scope of the present invention is indicated by the claims, not in the sense described above, and all modifications within the sense and scope equivalent to the claims are intended.

[0048] The matters described in each embodiment can be combined with each other. Furthermore, the independent and dependent claims described in the claims can be combined with each other in any combination, regardless of the form of reference. In addition, the claims use a form in which claims referencing two or more other claims (multi-claim form), but are not limited to this. A form in which multi-claims referencing at least one multi-claim (multi-multi-claim) may also be used. [Explanation of Symbols]

[0049] 1. Server (Information Processing Device) 11 Control Unit 12 Main memory 13 Communications Department 14 Auxiliary storage P Program 100 Expense Regulations File 141 Organization DB 142 UserDB 143 Application DB 2. Applicant's terminal 3. Accounting staff terminal 4. Generation Server 40. LLM (Language Model) 5 Operator terminal

Claims

1. We obtain expense application information related to expense applications submitted by expense applicants. By inputting a prompt into the language model that includes the aforementioned expense application information and expense regulations information indicating the approval conditions for expenses within the organization to which the expense applicant belongs, a check result is generated indicating whether or not to approve the expense application. The aforementioned expense application information and the check results are output to an external operator who will re-check the expense application. The aforementioned operator receives input of the results of the recheck of the expense application. The expense application information, the check results, and the re-check results are output to the accounting staff within the organization. A program that instructs a computer to perform a process.

2. The aforementioned check results include the decision on whether or not to approve the expense claim, and the reason for the decision. The program according to claim 1.

3. The expense regulations information for each of the multiple organizations is stored in the storage unit, When the aforementioned expense application information is obtained, the expense regulations information of the organization corresponding to the expense applicant is obtained from the storage unit. The check result is generated by inputting a prompt containing the acquired expense application information and the acquired expense regulation information into the language model. The program according to claim 1.

4. The aforementioned expense regulation information is stored in the storage unit for each organization and each expense item. Based on the aforementioned expense claim information, the expense claim is classified into one of several expense categories. The expense regulations information corresponding to the organization to which the expense applicant belongs and to the classified expense item is obtained from the storage unit. The program according to claim 3.

5. Obtain multiple expense application information items corresponding to multiple expense applications, The above check results are generated for each expense application. A report summarizing the results of the aforementioned checks for each expense claim will be sent to the accounting department. The program according to claim 1.

6. We obtain expense application information related to expense applications submitted by expense applicants. By inputting a prompt into the language model that includes the aforementioned expense application information and expense regulations information indicating the approval conditions for expenses within the organization to which the expense applicant belongs, a check result is generated indicating whether or not to approve the expense application. The aforementioned expense application information and the check results are output to an external operator who will re-check the expense application. The aforementioned operator receives input of the results of the recheck of the expense application. The expense application information, the check results, and the re-check results are output to the accounting staff within the organization. An information processing method in which a computer performs the processing.

7. An information processing device comprising a control unit, The control unit, We obtain expense application information related to expense applications submitted by expense applicants. By inputting a prompt into the language model that includes the aforementioned expense application information and expense regulations information indicating the approval conditions for expenses within the organization to which the expense applicant belongs, a check result is generated indicating whether or not to approve the expense application. The aforementioned expense application information and the check results are output to an external operator who will re-check the expense application. The aforementioned operator receives input of the results of the recheck of the expense application. The expense application information, the check results, and the re-check results are output to the accounting staff within the organization. Information processing device.

Images