Enterprise data-integrated ai-based electronic approval pre-verification method and review automation system

Abstract

An enterprise data-integrated AI-based electronic approval pre-verification method and a review automation system are disclosed. The AI-based electronic approval pre-verification method may comprise the steps of: receiving an approval document from an electronic approval system; and performing pre-verification on the approval document on the basis of natural language processing (NLP).

Description

Enterprise Data Integrated AI-Based Electronic Approval Pre-Verification Method and Review Automation System

[0001] The following description concerns the corporate electronic approval system.

[0002] With the advancement of digital and computer technologies, companies are seeking to improve processing efficiency and productivity by adopting digital and network technologies into their business operations. A prime example is the electronic approval system that eliminates paper usage in the office, and these technologies are also being applied to all business processes, including sales, accounting, and human resource management.

[0003] As an example of electronic approval technology, Korean Registered Patent No. 10-0454735 (registered on October 19, 2004) discloses a technology for providing a corporate portal service that integrates electronic approval, knowledge management systems, resource management, document management, and management information systems, enabling access and use through a single unified window such as a web browser.

[0004] In corporate electronic approval systems, the review of submitted documents largely relies on manual work by senior approvers, and approvers face a heavy workload as they must review a large number of documents. Furthermore, errors or omissions in the document content may go unnoticed, and it is inefficient as all reviewers in the approval chain repeatedly examine the same content; there are also issues with the difficulty of consistently verifying whether corporate policies or regulations have been violated.

[0005] It is difficult to judge the appropriateness of content based solely on simple form checks or mandatory item checks, and rule-based review systems struggle to handle new types of documents or exceptions. Furthermore, continuous improvement is difficult because individual feedback from approvers is not reflected in the system.

[0006] We can provide an AI-based electronic approval pre-verification method and system that utilizes AI (artificial intelligence) to support the automatic review and analysis of document content and improves review accuracy through learning based on approver feedback.

[0007] We can provide an AI-based electronic approval pre-verification method and system that supports data consistency verification through integration with ERP / CRM systems, automatic verification of compliance with corporate policies and regulations, and the presentation of customized review points for selected approvals.

[0008] An AI-based electronic approval pre-verification method performed by a computer device comprising at least one processor comprises: a step of receiving an approval document from an electronic approval system; and a step of performing pre-verification of said approval document based on natural language processing (NLP) through integration with an ERP (Enterprise Resource Planning) system or a CRM (Customer Relationship Management) system, wherein the step of performing the pre-verification includes: a step of identifying a regulation document related to said approval document; a step of determining whether said approval document complies with regulations based on said regulation document; and a step of suggesting an alternative solution if said approval document violates regulations.

[0009] According to one aspect, the receiving step may include the step of detecting and receiving a new submission document of the electronic approval system as the approval document.

[0010] According to another aspect, the receiving step may include a step of extracting metadata including the author, approval line, and document type of the approval document.

[0011] According to another aspect, the receiving step may include a step of verifying the validity of the document format for the approval document; and a step of converting the document content of the approval document into a standardized format.

[0012] According to another aspect, the steps performed above may include: a step of performing sentence semantic analysis on the approval document using a language model; and a step of detecting contextual errors in the content based on contextual information of the approval document.

[0013] According to another aspect, the step of performing the above may include a step of detecting anomaly patterns by comparing the approval document with similar documents.

[0014] According to another aspect, the step of performing the above may include comparing system data retrieved from the ERP (Enterprise Resource Planning) system or the CRM (Customer Relationship Management) system with the data of the approval document.

[0015] According to another aspect, the step of performing the above may include a step of extracting review items from the approval document based on the job or authority information of the approver.

[0016] According to another aspect, the step of performing the above may include a step of analyzing review patterns based on past approval history for the approval document.

[0017] According to another aspect, the step of performing the above may include providing the author with the results of a preliminary verification of the approval document for an opportunity to modify the approval document.

[0018] According to another aspect, the step of performing the above may include collecting user feedback on the results of the preliminary verification of the approval document and updating a model for analyzing the document content.

[0019] According to another aspect, the step of performing the above may include collecting user feedback on the results of the prior verification of the approval document and updating a model for analyzing document content through reinforcement learning based on the user feedback.

[0020] According to another aspect, the step of performing the above may include the step of highlighting and providing the review items of the approval document to the approver of the approval document.

[0021] According to another aspect, the step of performing the above may include, in providing a pre-verification report including data consistency verification results and compliance verification results, analyzing the job or authority information of each approver included in the approval line of the approval document to extract review items for each approver and highlighting the corresponding review items within the approval document to provide a differentiated review guide for each approver.

[0022] A computer device is provided that includes at least one processor implemented to execute a readable command on a computer device, wherein the at least one processor processes the process of receiving an approval document from an electronic approval system; and the process of performing a preliminary verification of the approval document based on natural language processing (NLP) through integration with an ERP (Enterprise Resource Planning) system or a CRM (Customer Relationship Management) system, and wherein the at least one processor processes the process of identifying a regulation document related to the approval document; the process of determining whether the approval document complies with regulations based on the regulation document; and the process of suggesting an alternative solution if the approval document violates regulations.

[0023] According to embodiments of the present invention, automatic review and analysis of document content can be supported by utilizing AI (artificial intelligence), and review accuracy can be improved through learning based on approver feedback.

[0024] According to embodiments of the present invention, data consistency verification can be supported through linkage with an ERP / CRM system, automatic verification of compliance with corporate policies and regulations, and presentation of customized review points for approval selection can be supported.

[0025] FIG. 1 is a block diagram illustrating an example of the internal configuration of a computer device in an embodiment of the present invention.

[0026] FIG. 2 illustrates the overall architecture of an AI-based electronic payment pre-verification system in one embodiment of the present invention.

[0027] FIG. 3 illustrates a detailed configuration diagram of an AI engine in one embodiment of the present invention.

[0028] FIG. 4 illustrates a detailed configuration diagram of a basic model-based review engine in one embodiment of the present invention.

[0029] FIG. 5 illustrates a detailed configuration diagram of a knowledge abstraction and transfer system in one embodiment of the present invention.

[0030] FIG. 6 illustrates a process sequence for the automation of pre-verification and review of corporate data-integrated AI-based electronic approval in one embodiment of the present invention.

[0031] FIGS. 7 and 8 illustrate a system linkage sequence for AI-based electronic approval in an embodiment of the present invention.

[0032] FIG. 9 illustrates a state diagram of an electronic approval submission process in an embodiment of the present invention.

[0033] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings.

[0034]

[0035] The embodiments of the present invention relate to in-enterprise electronic approval technology.

[0036] Embodiments including those specifically disclosed in this specification can construct an AI-based document pre-review system to automatically extract and analyze key review elements by document type.

[0037] An AI-based electronic approval pre-verification system according to embodiments of the present invention may be implemented by at least one computer device, and an AI-based electronic approval pre-verification method according to embodiments of the present invention may be performed through at least one computer device included in the AI-based electronic approval pre-verification system. At this time, a computer program according to one embodiment of the present invention may be installed and run on the computer device, and the computer device may perform an AI-based electronic approval pre-verification method according to embodiments of the present invention under the control of the run computer program. The above-described computer program may be stored on a computer-readable recording medium to be combined with the computer device to execute the AI-based electronic approval pre-verification method on the computer.

[0038] FIG. 1 is a block diagram illustrating an example of a computer device according to an embodiment of the present invention. For example, an AI-based electronic approval pre-verification system according to embodiments of the present invention can be implemented by a computer device (100) illustrated in FIG. 1.

[0039] As illustrated in FIG. 1, a computer device (100) may include a memory (110), a processor (120), a communication interface (130), and an input / output interface (140) as components for executing an AI-based electronic payment pre-verification method according to embodiments of the present invention.

[0040] Memory (110) is a computer-readable recording medium and may include a non-perishable mass storage device such as RAM (random access memory), ROM (read only memory), and a disk drive. Here, a non-perishable mass storage device such as a ROM and a disk drive may be included in the computer device (100) as a separate permanent storage device distinct from memory (110). Additionally, an operating system and at least one program code may be stored in memory (110). These software components may be loaded into memory (110) from a computer-readable recording medium separate from memory (110). This separate computer-readable recording medium may include a computer-readable recording medium such as a floppy drive, disk, tape, DVD / CD-ROM drive, or memory card. In another embodiment, software components may be loaded into memory (110) through a communication interface (130) rather than a computer-readable recording medium. For example, software components can be loaded into the memory (110) of the computer device (100) based on a computer program installed by files received through the network (160).

[0041] The processor (120) may be configured to process instructions of a computer program by performing basic arithmetic, logic, and input / output operations. Instructions may be provided to the processor (120) via memory (110) or a communication interface (130). For example, the processor (120) may be configured to execute instructions received according to program code stored in a recording device such as memory (110).

[0042] The communication interface (130) may provide a function for the computer device (100) to communicate with other devices through a network (160). For example, requests, commands, data, files, etc. generated by the processor (120) of the computer device (100) according to program code stored in a recording device such as memory (110) may be transmitted to other devices through the network (160) under the control of the communication interface (130). Conversely, signals, commands, data, files, etc. from other devices may be received by the computer device (100) through the communication interface (130) of the computer device (100) via the network (160). Signals, commands, data, etc. received through the communication interface (130) may be transmitted to the processor (120) or memory (110), and files, etc. may be stored in a storage medium (the permanent storage device described above) that the computer device (100) may further include.

[0043] The communication method is not limited and may include not only communication methods utilizing communication networks (e.g., mobile communication networks, wired internet, wireless internet, broadcasting networks) that the network (160) may include, but also short-range wired / wireless communication between devices. For example, the network (160) may include any one or more networks such as a PAN (personal area network), LAN (local area network), CAN (campus area network), MAN (metropolitan area network), WAN (wide area network), BBN (broadband network), and the Internet. Additionally, the network (160) may include any one or more network topologies such as a bus network, star network, ring network, mesh network, star-bus network, tree or hierarchical network, but is not limited thereto.

[0044] The input / output interface (140) may be a means for interfacing with an input / output device (150). For example, the input device may include a device such as a microphone, keyboard, camera, or mouse, and the output device may include a device such as a display or speaker. As another example, the input / output interface (140) may be a means for interfacing with a device in which the functions for input and output are integrated into one, such as a touchscreen. The input / output device (150) may be composed of a computer device (100) and a single device.

[0045] Additionally, in other embodiments, the computer device (100) may include fewer or more components than the components of FIG. 1. However, it is not necessary to clearly illustrate most of the prior art components. For example, the computer device (100) may be implemented to include at least some of the input / output devices (150) described above, or may include other components such as a transceiver, a camera, various sensors, a database, etc.

[0046] Below, specific embodiments of the technology for automating pre-verification and review of electronic approvals based on AI-integrated enterprise data will be described.

[0047] The AI-based electronic payment pre-verification system according to the present invention may include the following technical features.

[0048] (1) Establishment of an AI-based document pre-review system

[0049] It can automatically extract and analyze key review elements by document type, and perform data consistency verification by integrating with enterprise ERP (Enterprise Resource Planning) and CRM (Customer Relationship Management) systems. Furthermore, it can verify compliance by linking with corporate policy and regulation databases and present differentiated key review points based on the approval process.

[0050] (2) Training data configuration

[0051] The content and structure of existing approved documents can be utilized as training data, and the approvers' revision / rejection history and reasons can be included. Mapping information with relevant data from ERP / CRM systems can be added, and cases of policy / regulation violations and exception approvals can be used as reference data.

[0052] (3) Machine learning model configuration

[0053] Natural language processing models can be applied for document content analysis, and pattern recognition models can be implemented for data consistency verification. Additionally, classification models can be used to identify key review items for approval selection, and continuous improvement can be performed by applying feedback-based reinforcement learning models.

[0054] (4) Output data configuration

[0055] It supports identifying errors or omissions in document content and suggesting improvements, and allows specifying discrepancies with ERP / CRM data. Additionally, it can mark items with potential policy or regulation violations and categorize and present matters requiring focused review during the approval screening process.

[0056] (5) System application plan

[0057] It can be automatically executed during the pre-review stage prior to document submission, providing immediate feedback on review results to the submitter to offer an opportunity for revision. Additionally, review points can be highlighted for approvers, and the history of acceptance or rejection of system feedback can be incorporated into the learning process.

[0058] The AI-based electronic approval pre-verification system according to the present invention may apply the following core technologies.

[0059] (1) Document content analysis technology

[0060] Natural Language Processing (NLP)-based document content parsing

[0061] Document content vectorization can be performed using language models such as BERT, GPT, RoBERTa, and XLNet. Additionally, keyword extraction can be performed using embedding technologies such as Word2Vec, GloVe, and FastText, and contextual information can be analyzed using technologies such as LSTM, GRU, and Attention mechanisms. Furthermore, key keywords and contextual information can be extracted by document type, and the logical structure and flow of the document can be analyzed.

[0062] Document quality verification

[0063] It is possible to identify missing essential information and derive supplementary requirements; for example, missing information can be identified by utilizing methods such as rule-based verification, statistical-based verification, and machine learning-based verification. Additionally, context-based errors and contradictions can be detected; for example, contextual errors can be detected by utilizing techniques such as sentence similarity analysis, semantic redundancy checks, and logical flow analysis. Furthermore, anomalous patterns can be detected through comparison with similar documents; for example, anomalous patterns can be detected by utilizing clustering techniques such as K-means, DBSCAN, and Hierarchical Clustering.

[0064] (2) Data consistency verification technology

[0065] ERP / CRM Data Integration

[0066] Real-time data retrieval can be performed through REST APIs and the like. In this process, protocols such as REST APIs, GraphQL, and SOAP can be utilized. Furthermore, data linkage between systems can be implemented using methods such as ETL, CDC, and real-time synchronization. Additionally, system integration can be achieved by utilizing architectures such as microservices, API Gateways, and service meshes.

[0067] In addition, it can perform comparative verification between numerical values / information within the document and system data, and can analyze the causes of data discrepancies.

[0068] Data Mapping and Transformation

[0069] Automatic data format conversion between systems can be performed, utilizing tools such as regular expressions, parser generators, and data converters. Additionally, unstructured data can be standardized, and data normalization can be carried out using technologies such as ontology, semantic mapping, and metadata mapping. Furthermore, data correction considering time differences can be performed, utilizing techniques such as time series analysis, statistical correction, and outlier removal.

[0070] Model Update

[0071] It is possible to perform model retraining based on collected feedback, measure review accuracy and improve performance, and learn new types / patterns.

[0072] (3) Policy / regulatory compliance verification technology

[0073] Establishment of a Regulation Database

[0074] Automatic parsing and rule formulation of policy / regulation documents can be performed, and for example, automatic parsing of regulations can be carried out by utilizing technologies such as natural language processing, information extraction, and knowledge graphs. In addition, relationships and priorities between regulations can be defined, and relationships between regulations can be defined by utilizing technologies such as ontology, RDF, and OWL. Furthermore, conditions and processes for exception approval cases can be defined, and exception handling can be implemented by utilizing technologies such as rule engines, inference engines, and decision engines.

[0075] Determining compliance

[0076] It is possible to perform mapping between document content and relevant regulations, utilizing technologies such as rule-based engines, expert systems, and fuzzy logic. Additionally, it can assess the likelihood of regulatory violations and calculate risk levels, employing techniques such as Bayesian networks, decision trees, and random forests to evaluate the probability of violation. Furthermore, in the event of a violation, alternative measures or exception approval procedures can be proposed; for instance, alternative solutions can be suggested using technologies such as case-based reasoning, similarity analysis, and pattern matching.

[0077] (4) Technology for deriving review points for approval screening

[0078] Approver Profile Analysis

[0079] Key areas of interest can be identified based on approvers' job duties and authority information, utilizing techniques such as collaborative filtering, content-based filtering, and hybrid filtering. Additionally, review patterns can be analyzed by utilizing technologies such as time-series analysis, pattern mining, and sequence analysis to examine focused review patterns based on past approval history. Furthermore, specialized review items can be defined by department or organization, employing techniques such as clustering, topic modeling, and text mining.

[0080] Presentation of customized review points

[0081] Differentiated review items can be extracted based on the approval screening process, utilizing technologies such as ranking learning, recommendation systems, and personalization engines. Additionally, importance-based priorities are assigned to review items, and these priorities can be calculated using techniques such as weighted learning, importance scoring, and priority determination. Furthermore, guidelines regarding specific review requirements can be provided, and these guidelines can be generated using technologies such as explainable AI, interpretable models, and evidence extraction.

[0082] (5) Continuous learning / improvement skills

[0083] Collection of feedback information

[0084] Review comments and revision history can be automatically collected, utilizing technologies such as user behavior analysis, log mining, and event tracking to gather feedback. Additionally, reasons for acceptance or rejection can be analyzed and patterned using techniques such as sentiment analysis, opinion mining, and reputation analysis. Furthermore, exception handling cases can be managed separately, employing technologies such as anomaly detection, pattern recognition, and cluster analysis.

[0085] Model Update

[0086] Model retraining based on collected feedback can be performed using techniques such as incremental learning, transfer learning, and online learning. Additionally, review accuracy can be measured and performance improved; for instance, accuracy can be measured using techniques such as A / B testing, cross-validation, and performance metrics. Furthermore, learning of new types or patterns can be performed, utilizing methods such as active learning, semi-supervised learning, and self-learning.

[0087] The computer device (100) according to the present embodiment can provide an AI-based electronic payment pre-verification service to a client by accessing a dedicated application installed on the client or a web / mobile site related to the computer device (100). The computer device (100) may be configured with an AI-based electronic payment pre-verification system implemented on a computer. For example, the AI-based electronic payment pre-verification system may be implemented in the form of a program that operates independently, or configured as an in-app of a specific application so that it can operate on said specific application.

[0088] The processor (120) of the computer device (100) may be implemented as a component for performing the following AI-based electronic approval pre-verification method. Depending on the embodiment, the components of the processor (120) may be optionally included in or excluded from the processor (120). Additionally, depending on the embodiment, the components of the processor (120) may be separated or merged to represent the function of the processor (120).

[0089] These processors (120) and components of the processor (120) can control a computer device (100) to perform steps included in the following AI-based electronic approval pre-verification method. For example, the processor (120) and components of the processor (120) may be implemented to execute instructions according to the code of an operating system included in memory (110) and the code of at least one program.

[0090] Here, the components of the processor (120) may be representations of different functions performed by the processor (120) according to instructions provided by program code stored in the computer device (100).

[0091] The processor (120) can read necessary instructions from memory (110) in which instructions related to the control of the computer device (100) are loaded. In this case, the read instructions may include instructions for controlling the processor (120) to execute the steps to be described later.

[0092] The steps included in the AI-based electronic approval pre-verification method to be described later may be performed in a different order than the one described, and some of the steps may be omitted or additional processes may be included.

[0093] FIG. 2 illustrates the overall architecture of an AI-based electronic payment pre-verification system in one embodiment of the present invention.

[0094] Referring to FIG. 2, the main modules of the AI-based electronic approval pre-verification system according to the present invention may include a document input module (210), a preprocessing module (220), an AI review engine (230), a feedback processing module (240), a learning model update module (250), and a review result output module (260).

[0095] The document input module (210) is responsible for receiving documents to be reviewed from the electronic approval system. The document input module (210) may include a document receiving submodule and a document classification submodule. At this time, the document receiving submodule can detect and receive new submitted documents from the electronic approval system. The document receiving submodule can extract metadata of the document (author, approval line, document type, etc.) and can perform a first verification of the validity of the document format. In addition, the document classification submodule can perform category classification according to the document type and content. The document classification submodule can determine the review priority and importance and determine the set of review rules to be applied.

[0096] The preprocessing module (220) can perform document format conversion and normalization on the document to be reviewed. The preprocessing module (220) may include a data normalization submodule and a text analysis submodule. In this case, the data normalization submodule can convert the document content into a structured format. The data normalization submodule can standardize data formats such as numbers, dates, and currencies, and can process special characters and formatting. Additionally, the text analysis submodule can perform sentence structure analysis through natural language processing. The text analysis submodule can extract key keywords and contextual information, and can identify the logical structure of the document.

[0097] The AI ​​review engine (230) serves as a main engine composed of four core verification modules. The AI ​​review engine (230) may include a document content analyzer (231), a data consistency verifier (232), a policy / regulatory compliance verifier (233), an approval screening review point generator (234), and a review result integrator (235).

[0098] The document content analyzer (231) can perform NLP-based document content verification. The document content analyzer (231) may include a context understanding unit that performs BERT-based sentence semantic analysis, an error detection unit that detects contradictions / omissions in the content, and a quality evaluation unit that evaluates the completeness of the document.

[0099] The data consistency verifier (232) can verify consistency with ERP / CRM data. The data consistency verifier (232) may include a data inquiry unit that queries ERP / CRM system data in real time, a comparison analysis unit that compares document data and system data, and a discrepancy analysis unit that analyzes the cause of the difference.

[0100] The policy / regulatory compliance verifier (233) can verify whether internal regulations are complied with. The policy / regulatory compliance verifier (233) may include a regulation mapping unit that identifies / maps relevant regulations, a compliance evaluation unit that determines compliance, a risk calculation unit that calculates the risk of violation of regulations, etc.

[0101] The approval selection review point generator (234) can derive customized review items. The approval selection review point generator (234) may include a profile analysis unit that analyzes interests of each approver, a point extraction unit that derives key review items, and a priority determination unit that evaluates the importance of review items.

[0102] As a result of the review, the integrator (235) can synthesize each verification result to generate a final result.

[0103] The feedback processing module (240) can process user feedback regarding the review results through the AI ​​review engine (230). The feedback processing module (240) may include a feedback collection submodule and a feedback analysis submodule. In this case, the feedback collection submodule can collect user response data. The feedback collection submodule can record revision / rejection history and identify exception handling cases. Additionally, the feedback analysis submodule can analyze feedback patterns, derive items requiring improvement, and perform training data processing.

[0104] The learning model update module (250) can support model updates based on collected feedback. The learning model update module (250) may include a model evaluation submodule and a model update submodule. In this case, the model evaluation submodule can measure the performance of the current model, identify areas requiring improvement, and determine the timing for model updates. Additionally, the model update submodule can retrain the model with new training data, verify model performance, and distribute the updated model.

[0105] The review result output module (260) may include a review integration submodule and a report generation submodule. In this case, the result integration submodule can synthesize each verification result, remove duplicate items, and perform sorting by priority. Additionally, the report generation submodule generates a review result report, and can create a customized report for approval selection and generate visualization data.

[0106] The AI-based electronic approval pre-verification system according to the present invention may include an ERP system (241), a CRM system (242), a policy / regulation DB (243), an approval history DB (244), and a learning model repository (245) as data repositories.

[0107] The ERP system (241) and CRM system (242) can provide core corporate data.

[0108] The policy / regulation DB (243) can store internal regulations and policy information.

[0109] The approval history DB (244) can store past approval records and pattern information.

[0110] The learning model repository (245) can manage AI models and training data.

[0111] The AI-based electronic payment pre-verification system according to the present invention can perform secure data exchange with an external system through an API gateway (250) in a system linkage environment, and can provide real-time data synchronization and integrity guarantee.

[0112] FIG. 3 illustrates a detailed configuration diagram of an AI engine in one embodiment of the present invention.

[0113] The AI ​​engine of FIG. 3 is included on the AI ​​review engine (230) of FIG. 2 and may include a configuration that utilizes multiple basic models reflecting various characteristics of the company's composition (e.g., by department) when processing the company's data.

[0114] In other words, the AI ​​engine for automating pre-verification and review of electronic approval can operate universally when utilizing AI in the document content analyzer (231), data consistency verifier (232), policy / regulation compliance verifier (233), and approval selection review point generator (234) of the AI ​​review engine (230).

[0115] Referring to FIG. 3, the AI ​​engine for automating pre-verification and review of electronic approval may include a Foundation Model Layer (310), a Knowledge Abstraction Layer (320), a Data Integration Layer (330), a Context Processing Layer (340), and a Model Management Layer (350).

[0116] The foundation model layer (310) may include a foundation model interface and a model replacement mechanism. In this case, the foundation model interface may provide a standardized interface with various foundation models. Through this, the impact on lower layers can be minimized when replacing models, and model-specific specialized functions can be abstracted and provided. Additionally, the model replacement mechanism may provide a zero-learning or few-shot learning support interface, standardize prompt engineering templates, and automate the performance verification process when applying a new foundation model.

[0117] The knowledge abstraction layer (320) may include a knowledge abstraction module, a knowledge management module, and a transfer learning module. The knowledge abstraction module can abstract domain-specific knowledge into a standardized form, formalize enterprise-specific terms and rules, and maintain an independent knowledge representation structure of the basic model. Additionally, the knowledge management module can perform version control of the abstracted knowledge, manage dependencies between knowledge, and manage rule-based knowledge and learning-based knowledge in an integrated manner. Furthermore, the transfer learning module can extract and preserve features of existing learning results, transform knowledge into a form suitable for the new basic model, and manage incremental learning and relearning processes.

[0118] The data integration layer (330) may include a data collection interface and a data refinement and integration module. In this case, the data collection interface may provide an email server integration module, a meeting minutes storage access module, and a messenger system integration module. Additionally, the data refinement and integration module may perform unstructured data normalization, perform duplicate information removal and integration, and perform data reliability evaluation.

[0119] The context processing layer (340) may include a context analysis engine, a relationship extraction engine, and a time series analysis engine. In this case, the context analysis engine analyzes the relationships between documents, can identify the flow of information in chronological order, and can extract decision patterns. Additionally, the relationship extraction engine can analyze work relationships between departments / personnel, extract processing patterns by document type, and identify exception handling cases. Furthermore, the time series analysis engine can track changes in information over time, identify periodic patterns, and detect abnormal patterns.

[0120] The model management layer (350) may include a model version management module, a performance monitoring module, and a training data management module. In this case, the model version management module manages the performance history of the base model by version, can provide support for model rollback, and can analyze the impact of model changes. Additionally, the performance monitoring module measures real-time performance indicators, can analyze factors causing performance degradation, and can identify improvement points. Additionally, the training data management module evaluates data quality, can identify training data reinforcement points, and can monitor data imbalance.

[0121] FIG. 4 illustrates a detailed configuration diagram of a basic model-based review engine in one embodiment of the present invention.

[0122] Referring to FIG. 4, the basic model-based review engine may include a model management area, an inference processing area, a training data management area, and an integrated interface area.

[0123] The model management area may include a model version controller that performs version control and history tracking of the underlying model, a performance monitoring engine that performs real-time performance measurement and analysis, and a model replacement manager module that supports non-disruptive model replacement and rollback.

[0124] The inference processing area may include a context understanding processor that performs semantic analysis of document content, a rule verification processor that verifies policy / regulation compliance, and a consistency check processor that verifies data consistency.

[0125] The training data management area may include a data quality manager module that ensures the quality of training data, a training data selector that selects optimal training data, and a data augmentation processor that reinforces or expands training data.

[0126] The integrated interface area may include an enterprise system integration module that provides real-time integration with ERP, CRM, etc., a data source integration module that integrates with data sources such as email and meeting minutes, and an external system integration module that supports secure integration with external systems.

[0127] FIG. 5 illustrates a detailed configuration diagram of a knowledge abstraction and transfer system in one embodiment of the present invention.

[0128] Referring to FIG. 5, the knowledge abstraction and transfer system may include a knowledge abstraction area, a transfer learning area, and an integrated repository area.

[0129] The knowledge abstraction domain may include a rule abstraction engine that converts business rules into a standardized form, a pattern abstraction engine that identifies / abstracts repetitive patterns, a policy abstraction engine that converts corporate policies into a structured form, and a process abstraction engine that models business processes.

[0130] The transfer learning domain may include a feature extractor that extracts key features of an existing model, a model transformer that transforms the model to fit a new base model, a knowledge mapping engine that maps relationships between knowledge, and a performance optimization engine that optimizes transfer learning performance.

[0131] The integrated repository area may include an abstract knowledge repository that systematically stores abstract knowledge, a transfer learning result repository that manages transfer learning results, and a metadata repository that manages system operation metadata.

[0132] FIG. 6 illustrates a process sequence for the automation of pre-verification and review of corporate data-integrated AI-based electronic approval in one embodiment of the present invention.

[0133] The document review process may include an initial data collection phase and a review execution phase.

[0134] In the initial data collection phase, relevant contextual information can be collected when a user submits a document. For example, related data such as emails and meeting minutes can be automatically searched, and the history of similar past cases can be retrieved.

[0135] In the review execution phase, integrated data collected through the initial data collection phase can be transmitted to a review engine based on an underlying model. At this stage, relevant rules / policies can be applied through context analysis, and results can be provided to the user after a multi-faceted review.

[0136] The learning and feedback process may include a feedback collection step and a learning data update step.

[0137] In the feedback collection stage, feedback regarding the review results can be collected from users. During this stage, the feedback content can be analyzed and patterned, and the analyzed feedback can be stored in a database.

[0138] In the training data update stage, training data can be updated based on feedback collected through the feedback collection stage, the quality of the updated data can be verified, and the consistency of the training results can be checked.

[0139] The basic model replacement process may include a replacement preparation phase and a model transition phase.

[0140] In the replacement preparation phase, existing training data and knowledge can be extracted and transformed into a form suitable for the new foundation. At this stage, preparations for knowledge transfer through transfer learning can be made.

[0141] During the model transition phase, the performance of the new foundational model can be pre-verified, and the model can be transitioned in a non-disruptive manner while monitoring stability after the transition.

[0142] The enterprise data-integrated AI-based electronic approval pre-verification and review automation process according to the present invention can support real-time data processing at each stage through a real-time interconnection environment, enable the reflection of immediate feedback, and provide seamless connectivity between systems. Furthermore, in a non-stop operation environment, there is no service interruption when replacing the base model, it supports gradual model transitions, and allows for the preparation of rollback scenarios. Moreover, the present invention supports scalability, making it easy to add new data sources. Not only is it possible to apply various base models, but it can also support the dynamic expansion of review rules.

[0143] FIGS. 7 and 8 illustrate a system linkage sequence for AI-based electronic approval in an embodiment of the present invention.

[0144] The present invention can support groupware integration and legacy system integration through the integration of the initial review process.

[0145] In the groupware integration stage, automatic review can be initiated upon receipt of an electronic approval document, the scope of the review can be set based on approval line information, and the review results can be automatically attached to the approval document.

[0146] During the legacy system integration phase, financial and HR data from the ERP system can be viewed in real time, customer and contract information from the CRM system can be linked and verified, and data consistency verification can be performed automatically.

[0147] The present invention can support integration with mail / log systems and external systems as a context data collection integration environment.

[0148] During the mail / log system integration phase, related mail threads can be automatically tracked, relevant meeting minutes can be searched and referenced, and abnormal patterns can be identified through system log analysis.

[0149] During the external system integration phase, secure external connections can be performed through the API Gateway. At this stage, real-time data verification can be requested, and the results of the external verification can be reflected in decision-making.

[0150] The present invention can support system data updates and learning data management through processing integration after payment approval.

[0151] During the system data update phase, relevant information in the ERP system can be automatically updated and contract information in the CRM system can be updated, and the change history can be automatically recorded.

[0152] In the training data management stage, approved cases are converted into training data to learn linkage patterns between systems, thereby improving the accuracy of the review model.

[0153] The present invention can support daily synchronization and anomaly detection through a regular synchronization process.

[0154] In the daily synchronization phase, the consistency of the entire system data can be verified, changes can be automatically synchronized, and synchronization results can be monitored.

[0155] In the anomaly detection stage, data inconsistencies can be automatically detected, and in the event of an error, immediate notifications can be provided and the recovery process can be automatically initiated.

[0156] The present invention can support access control and fault response in an integrated security management environment.

[0157] In the access control stage, access rights can be granularly defined by system, encrypted data transmission can be guaranteed, and access history can be recorded in detail.

[0158] In the failure response phase, the inter-system connectivity status is monitored in real time to provide alternative routes and execute automatic recovery processes in the event of a failure.

[0159] FIG. 9 illustrates a state diagram of an electronic approval submission process in an embodiment of the present invention.

[0160] Referring to Fig. 9, during the document creation stage, real-time guidance can be provided through AI intervention while creating the document, similar document templates can be automatically recommended, and immediate notifications can be provided if required items are missing.

[0161] Context collection can be performed during the AI ​​preliminary review stage. At this stage, relevant materials such as emails, chats, and meeting minutes can be automatically collected, reference data can be secured based on previous similar cases, and correlation analysis of ongoing cases in related departments can be conducted.

[0162] During the AI ​​preliminary review stage, consistency verification can be performed, which involves real-time comparison with ERP / CRM data, automatic verification of budget / authority scopes, and automatic identification of data inconsistencies.

[0163] During the AI ​​preliminary review phase, compliance verification and risk assessment can be performed. At this stage, compliance verification can confirm adherence to internal regulations and laws, conduct adequacy assessments based on precedents and case law, and provide guidelines for handling exceptions. In the risk assessment process, financial and legal risks can be automatically calculated, the impact of reputational and operational risks can be analyzed, and review points can be presented based on risk levels.

[0164] During the approval process, the AI ​​reviewer analysis process can extract key review items for each reviewer, analyze past revision / criteria patterns, and provide guidelines for review efficiency.

[0165] During the approval process, AI analysis of approvers can highlight individual interests, present the history of handling similar agenda items, and emphasize key decision-making points.

[0166] Automation in the approval process can reduce review time, decrease approval backlog, and improve approval quality.

[0167] At the approval completion stage, automatic data from related systems can be linked through system integration, and automatic processes in business systems can be initiated to provide automatic notifications to related departments.

[0168] At the approval completion stage, data updates such as data synchronization with enterprise systems, automatic updating of relevant status / statistics, and automatic reflection of budget / resource status can be performed.

[0169] History management can be performed at the approval completion stage, during which approval history can be automatically classified and saved, automatically converted into AI training data, or audit trail data generated.

[0170] Accordingly, according to embodiments of the present invention, flexibility for replacing the base model can be provided. With a modular structure, the application of a new base model is easy, transfer is possible without loss of existing training data, and uninterrupted model replacement can be supported.

[0171] According to embodiments of the present invention, integrated analysis of various unstructured data can be performed through intelligent context analysis, and not only can related information such as emails and meeting minutes be automatically linked, but the work context according to the flow of time can also be identified.

[0172] According to embodiments of the present invention, as an advanced learning system, evolution through continuous learning of approval patterns is possible, specialized review points by department / job function can be derived, and adaptive learning for exception cases can be performed.

[0173] According to embodiments of the present invention, potential problems can be identified at the pre-submission stage through preemptive error prevention, data consistency can be verified in real time, and the possibility of regulatory violations can be prevented in advance.

[0174] According to embodiments of the present invention, by supporting a customized review environment, interests of each approver can be automatically analyzed, key review points can be highlighted, and decision-making based on similar cases can be supported.

[0175] According to embodiments of the present invention, financial / legal / reputational risks can be comprehensively evaluated through integrated risk management, the grounds and history of exception handling can be systematized, and evidence for audit response can be automatically secured.

[0176] According to embodiments of the present invention, workforce management efficiency can be realized by reducing the effort required for review tasks and automating simple repetitive tasks to improve the concentration of specialized personnel on core tasks.

[0177] According to embodiments of the present invention, error costs can be reduced by reducing rework caused by data errors, preventing penalties for violations of regulations, and reducing system linkage errors.

[0178] According to embodiments of the present invention, the overall processing time of system operation can be reduced by shortening the approval processing time and system integration time, and by reducing the audit response time.

[0179] According to embodiments of the present invention, through system scalability, it is easy to apply to new business areas, linkage with various enterprise systems is possible, and the addition of new data sources can be freely done.

[0180] According to embodiments of the present invention, real-time reflection of policy / regulation changes is possible through rule management flexibility, independent management of department-specific rules is possible, and dynamic addition of exception handling criteria is possible.

[0181] According to embodiments of the present invention, continuous improvement based on user feedback is possible through user-customized support, it is easy to add specialized functions by department / job function, and flexible changes to the user interface are possible.

[0182] According to embodiments of the present invention, data consistency between systems can be ensured through data quality management, unified management of master data is possible, and automatic detection and correction of data errors is possible.

[0183] According to embodiments of the present invention, automatic classification of security grades by document is possible with enhanced security, fine-grained control of access rights is possible, and preemptive prevention of the risk of information leakage is possible.

[0184] According to embodiments of the present invention, systematic management of approval history is possible through audit responsiveness, clear presentation of the basis for decision-making is possible, and automatic generation and management of audit evidence is possible.

[0185] The device described above may be implemented as a hardware component, a software component, and / or a combination of a hardware component and a software component. For example, the device and components described in the embodiments may be implemented using one or more general-purpose or special-purpose computers, such as a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing unit may execute an operating system (OS) and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing unit may be described as being used as a single unit, but those skilled in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements. For example, the processing unit may include multiple processors or one processor and one controller. In addition, other processing configurations, such as parallel processors, are also possible.

[0186] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or instruct the processing unit independently or collectively. Software and / or data may be embodied in any type of machine, component, physical device, computer storage medium, or device so as to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be distributed over networked computer systems and may be stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

[0187] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. In this case, the medium may continuously store a program executable by a computer, or temporarily store it for execution or download. Additionally, the medium may be various recording or storage means in the form of a single or several hardware combined, and may not be limited to a medium directly connected to a computer system but may exist distributed over a network. Examples of media may include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and media configured to store program instructions, including ROM, RAM, and flash memory. Additionally, other examples of media may include recording or storage media managed by app stores that distribute applications or sites and servers that supply or distribute various other software.

[0188] Although the embodiments have been described above with reference to limited examples and drawings, those skilled in the art can make various modifications and variations from the description above. For example, suitable results can be achieved even if the described techniques are performed in a different order than described, and / or the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents.

[0189] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below.

Claims

1. An AI-based electronic approval pre-verification method performed by a computer device comprising at least one processor, A step of receiving an approval document from an electronic approval system; and A step of performing pre-validation of the aforementioned approval document based on Natural Language Processing (NLP) through integration with an ERP (Enterprise Resource Planning) system or a CRM (Customer Relationship Management) system. Includes, The steps performed above are, A step of identifying regulatory documents related to the above approval document; A step of determining whether the approval document complies with regulations based on the above regulation document; and Step of presenting alternative solutions in case the above approval document violates regulations AI-based electronic approval pre-verification method that performs 2. In Paragraph 1, The above receiving step is, A step of detecting and receiving a new submitted document of the electronic approval system as the above approval document. AI-based electronic approval pre-verification method including 3. In Paragraph 1, The above receiving step is, Step of extracting metadata including the author, approval line, and document type of the above approval document AI-based electronic approval pre-verification method including 4. In Paragraph 1, The above receiving step is, A step of verifying the validity of the document format for the above approval document; and Step of converting the document content of the above approval document into a standardized format AI-based electronic approval pre-verification method including 5. In Paragraph 1, The steps performed above are, A step of performing sentence semantic analysis on the above approval document using a language model; and A step of detecting contextual errors in the content based on the contextual information of the above approval document AI-based electronic approval pre-verification method including 6. In Paragraph 1, The steps performed above are, A step of detecting abnormal patterns in the above approval document through comparison with similar documents. AI-based electronic approval pre-verification method including 7. In Paragraph 1, The steps performed above are, A step of comparing system data retrieved from the above ERP (Enterprise Resource Planning) system or the above CRM (Customer Relationship Management) system with the data of the above approval document AI-based electronic approval pre-verification method including 8. In Paragraph 1, The steps performed above are, A step of extracting review items from the above approval document based on the approver's job duties or authority information. AI-based electronic approval pre-verification method including 9. In Paragraph 1, The steps performed above are, A step of analyzing review patterns for the above approval document based on past approval history. AI-based electronic approval pre-verification method including 10. In Paragraph 1, The steps performed above are, A step of providing the author with the results of the preliminary verification of the aforementioned approval document to provide an opportunity to modify the aforementioned approval document. AI-based electronic approval pre-verification method including 11. In Paragraph 1, The steps performed above are, A step of updating a model for document content analysis by collecting user feedback on the results of the preliminary verification of the above approval document. AI-based electronic approval pre-verification method including 12. In Paragraph 1, The steps performed above are, A step of collecting user feedback on the results of the prior verification of the aforementioned approval document and updating a model for document content analysis through reinforcement learning based on the said user feedback. AI-based electronic approval pre-verification method including 13. In Paragraph 1, The steps performed above are, A step of providing highlighted review items of the aforementioned approval document to the approver of the aforementioned approval document. AI-based electronic approval pre-verification method including 14. In Paragraph 1, The steps performed above are, In providing a pre-verification report including data consistency verification results and regulatory compliance verification results, the step of analyzing the job duties or authority information of each approver included in the approval line of the aforementioned approval document to extract review items for each approver and highlighting the corresponding review items within the aforementioned approval document to provide a differentiated review guide for each approver. AI-based electronic approval pre-verification method including 15. At least one processor implemented to execute readable instructions on a computer device Includes, The above-mentioned at least one processor is, The process of receiving approval documents from an electronic approval system; and A process of performing pre-validation of the aforementioned approval documents based on Natural Language Processing (NLP) through integration with an ERP (Enterprise Resource Planning) system or a CRM (Customer Relationship Management) system Process, The above-mentioned at least one processor is, A process of identifying regulatory documents related to the above approval document; A process of determining whether the above approval document complies with regulations based on the above regulation document; and The process of presenting alternative solutions in case the above approval document violates regulations A computer device that processes.

Images