An AI-based intelligent office automation process management method and system

Abstract

The application relates to the technical field of artificial intelligence and office automation, and discloses an AI-based intelligent office automation process management method and system, wherein the method can recognize non-standard path adjustment of experienced personnel by acquiring operation behavior data of the approval personnel and comparing the operation behavior data with preset process path information. When the occurrence frequency of the non-standard path adjustment reaches a preset threshold value, the system actively provides an interactive interface to the experienced personnel to acquire experience information behind the non-standard operation. The experience information is then converted into structured experience rule segments and displayed as a recommended path in a subsequent process recommendation process.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and office automation, and more specifically, to an AI-based intelligent office automation process management method and system. Background Technology

[0002] Modern enterprise office processes commonly suffer from cross-departmental collaboration barriers and delays caused by manual approvals. This is especially true when dealing with complex form workflows and multi-level approvals. Traditional office automation systems struggle to adapt flexibly to changing business needs, leading to resource waste and delayed decision-making. To address these challenges, enterprises typically deploy intelligent office automation systems. The core objective of these systems is to streamline business form workflows and approvals. Through intelligent judgment mechanisms that learn from established business rules and past approval records, they provide efficient and accurate path suggestions for routine form workflows.

[0003] However, with the rapid development of the company's business and the increasing sophistication of its operating models, the complexity of internal processes has significantly increased. Different business units, when processing similar business forms, develop and frequently update numerous field validity judgment rules due to business segmentation and strict compliance considerations. These department-specific rules are not static; they continuously evolve based on external market changes, new regulatory requirements, or internal policy adjustments. This continuous rule updating and inter-departmental differentiation means that the intelligent judgment mechanism, which initially relied on a relatively unified and stable set of rules, now faces a highly fragmented and constantly changing rule environment.

[0004] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0005] This invention provides an AI-based intelligent office automation process management method and system, aiming to solve the problem that existing intelligent office automation systems cannot keep up with and provide accurate and timely process recommendations when faced with fragmented and constantly changing business rules, implicit experience preferences of senior personnel, and rapid evolution of business processes. This results in poor accuracy and practicality of path recommendations, which seriously restricts the level of intelligent office processes.

[0006] The technical solution of this application is as follows:

[0007] Firstly, this application discloses an AI-based intelligent office automation process management method, specifically including:

[0008] Acquire the operational behavior data of the approvers and store the preset process path information;

[0009] By comparing operational behavior data with preset process path information, operational sequences that differ from the preset process path information are identified as non-standard path adjustments, and the frequency of non-standard path adjustments is statistically analyzed.

[0010] When the frequency of non-standard path adjustments reaches a preset threshold, an interactive interface is provided to senior personnel who perform non-standard path adjustments to obtain experience information related to non-standard path adjustments.

[0011] Transform experiential information into structured fragments of experiential rules;

[0012] Store structured fragments of experience rules, and display these fragments as suggested paths during subsequent process recommendation.

[0013] Record the adoption status of suggested paths and evaluate the efficiency improvement effect of suggested paths. Based on the adoption status and efficiency improvement effect, determine whether the structured empirical rule fragments have passed the validation.

[0014] Integrate validated, structured pieces of empirical rules into the process recommendation logic to generate process recommendation strategies;

[0015] Continuously evaluate the performance of the process recommendation strategy and adjust the strategy based on the evaluation results.

[0016] Through this technical solution, this application can proactively identify non-standard operations of senior personnel and transform their implicit experience into structured rules, which are then integrated into the process recommendation strategy. This achieves dynamic adaptation to the evolution of business processes and the experience preferences of senior personnel, effectively solving the problem of insufficient accuracy and timeliness of recommendations in existing systems under complex and ever-changing business environments.

[0017] Furthermore, based on the above, when the frequency of non-standard path adjustments reaches a preset threshold, an interactive interface is provided to senior personnel performing non-standard path adjustments to obtain relevant experience information, specifically including:

[0018] When senior personnel are about to perform a critical operation, monitor the critical operation and, in conjunction with the business type of the form, the field values ​​already filled in, and the senior personnel's historical operation patterns, determine whether the critical operation will cause a deviation from the preset process path information or involve potential compliance risks.

[0019] When it is determined that there is a deviation between the key operation and the preset process path information or that there is a potential compliance risk, a first prompt box is provided to senior personnel. The first prompt box asks senior personnel about the informal rules or off-process risk mitigation mechanisms on which the key operation is based, based on the business context of the form and the predicted result of the key operation.

[0020] The first prompt box guides senior personnel to clarify the multiple objectives behind key operations, including improving approval efficiency, mitigating supply chain disruption risks, or ensuring compliance with quarterly cumulative procurement requirements.

[0021] After senior personnel complete the declaration of informal rules, off-process risk mitigation mechanisms, and the intention to balance multiple objectives, a risk and compliance assessment is conducted based on the declaration, the business type of the form, the field values ​​filled in, and the known formal and informal rules.

[0022] When the risk and compliance assessment results indicate potential compliance risks, a second prompt box is provided to senior personnel. This second prompt box points out the potential risk points and provides compliance recommendations, which senior personnel can choose to accept and adjust their operations accordingly.

[0023] Through this technical solution, this application can proactively intervene before key operations, guide senior personnel to clarify the underlying logic and multiple objective balancing intentions behind non-standard operations, and conduct real-time risk and compliance assessments. This effectively identifies and avoids potential risks during the experience acquisition phase, ensuring the quality and compliance of the acquired experience and preventing non-compliant or high-risk experiences from being included in recommended strategies.

[0024] Based on this, the adoption status of the suggested paths is recorded, and the efficiency improvement effect of the suggested paths is evaluated. Based on the adoption status and efficiency improvement effect, it is determined whether the structured empirical rule fragments have passed validation, specifically including:

[0025] Identify the triggering conditions contained in the structured experience rule fragments, and extract the business context features of the structured experience rule fragments based on the triggering conditions;

[0026] Based on the characteristics of the business context, filter out form instances that match the triggering conditions from historical operation behavior data;

[0027] Analyze whether any anomalies occurred during the actual workflow of the form instance due to the adoption of similar non-standard paths;

[0028] Calculate the situational risk index of structured empirical rule fragments based on the frequency and severity of abnormal situations.

[0029] Based on the execution actions specified in the structured rule of experience fragments, simulate the execution of the structured rule of experience fragments in test environments of different business departments;

[0030] Assess the impact of simulation execution on different segmented business scenarios and generate a scenario universality assessment report with structured empirical rule fragments;

[0031] When the situational risk index is below the preset threshold and the scenario universality assessment report shows that the structured empirical rule fragment performs well in multiple relevant sub-scenarios, the structured empirical rule fragment is judged to have passed the verification.

[0032] Through this technical solution, this application can conduct multi-dimensional and in-depth verification of empirical rule fragments through situational risk index and scenario universality assessment report, ensuring their safety and effectiveness in practical applications, avoiding the introduction of new risks or reduction of overall process efficiency, thereby improving the reliability of empirical rule fragments.

[0033] Furthermore, the performance of the process recommendation strategy is continuously evaluated, and the strategy is adjusted based on the evaluation results. Specifically, this includes:

[0034] For each integrated process recommendation strategy, identify the associated business departments and the downstream process nodes that may be affected;

[0035] Collect process data from business departments associated with the process recommendation strategy. This process data includes approval time, task backlog, abnormal rejection rate, or compliance check pass rate of downstream departments.

[0036] Maintain a hidden risk rule base, which contains potential compliance risk patterns arising from localized efficiency improvements;

[0037] When a process recommendation strategy improves the efficiency of its main related departments, but leads to increased approval time, higher rejection rate, or triggers risk patterns in the implicit risk rule base, calculate the comprehensive impact index of the process recommendation strategy on overall process efficiency and compliance.

[0038] Based on the comprehensive impact index, adjust the priority of the recommended process strategies, or send early warning notices to relevant department managers.

[0039] Through this technical solution, this application can conduct a cross-departmental, end-to-end macro-evaluation of the process recommendation strategy by introducing an implicit risk rule base and a comprehensive impact index. This effectively avoids the overall efficiency decline or compliance risks that may be caused by local optimization, and ensures the global optimality and sustainability of the process recommendation strategy.

[0040] In some preferred embodiments, when the frequency of non-standard path adjustments reaches a preset threshold, an interactive interface is provided to experienced personnel performing the non-standard path adjustments to obtain relevant experience information, specifically including:

[0041] An interactive interface is provided after senior personnel make adjustments to non-standard paths;

[0042] Based on the business type of the form and the historical operation context of senior personnel, preset experience templates and keyword tags are dynamically generated and displayed in the interactive interface;

[0043] Receive experience templates, keyword tags, or descriptive information selected by senior personnel through the interactive interface as experience information;

[0044] Provides a context-related backtracking function, which displays the operation sequence and related data fragments of senior personnel before this non-standard adjustment;

[0045] Receive annotations or supplementary notes from senior personnel on operation sequences and data fragments as part of their experiential information;

[0046] A conflict pre-check is performed on the experience information input by senior personnel, and the conflict pre-check is compared with the experience information and the existing rule base.

[0047] When a conflict pre-detection identifies a potential conflict or inconsistency, it should alert senior personnel to the potential conflict or inconsistency.

[0048] Receive corrections or supplements to experience-based information from senior personnel.

[0049] Through this technical solution, this application can provide an interactive interface after adjusting non-standard paths, and combine contextual backtracking function and conflict pre-detection mechanism to guide senior personnel to describe experience information more accurately and completely, and to discover and correct potential conflicts in a timely manner, thereby improving the efficiency and quality of experience information acquisition.

[0050] More specifically, transforming experiential information into structured fragments of experiential rules includes:

[0051] Receive experience information;

[0052] Sentence segmentation and word recognition are performed on experiential information to obtain language units;

[0053] Based on a pre-defined vocabulary and synonym set for the business domain, language units are standardized to obtain standardized language units.

[0054] Identify the business entities contained in the experience information, and determine the type and attributes of the business entities based on the business ontology definition;

[0055] Analyze the syntactic structure between language units in experiential information to identify conditions, actions, and results;

[0056] Based on conditions, actions, and results, and combined with preset rule templates, experiential information is populated into structured rule fragments;

[0057] When experience information cannot be directly mapped to the rule template, an interactive clarification interface is provided to senior personnel, which displays the rule fragments that the system initially understands.

[0058] Receive corrections or additions to rule fragments from senior personnel.

[0059] Through this technical solution, this application can efficiently and accurately transform unstructured senior experience information into structured rule fragments that can be understood and executed by the system through multi-step natural language processing and interactive clarification mechanisms, thus solving the problem that experience information is difficult to be learned and utilized by the system.

[0060] Preferably, the adoption status of suggested paths is recorded, and the efficiency improvement effect of suggested paths is evaluated. Based on the adoption status and efficiency improvement effect, it is determined whether the structured empirical rule fragment passes the validation, specifically including:

[0061] Identify the user roles and department information for the adoption of suggested paths;

[0062] Determine whether the user who adopted the recommendation is a senior or highly influential person;

[0063] If so, monitor the adoption rate trend of the suggested path among non-senior or non-high-influence personnel;

[0064] When senior staff adopt a proposal and the adoption rate of non-senior staff increases significantly, the adoption behavior is marked as "affected adoption".

[0065] Provide a feedback interface to users who adopt the suggested path, and ask them for the basis of their adoption decision;

[0066] Receive decision-making information provided by users;

[0067] The adoption behavior is classified according to the type and content of the decision-making basis;

[0068] When an adoption is marked as not "affected adoption" and the decision is based on actual business value, it is counted as a valid adoption. Based on the valid adoption and the efficiency improvement effect, it is determined whether the structured rule of experience fragment has passed the validation.

[0069] Through this technical solution, this application can more accurately assess the actual business value and universality of empirical rule fragments by distinguishing the adoption behavior of different user roles and combining it with the analysis of decision-making basis, thereby avoiding "blind adoption" caused by the influence of senior personnel and improving the objectivity and reliability of the verification results.

[0070] Based on the above, the performance of the process recommendation strategy is continuously evaluated, and the strategy is adjusted according to the evaluation results, specifically including:

[0071] For each integrated process recommendation strategy, collect objective performance data on the adoption rate, process duration, and error rate of the process recommendation strategy during execution.

[0072] After senior personnel complete the process operation recommended by the process strategy, a feedback interface is provided. This feedback interface generates evaluation dimensions related to the business context of the senior personnel's operation. These evaluation dimensions include efficiency improvement, risk control, and ease of operation, and provide quantitative scoring options and text input boxes.

[0073] Senior staff can evaluate the performance of the process recommendation strategy across evaluation dimensions using quantitative scoring options and text input boxes.

[0074] Keyword extraction and sentiment analysis are performed on the text feedback entered by senior personnel to transform the text feedback into structured feedback data;

[0075] Structured feedback data is integrated with objective performance data to calculate the overall performance score of the process recommendation strategy;

[0076] When the overall performance score is lower than a preset threshold, the key dimensions that cause the overall performance score to drop are identified.

[0077] Based on key dimensions, generate adjustment suggestions for the process recommendation strategy. These suggestions include adjusting the triggering conditions, recommendation priority, or related parameters of the process recommendation strategy.

[0078] Receive confirmation or corrections from senior staff regarding the proposed adjustments.

[0079] Through this technical solution, this application can comprehensively and multidimensionally evaluate the process recommendation strategy by integrating objective performance data and subjective feedback from senior personnel, and generate specific adjustment suggestions based on the evaluation results, thereby realizing the continuous optimization and adaptive adjustment of the process recommendation strategy and ensuring its long-term effectiveness.

[0080] As a technological improvement, structured feedback data is integrated with objective performance data to calculate a comprehensive performance score for the recommendation strategy, specifically including:

[0081] The data provided by senior staff is timestamped to obtain calibrated feedback data.

[0082] Perform consistency cross-validation on the calibrated feedback data and objective performance data to identify any inconsistencies between the feedback data and objective performance data that fall outside a preset threshold range;

[0083] When an inconsistency is detected between the feedback data and the objective performance data that is outside the preset threshold range, the abnormal data marking mechanism is activated to mark the inconsistent data pairs as data to be reviewed.

[0084] Based on the severity of data inconsistency or temporal misalignment, the weights of data pairs in the fusion computation are dynamically adjusted, and the overall performance score of the process recommendation strategy is calculated.

[0085] Generate a data inconsistency report.

[0086] Through this technical solution, this application can effectively handle the inconsistencies that may exist between subjective feedback and objective data by using timestamp calibration, consistency cross-validation, and dynamic weight adjustment. This improves the accuracy and robustness of the comprehensive performance score calculation and avoids evaluation bias caused by data quality issues.

[0087] Secondly, this application also discloses an AI-based intelligent office automation process management system, specifically including:

[0088] The identification end is used to acquire the operation behavior data of the approvers and store the preset process path information; compare the operation behavior data with the preset process path information, identify the operation sequence that differs from the preset process path information as non-standard path adjustment, and count the frequency of non-standard path adjustment.

[0089] The interactive interface is used to provide an interactive interface to senior personnel who perform non-standard path adjustments when the frequency of non-standard path adjustments reaches a preset threshold, so as to obtain the experience information corresponding to the non-standard path adjustments; to convert the experience information into structured experience rule fragments; to store the structured experience rule fragments, and to display the structured experience rule fragments as suggested paths in the subsequent process recommendation process.

[0090] The judgment end is used to record the adoption status of suggested paths and evaluate the efficiency improvement effect of suggested paths. Based on the adoption status and efficiency improvement effect, it determines whether the structured empirical rule fragments have passed the validation.

[0091] The adjustment module integrates validated, structured rule fragments into the process recommendation logic to generate a process recommendation strategy; it continuously evaluates the performance of the process recommendation strategy and adjusts it based on the evaluation results.

[0092] Through this technical solution, this application can achieve automated identification, acquisition, transformation, verification and integration of the tacit experience of senior personnel through a systematic modular design, and can continuously evaluate and adjust process recommendation strategies, thereby building an intelligent office automation management system that can dynamically adapt to business changes and experience evolution.

[0093] Beneficial effects

[0094] This application provides an AI-based intelligent office automation process management method and system. By acquiring operational behavior data of approvers and comparing it with preset process path information, the system can identify non-standard path adjustments made by experienced personnel. When the frequency of these non-standard path adjustments reaches a preset threshold, the system proactively provides an interactive interface to experienced personnel to obtain the experience information behind their non-standard operations. This experience information is then transformed into structured experience rule fragments and displayed as suggested paths in subsequent process recommendation. The system further records the adoption status and efficiency improvement effect of suggested paths, and determines whether the experience rule fragments have passed verification. Verified experience rule fragments are integrated into the process recommendation logic to generate a process recommendation strategy, and the strategy is continuously evaluated and adjusted.

[0095] This method effectively solves the problem that existing intelligent office automation systems cannot keep up with and provide accurate and timely process recommendations when faced with fragmented and constantly changing business rules, implicit experience preferences of senior personnel, and rapid evolution of business processes. Specifically, this application achieves excellent technical results through the following methods:

[0096] First, this application proactively captures non-standard operational behaviors of senior personnel and treats them as potential optimization experiences, rather than simply as deviations. This overcomes the limitation of existing systems in extracting implicit experiences from unstructured data.

[0097] Secondly, by providing an interactive interface for experienced personnel, this application can directly access the deep experiential information behind non-standard path adjustments, including their decision-making logic, multi-objective balancing intentions, and potential risk mitigation mechanisms. This enables the system to understand "why" a particular path is taken, rather than merely observing "what," thereby bridging the gap between the system's recommended path and the path optimized by experienced personnel in practice.

[0098] Furthermore, transforming the acquired experiential information into structured fragments of experiential rules enables these implicit experiences to be understood, stored, and reused by the system, providing a clear and learnable data source for intelligent judgment mechanisms.

[0099] Furthermore, by evaluating the adoption of suggested paths and the efficiency improvement effects, and by determining whether the empirical rule fragments have passed verification, this application ensures that the empirical rules integrated into the process recommendation logic are tested in practice and have practical business value, thus avoiding the introduction of inefficient or non-compliant practices.

[0100] Finally, the mechanism for continuously evaluating and adjusting the process recommendation strategy enables the system to dynamically adapt to changes in the external market environment, adjustments in the enterprise's internal organizational structure, and the continuous evolution of senior personnel's experience. This ensures the accuracy and timeliness of the process recommendation strategy and effectively solves the problems of recommendation bias and computational pressure caused by frequent relearning in the existing system due to business evolution.

[0101] In summary, this application makes the tacit experience of senior personnel explicit, structured, and integrated into intelligent office automation process management, thereby achieving dynamic adaptation to business process evolution and experience preferences, significantly improving the accuracy and practicality of process recommendations, and thus comprehensively enhancing the level of intelligence in office processes. Attached Figure Description

[0102] Figure 1 This is a flowchart illustrating an AI-based intelligent office automation process management method provided in an embodiment of the present invention.

[0103] Figure 2 This is a schematic diagram of the structure of an AI-based intelligent office automation process management system provided in an embodiment of the present invention. Detailed Implementation

[0104] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0105] Reference Figure 1 , Figure 1 This is a flowchart illustrating an AI-based intelligent office automation process management method provided in an embodiment of the present invention, including:

[0106] S11, acquire the operational behavior data of the approver and store the preset process path information;

[0107] S12, compare the operation behavior data with the preset process path information, identify operation sequences that differ from the preset process path information as non-standard path adjustments, and count the frequency of occurrence of the non-standard path adjustments.

[0108] S13, when the frequency of the non-standard path adjustment reaches a preset threshold, an interactive interface is provided to the senior personnel who perform the non-standard path adjustment to obtain the experience information corresponding to the non-standard path adjustment.

[0109] S14, the experience information is transformed into structured fragments of experience rules;

[0110] S15, store the structured experience rule fragments, and display the structured experience rule fragments as suggested paths in the subsequent process recommendation process;

[0111] S16, record the adoption status of the suggested path, evaluate the efficiency improvement effect of the suggested path, and determine whether the structured empirical rule fragment passes the verification based on the adoption status and the efficiency improvement effect;

[0112] S17, integrate the validated structured empirical rule fragments into the process recommendation logic to generate a process recommendation strategy; continuously evaluate the performance of the process recommendation strategy and adjust the process recommendation strategy based on the evaluation results.

[0113] This application acquires and analyzes operational behavior data of approvers to identify non-standard path adjustments and proactively seeks experience information from senior personnel, transforming it into structured experience rule fragments. These experience rule fragments, after verification, are integrated into the process recommendation logic to form a process recommendation strategy. Through continuous evaluation and adjustment, this application can adapt to the evolution of business processes in a timely manner, effectively bridging the gap between the system's recommended path and the actual optimal operational path, and significantly improving the accuracy and timeliness of path recommendation in the intelligent office automation system.

[0114] The "operational behavior data" mentioned in this application refers to records of a series of operations performed by approvers in the office automation system, such as approval, form filling, and task transfer. This data typically includes operation time, operation type, relevant form information, and approval comments. "Preset process path information" refers to standard business process templates predefined by the system or configured by the administrator, which specify the legal flow order and approval nodes for specific business forms. "Non-standard path adjustment" refers to the sequence of operations performed by approvers that deviates from the preset process path but still achieves the business objective. "Experience information" refers to the implicit knowledge, such as informal rules, risk considerations, and efficiency optimizations, that experienced personnel rely on when adjusting non-standard paths. "Structured experience rule fragments" are rules that transform unstructured experience information into a condition-action-result form that machines can understand and execute. "Suggested path" refers to a potentially better process path recommended to the user by the system based on verified experience rule fragments. "Process recommendation strategy" is an algorithm or model that integrates multiple experience rule fragments and formal rules to guide the system in intelligent path recommendation. The implementation environment of this application is typically an intelligent office automation system within an enterprise, which is deployed on a server cluster and interacts with approvers through client applications or web interfaces.

[0115] In practical implementation, the first step is to acquire the operational behavior data of approvers and store the preset process path information. Operational behavior data can be collected in real-time or in batches through system logs, user interface event listeners, and other methods. For example, when approvers submit, reject, transfer, or modify forms in the system, the system records detailed information about these operations, including the operator's identity, operation time, the type of form involved, changes in form field values, and the operation result. The preset process path information is typically stored in the database in the form of flowcharts, state machine models, or rule sets. This information defines the various nodes, flow conditions, and approval permissions of the standard business process.

[0116] Next, the system compares the acquired operational behavior data with the stored preset process path information. This comparison process can be implemented using pattern matching algorithms, for example, aligning the actual operation sequence of the approver with the preset process path to identify any discrepancies. Any operation sequence that does not conform to the preset process path information, such as skipping an approval step, changing the approval order, or using an unconventional approver, will be identified as a non-standard path adjustment. Subsequently, the system will count the frequency of these non-standard path adjustments. For example, a counter can be used to accumulate each identified non-standard path adjustment and periodically update its frequency.

[0117] When the frequency of non-standard path adjustments reaches a preset threshold, the system will provide an interactive interface to the senior personnel who performed the adjustment, allowing them to obtain relevant experience information. For example, if a specific non-standard path adjustment occurs more than 10 times within a month, and is performed by the same senior approver, the system will trigger this interaction. This interface can be a pop-up window or a dedicated feedback page, guiding the senior personnel to input the reasons for the adjustment, the informal rules followed, the risk factors considered, or the desired efficiency goals.

[0118] Subsequently, the acquired experience information will be transformed into structured fragments of experience rules. Since the experience information input by senior personnel is typically unstructured text in natural language form, it requires parsing using Natural Language Processing (NLP) techniques. For example, Named Entity Recognition (NER) technology can be used to extract business entities (such as "purchase order" and "supplier"), and relation extraction technology can be used to identify conditions (such as "amount greater than 100,000"), actions (such as "submit directly to the general manager for approval"), and results (such as "speed up the approval process"). Finally, this information will be mapped to a pre-defined rule template, forming structured rule fragments that follow the pattern "if [condition], then [action], to achieve [result]".

[0119] These structured rule-of-fact fragments are stored and displayed as suggested paths during subsequent process recommendations. For example, when a new business form enters the approval process, the system matches the stored structured rule-of-fact fragments with the current form's business context and field values. If a match is successful, the system will display the suggested path corresponding to that rule-of-fact fragment to the approver in a visual manner (such as a flowchart or text prompt) as an optional optimization path.

[0120] To ensure the effectiveness and reliability of suggested paths, the system records their adoption status and evaluates their efficiency improvement. Adoption status can be recorded through user actions such as clicking, selecting, or actually executing suggested paths. Efficiency improvement can be quantified by comparing metrics such as process time, resource consumption, and error rate before and after adopting suggested paths. For example, if adopting a suggested path reduces the average approval time by 20%, its efficiency improvement is considered significant. Based on adoption status and efficiency improvement, the system determines whether the structured rule of thumb fragment passes validation. For instance, only when the adoption rate of a suggested path reaches a certain percentage and the efficiency improvement is significant will the rule of thumb fragment be marked as "validated."

[0121] Validated, structured pieces of empirical rules will be integrated into the process recommendation logic to generate process recommendation strategies. This means that these empirical rule pieces will no longer be merely suggestions, but will become a core component of the system's intelligent judgment and recommendation. For example, validated empirical rule pieces can be added as new rules to the rule engine, or used to train machine learning models, enabling them to more accurately predict and recommend optimal process paths.

[0122] Finally, the system continuously evaluates the performance of the generated process recommendation strategies and adjusts them based on the evaluation results. Evaluation metrics may include recommendation accuracy, user satisfaction, overall process efficiency, and compliance risks. For example, if a process recommendation strategy is found to lead to a high error rate or user complaints, the system will automatically or semi-automatically adjust the priority, triggering conditions, or associated parameters of that strategy based on the evaluation results to optimize its performance.

[0123] Compared to existing technologies, the core innovation of this application lies in its ability to proactively acquire, structure, verify, and integrate the tacit experience of senior personnel, and the adaptive, self-learning process recommendation mechanism built upon this foundation. Traditional intelligent office automation systems mainly rely on preset formal rules and historical structured data for process recommendation, making it difficult to adapt to rapid business changes and non-standard optimizations based on experience by senior personnel. When business rules are frequently updated, there are significant differences between departments, and senior personnel adopt non-standard but efficient temporary paths, existing systems often lack the understanding and learning ability of these tacit experiences, leading to deviations between recommended paths and the actual optimal operating paths, and even "outdated" recommendations.

[0124] Traditional intelligent office automation process management methods may rely solely on simple interactive interfaces to collect experience from senior staff regarding non-standard path adjustments. This can lead to incomplete information, a lack of necessary business context, or failure to fully reveal the deeper considerations behind these adjustments, such as the intention to balance multiple objectives. If these issues are not addressed, the accuracy and effectiveness of subsequently transforming this experience into structured rule fragments will be affected, consequently impacting the quality and reliability of process recommendation strategies.

[0125] In this regard, this application further proposes the following steps for obtaining the empirical information corresponding to the aforementioned non-standard path adjustment:

[0126] When the senior personnel are about to perform a key operation, monitor the key operation and, in conjunction with the business type of the form, the field values ​​already filled in, and the senior personnel's historical operation patterns, determine whether the key operation will cause a deviation from the preset process path information or involve potential compliance risks.

[0127] When it is determined that the key operation deviates from the preset process path information or involves potential compliance risks, a first prompt box is provided to the senior personnel. The first prompt box asks the senior personnel about the informal rules or off-process risk mitigation mechanisms on which the key operation is based, based on the business context of the form and the predicted result of the key operation.

[0128] In the first prompt box, the senior personnel are guided to understand the multiple objectives behind the key operation, including improving approval efficiency, avoiding supply chain disruption risks, or ensuring compliance with quarterly cumulative procurement requirements.

[0129] After the senior personnel complete the declaration of the informal rules, the out-of-process risk mitigation mechanism, and the intention to balance multiple objectives, a risk and compliance assessment is conducted based on the declaration, the business type of the form, the filled-in field values, and the known formal and informal rules.

[0130] When the risk and compliance assessment results indicate potential compliance risks, a second prompt box is provided to the senior personnel. The second prompt box points out the potential risk points and provides compliance suggestions, which the senior personnel can choose to accept and adjust their operations accordingly.

[0131] Specifically, critical operations refer to actions that have a significant impact on the business process or may cause significant deviations from the process path, such as special approvals, amount adjustments, and supplier changes in the approval process. Monitoring these critical operations can be achieved through system log analysis, user behavior tracking, or a pre-defined rule engine.

[0132] The business type of the form can be understood as the business function or business area that the form carries, such as a purchase requisition form, expense reimbursement form, or contract approval form. Its purpose is to provide contextual information for subsequent judgments and prompts. The filled field values ​​refer to the data that has already been entered or selected in the form, such as the purchase amount, supplier name, and approval comments. This data is an important basis for judging whether the operation deviates from the preset path and assessing risks. The senior staff's historical operation patterns refer to the habitual operation paths, decision-making preferences, and handling methods for non-standard situations that the senior staff have used in the past when handling similar business. This is analyzed and modeled using machine learning models to more accurately predict the intent and potential impact of their current operations.

[0133] In practical applications, the first prompt is an interactive window that automatically pops up when the system detects that a senior staff member is about to perform a critical operation that may deviate from the preset path or pose a risk. This prompt aims to proactively guide the senior staff member to provide the underlying reasons for their non-standard operation, such as asking if they followed any informal rules not formally recorded by the system, or whether they took risk mitigation measures outside of the established procedures. For example, a senior staff member might have temporarily taken a non-standard operation to avoid an emergency, while simultaneously mitigating risk through methods such as telephone communication.

[0134] Furthermore, the intention to balance multiple objectives refers to the fact that when senior personnel perform non-standard operations, they may not be pursuing a single objective, but rather making trade-off decisions based on a comprehensive consideration of multiple interrelated or even potentially conflicting objectives. For example, to improve approval efficiency, senior personnel may skip certain routine steps while simultaneously ensuring compliance through other means to mitigate supply chain disruption risks, or they may adjust procurement plans to meet quarterly cumulative procurement targets. Guiding senior personnel to clarify these intentions helps the system to more comprehensively understand the rationale and context of non-standard operations.

[0135] Furthermore, risk and compliance assessment refers to the system's comprehensive risk and compliance analysis of senior personnel's non-standard operations after obtaining their statements, taking into account the business type of the form, the values ​​of the fields already filled in, and known formal and informal rules. This can include assessing the legality of the operation, potential financial risks, operational risks, information security risks, and whether it complies with internal policies and external regulations.

[0136] When the assessment results indicate potential compliance risks, the system will provide a second prompt box to senior personnel. This prompt box will clearly point out the specific risk points, such as "This purchase amount exceeds the department's budget limit, which may lead to budget overruns," and provide corresponding compliance suggestions, such as "It is recommended to adjust the purchase amount to within the budget or submit a special budget application." Senior personnel can choose to accept the suggestions and adjust their actions accordingly, thereby effectively reducing potential risks while gaining experience and information.

[0137] This application's solution proactively identifies potential non-standard path adjustments and compliance risks by monitoring and predicting key operations performed by senior personnel, combined with multi-dimensional information (such as form business types, field values, and historical operation patterns). Through this technical solution, the application significantly improves the quality and depth of acquired non-standard path adjustment experience information. Specifically, by intelligently monitoring and predicting risks before key operations, the system proactively captures non-standard behaviors of senior personnel and provides contextualized guided interaction, enabling them to more comprehensively explain their decision-making basis, including informal rules, off-process risk mitigation mechanisms, and the intention to balance multiple objectives. This approach avoids the ambiguity, incompleteness, or lack of deep logical support that may exist in traditional methods of experience information. Furthermore, by integrating risk and compliance assessment and recommendation mechanisms into the experience information acquisition process, the collected experience rule fragments possess higher compliance and reliability from the source, effectively reducing the potential risks of converting non-standard experience into recommended strategies. Therefore, this application can provide more accurate, reliable, and secure input for subsequent experience rule fragment conversion and process recommendation strategy generation, thereby improving the intelligence level and business value of the entire intelligent office automation process management system.

[0138] In some of the embodiments described above, this application proposes an AI-based intelligent office automation process management method. This method determines whether a structured piece of empirical rule data passes verification by recording the adoption status of suggested paths and evaluating their efficiency improvement effects. However, relying solely on adoption status and efficiency improvement effects for verification may not fully assess the potential risks of empirical rule data or its universality across different business scenarios. For example, a non-standard path that performs efficiently in a specific department may trigger compliance risks or lead to new efficiency bottlenecks in other departments or specific business scenarios. If these issues are not addressed, immature or risky empirical rules may be integrated into the process recommendation logic, thereby affecting the overall stability and compliance of the process. Therefore, this application further proposes a more rigorous verification mechanism that comprehensively considers situational risks and scenario universality to determine whether a structured piece of empirical rule data passes verification.

[0139] The above records the adoption status of the suggested paths and evaluates the efficiency improvement effect of the suggested paths. Based on the adoption status and the efficiency improvement effect, it determines whether the structured empirical rule fragment passes the validation, including:

[0140] Identify the triggering conditions contained in the structured experience rule fragments, and extract the business context features of the structured experience rule fragments based on the triggering conditions;

[0141] Based on the business context characteristics, form instances that match the triggering conditions are selected from historical operation behavior data;

[0142] Analyze whether any abnormal situations have occurred during the actual flow of the form instance due to the adoption of similar non-standard paths;

[0143] Calculate the situational risk index of the structured empirical rule fragment based on the frequency and severity of the abnormal situation;

[0144] Based on the execution actions specified in the structured empirical rule fragments, the execution of the structured empirical rule fragments is simulated in test environments of different business departments;

[0145] Assess the impact of the simulation execution on different segmented business scenarios, and generate a scenario universality assessment report of the structured empirical rule fragments;

[0146] When the situational risk index is lower than a preset threshold and the scenario universality assessment report shows that the structured empirical rule fragment performs well in multiple relevant sub-scenarios, the structured empirical rule fragment is judged to have passed the verification.

[0147] Specifically, identifying the triggering conditions contained in structured rule fragments refers to parsing the specific conditions under which the rule fragment takes effect from its description, such as a specific business type, a range of form field values, or the role of the operator. Based on these triggering conditions, the business context characteristics to which the rule fragment applies can be extracted, such as "the purchase amount is greater than 100,000 yuan and the supplier is of category A".

[0148] Among them, selecting form instances that match the triggering conditions from historical operational behavior data based on business context characteristics can be understood as finding historical business records that are similar to or the same as the business scenario described by the current empirical rule fragment to be verified in a massive amount of historical data.

[0149] In practical applications, analyzing whether form instances have encountered any abnormal situations due to the adoption of similar non-standard paths during the actual process means checking these selected historical form instances to see if they have encountered problems such as rejection, rework, compliance warnings, or the final result not meeting expectations during the process.

[0150] Furthermore, based on the frequency and severity of anomalies, a situational risk index is calculated for structured rule fragments. The purpose is to quantify the risks that a rule fragment may pose in a specific context. For example, if similar operations have frequently led to compliance issues in the past, their situational risk index will be higher.

[0151] Furthermore, simulating the execution of structured rule fragments in test environments of different business departments, based on the execution actions specified in the structured rule fragments, refers to simulating the actual operating effect of the rule fragments in different departments or different business processes through a simulation system or sandbox environment without affecting actual business operations.

[0152] Therefore, this assessment evaluates the impact of simulated execution on different segmented business scenarios and generates a scenario universality assessment report for structured rule fragments. The purpose is to comprehensively understand the applicability and potential impact of these rule fragments. The report details in which scenarios they perform well and in which scenarios they may encounter problems.

[0153] Finally, when the situational risk index is below the preset threshold and the scenario universality assessment report shows that the structured empirical rule fragment performs well in multiple relevant sub-scenarios, the structured empirical rule fragment is judged to have passed the verification. This means that the empirical rule fragment is not only risk-controllable in historical scenarios, but also has good applicability and stability in a variety of potential application scenarios in the future.

[0154] This application's solution addresses the limitations of relying solely on adoption and efficiency improvement for verification by introducing a situational risk index and a scenario universality assessment report. Through these technical solutions, this application significantly improves the comprehensiveness and accuracy of structured empirical rule fragment verification. The calculation of the situational risk index effectively identifies and mitigates potential compliance risks and operational anomalies, preventing the integration of potentially negative non-standard paths into core business processes. Simultaneously, the scenario universality assessment report ensures that empirical rule fragments are not only effective in specific scenarios but also operate stably across multiple relevant sub-scenarios, greatly expanding their applicability and value. This more rigorous verification process makes the final integrated process recommendation strategy more robust and reliable, effectively reducing system operational risks and improving the overall quality and user trust of intelligent office automation process management.

[0155] The above-mentioned continuous evaluation of the performance of the process recommendation strategy, and adjustment of the process recommendation strategy based on the evaluation results, specifically includes:

[0156] For each integrated process recommendation strategy, identify the associated business departments and the downstream process nodes that may be affected;

[0157] Collect process data from business departments associated with the process recommendation strategy. The process data includes approval time, task backlog, abnormal rejection rate, or compliance check pass rate of downstream departments.

[0158] Maintain a hidden risk rule base, which contains potential compliance risk patterns caused by local efficiency improvements;

[0159] When the process recommendation strategy improves the efficiency of its main related departments, but causes the approval time of the downstream departments to increase, the abnormal rejection rate to rise, or triggers the risk patterns in the implicit risk rule base, the comprehensive impact index of the process recommendation strategy on the overall process efficiency and compliance is calculated.

[0160] Based on the comprehensive impact index, adjust the priority of the process recommendation strategy, or send an early warning notification to relevant department managers.

[0161] Specifically, when evaluating a recommended process strategy, it is first necessary to identify the business departments associated with the strategy and the downstream process nodes that may be affected. This ensures that the scope of the evaluation is not limited to the departments directly affected by the strategy, but extends to the entire affected business chain. For example, a recommended strategy for a procurement approval process may primarily involve the procurement department, but its downstream process nodes may include payment approval in the finance department and receiving confirmation in the inventory management department.

[0162] Furthermore, to comprehensively assess the impact of the strategy, it is necessary to collect process data from the business departments associated with the recommended process strategy. This data may include approval times in downstream departments, used to measure whether the strategy leads to a decrease in efficiency in downstream processes; task backlog, reflecting whether the workload of downstream departments has increased due to the strategy; abnormal rejection rates, indicating whether the strategy has increased errors or non-compliant operations in the process; and compliance check pass rates, directly assessing the impact of the strategy on compliance. These multi-dimensional data collectively constitute the quantitative basis for assessing the overall impact of the strategy.

[0163] Furthermore, this application maintains a hidden risk rule base. This rule base aims to identify potential compliance risk patterns that may arise from localized efficiency improvements. For example, to expedite approvals, a strategy might recommend skipping certain routine risk review steps, which improves efficiency in the short term but could lead to compliance loopholes in the long run. The hidden risk rule base, through pre-defined pattern recognition, can promptly identify such potential risks.

[0164] Therefore, when it is found that while the process recommendation strategy improves the efficiency of its primary related department, it leads to increased approval time, higher rejection rates, or triggers risk patterns in the implicit risk rule base of downstream departments, the system will calculate the comprehensive impact index of the process recommendation strategy on overall process efficiency and compliance. This index is a comprehensive quantitative indicator used to measure the trade-offs of the strategy on a global scale; for example, it can be weighted by combining the positive effects of efficiency improvement and the negative effects of increased risk.

[0165] Finally, based on the comprehensive impact index, the system can take corresponding adjustment measures. For example, the priority of the process recommendation strategy can be adjusted so that it is no longer recommended in certain specific situations, or a warning notification can be sent to relevant department managers to remind them of potential problems that the strategy may cause, so that timely manual intervention or strategy correction can be carried out.

[0166] This application's solution expands the evaluation scope of process recommendation strategies from a single department to the entire business process chain, and introduces multi-dimensional data collection and an implicit risk rule base, thereby solving the problem of global imbalance caused by local optimization in traditional evaluation methods. Through the above technical solution, this application effectively avoids the negative impact on the overall business process or the introduction of new compliance risks while improving local efficiency. This solution ensures that the optimization of process recommendation strategies is based on a global perspective, thereby improving the robustness, compliance, and sustainability of the entire intelligent office automation process management. Furthermore, by promptly identifying and warning of potential problems, this application helps managers make more informed decisions, avoiding damage to the overall interests of the enterprise due to blindly pursuing local efficiency, and significantly enhancing the practical value and reliability of process recommendation strategies.

[0167] Traditional intelligent office automation process management methods may only provide a generic interface when acquiring experience information related to adjusting non-standard paths. This could lead to experienced personnel lacking necessary contextual support when describing their experience, or struggling to accurately and completely express their informal rules or risk mitigation mechanisms. If these issues are not addressed, the acquired experience information may be of low quality, ambiguous, or conflict with existing rules, thus affecting the conversion of subsequent experience rule fragments and the effectiveness of process recommendation strategies.

[0168] In response, this application proposes an optimized method for obtaining experience information corresponding to non-standard path adjustments. This method provides a contextualized interactive interface after senior personnel make non-standard path adjustments, and supplements it with contextual backtracking and conflict pre-detection mechanisms to ensure that the obtained experience information is more accurate, complete, and of higher quality.

[0169] When the frequency of non-standard path adjustments reaches a preset threshold, an interactive interface is provided to experienced personnel performing the non-standard path adjustments to obtain relevant experience information, including:

[0170] The interactive interface is provided after the senior personnel make the adjustment to the non-standard path;

[0171] Based on the business type of the form and the historical operation context of the senior personnel, preset experience templates and keyword tags are dynamically generated and displayed in the interactive interface;

[0172] The experience information is received from the senior personnel through the interactive interface, including the experience template, the keyword tags, or the descriptive information they select.

[0173] Provides a context-related backtracking function, which displays the senior personnel's operation sequence and related data fragments before this non-standard adjustment;

[0174] Receive annotations or supplementary notes from the senior personnel on the operation sequence and the data segment as part of the experience information;

[0175] A conflict pre-detection is performed on the experience information input by the senior personnel, and the conflict pre-detection compares the experience information with the existing rule base;

[0176] When the conflict pre-detection detects potential conflicts or inconsistencies, the senior personnel are alerted to the potential conflicts or inconsistencies.

[0177] Receive corrections or supplements from the senior personnel regarding the experience information.

[0178] Specifically, after a senior employee completes a non-standard path adjustment, the system immediately provides an interactive interface. This interface is not static but dynamically generates a series of preset experience templates and keyword tags based on the business type of the form being processed and the senior employee's past operating habits and patterns. These templates and tags are designed to guide senior employees to more efficiently and accurately describe the reasons and basis for their non-standard adjustments. For example, for a purchase request, keyword tags such as "urgent purchase," "supplier change," and "cost optimization" may be generated. Senior employees can provide their experience information by selecting these templates and tags or by directly entering detailed descriptive information.

[0179] The context-based backtracking function can be understood as a mechanism for reproducing historical operational trajectories. When senior personnel need to describe their non-standard adjustments, this function displays a series of operational sequences and related business data fragments preceding the current non-standard adjustment. For example, if a senior personnel modifies a purchase amount during the approval process, the context-based backtracking function will display information such as supplier quotations and historical purchase records viewed before the modification. Its purpose is to help senior personnel recall and accurately supplement their decision-making basis, avoiding memory fuzziness or information omissions due to time intervals. Senior personnel can directly annotate or supplement these backtracked operational sequences and data fragments, further enriching the details of their experience.

[0180] In practical applications, conflict pre-checking of experience information input by senior personnel refers to the system comparing this information with existing formal and informal rule bases. For example, if a non-standard path declared by a senior personnel contradicts a mandatory compliance requirement or logically conflicts with a validated rule of experience, the system will immediately identify these potential conflicts or inconsistencies. The aim is to promptly identify and resolve potential problems before experience information is adopted and transformed, avoiding the introduction of erroneous or contradictory rules. When the system detects a potential conflict, it will alert the senior personnel through prompts, indicating the specific points of conflict and providing possible corrective suggestions. The senior personnel can then revise or supplement the experience information based on these suggestions to ensure its accuracy and consistency.

[0181] This application's solution ensures the timeliness and accuracy of experience information acquisition by providing an interactive interface immediately after senior personnel adjust non-standard paths. Dynamically generating experience templates and keyword tags, combined with the business type of the form and the senior personnel's historical operational context, provides highly relevant guidance to senior personnel, thereby reducing their cognitive burden in describing experience and improving the structure and completeness of the acquired information. The contextual backtracking function effectively helps senior personnel recall and supplement their decision-making basis by reproducing the operation sequence and related data fragments, ensuring the contextual integrity of the experience information. Furthermore, a conflict pre-check mechanism verifies the validity of experience information before it is adopted, promptly identifying and resolving potential conflicts or inconsistencies with existing rules, thus guaranteeing the quality and reliability of experience information and preventing the introduction of erroneous or contradictory experiences into the process recommendation logic.

[0182] Through the aforementioned technical solutions, this application significantly improves the quality and efficiency of acquiring experience information for non-standard path adjustments. Compared to traditional methods that only provide a general interactive interface, this application, through contextual guidance and backtracking, enables senior personnel to express their informal rules and decision-making logic more accurately and completely, reducing information omissions and ambiguities. In particular, the introduction of a conflict pre-detection mechanism effectively avoids incorporating experience information that conflicts with or is inconsistent with existing rules into the system, thereby ensuring the accuracy and reliability of subsequent experience rule fragment transformations and process recommendation strategies. This reduces potential risks caused by introducing inappropriate experience and lays a solid foundation for building a high-quality intelligent office automation process management system.

[0183] In some of the embodiments described above in this application, the experience information of senior personnel is transformed into structured fragments of experience rules. However, during its implementation, if the experience information is not accurately and consistently transformed into machine-understandable structured rules, it may lead to deviations in subsequent process recommendation strategies or even introduce new operational risks. Traditionally, this transformation may rely on manual interpretation and coding, which is inefficient and prone to subjective errors, thereby affecting the quality of experience rules and the reliability of automated applications.

[0184] In this regard, this application further proposes a step for transforming the experiential information into structured fragments of experiential rules, including:

[0185] Receive the experience information;

[0186] The empirical information is segmented into sentences and words are identified to obtain language units;

[0187] Based on a pre-defined business domain vocabulary and a set of synonyms, the language units are standardized to obtain standardized language units.

[0188] Identify the business entities contained in the experience information, and determine the type and attributes of the business entities according to the business ontology definition;

[0189] Analyze the syntactic structure between language units in the empirical information to identify conditions, actions, and results;

[0190] Based on the conditions, actions, and results, and in conjunction with a preset rule template, the experience information is filled into the structured rule fragment;

[0191] When the experience information cannot be directly mapped to the rule template, an interactive clarification interface is provided to the senior personnel, which displays a rule fragment that the system has initially understood.

[0192] Receive corrections or additions to the rule fragments from the senior personnel.

[0193] Specifically, receiving the aforementioned experience information refers to the system acquiring descriptive content from the interactive interface of senior personnel, such as informal rules, off-process risk mitigation mechanisms, or intentions for balancing multiple objectives regarding adjustments to non-standard paths. This experience information is typically in natural language and may contain colloquial expressions, technical jargon, or tacit knowledge.

[0194] The process of segmenting and identifying words from the empirical information to obtain language units can be understood as using natural language processing (NLP) technology to decompose continuous text information into independent sentences and words. For example, a word segmenter can be used to split sentences into word sequences and tag words with parts of speech, laying the foundation for subsequent semantic analysis.

[0195] Furthermore, based on a pre-defined business domain vocabulary and synonym set, the language units are standardized to obtain standardized language units. The purpose of this is to eliminate ambiguity and inconsistency in natural language expressions. For example, synonyms such as "purchase order," "PO," and "order" are uniformly mapped to the standard business entity "purchase order," ensuring the standardization and consistency of rule expressions.

[0196] In practical applications, identifying the business entities contained in the experience information and determining the type and attributes of the business entities according to the business ontology definition means that the system uses a pre-built business ontology to identify key business objects involved in the experience information, such as "suppliers", "contract amounts", and "approvers", and clarifies the specific types of these entities (such as "personnel" and "financial data") and their related attributes (such as "supplier level" and "contract status").

[0197] Analyzing the syntactic structure between language units in the experiential information to identify conditions, actions, and results can be understood as using syntactic analysis techniques to resolve the relationships between sentence components, thereby extracting the core elements that constitute the rules. For example, identifying the logical structure "If (condition)...then (action)...to achieve (result)...".

[0198] Therefore, based on the conditions, actions, and results, and in conjunction with a preset rule template, the experience information is filled into the structured rule fragment. The purpose is to transform unstructured experience into a unified, machine-readable rule format. For example, the preset rule template could be "IF [condition] THEN [action] ELSE [result]", and the system will fill the identified conditions, actions, and results into the corresponding slots.

[0199] As a preferred implementation, when the experience information cannot be directly mapped to the rule template, an interactive clarification interface is provided to the senior personnel. This interface displays a rule fragment that the system initially understands, aiming to introduce human intervention to improve the accuracy of rule conversion. The interface can highlight parts that the system does not fully understand or provide multiple possible interpretations for the senior personnel to choose from.

[0200] Finally, the system receives corrections or additions to the rule fragments from the senior personnel, ensuring that the final generated structured rule fragments accurately reflect the true intentions and experience of the senior personnel, thus compensating for any shortcomings that may exist in automated conversion.

[0201] This application's solution effectively addresses the problems of low efficiency, poor accuracy, and difficulty in ensuring rule consistency in traditional methods by introducing a series of structured natural language processing and semantic understanding steps. Through the aforementioned technical solution, this application can significantly improve the efficiency and accuracy of converting experiential information into structured rules. Compared to traditional manual coding or simple keyword matching, this solution, through refined natural language processing and semantic understanding, ensures that the rules extracted from the experience of senior personnel are more accurate and consistent, and effectively avoids ambiguity. Furthermore, the introduced interactive clarification mechanism allows the system to obtain timely feedback and corrections from senior personnel when its automated processing capabilities are insufficient, greatly improving the success rate and quality of rule conversion. As a result, the generated structured experiential rule fragments can be more reliably integrated into the process recommendation logic, providing high-quality suggested paths for subsequent intelligent office automation process management, thereby improving the overall intelligence level and operational efficiency of the process.

[0202] In some embodiments described above, this application proposes recording the adoption status of suggested paths and evaluating their efficiency improvement effects. Based on the adoption status and efficiency improvement effects, it determines whether a structured empirical rule fragment has passed validation. However, in practical applications, relying solely on adoption rates and efficiency improvement effects to judge the effectiveness of empirical rule fragments may have limitations. For example, if a suggested path is mainly adopted by a few senior personnel, and their adoption behavior may be influenced by personal preferences or specific situations rather than its general applicability or intrinsic value, simply considering its adoption as validation might lead to the integration of empirical rules without universality into the process recommendation logic, thereby affecting the overall efficiency and stability of the process. Failure to address these issues may result in a decrease in the accuracy and reliability of the process recommendation strategy, and even introduce new operational risks.

[0203] To this end, this application further proposes to record the adoption status of suggested paths, evaluate the efficiency improvement effect of suggested paths, and determine whether the structured empirical rule fragments pass validation based on the adoption status and efficiency improvement effect. The specific steps include:

[0204] Identify the user roles and department information that adopt the suggested path;

[0205] Determine whether the user who adopted the recommendation is a senior or highly influential person;

[0206] If so, monitor the adoption rate trend of the suggested path among non-senior or non-high-influence personnel;

[0207] When the senior staff adopts the adoption and the adoption rate of the non-senior staff increases significantly, the adoption behavior is marked as "affected adoption".

[0208] Provide a feedback interface to users who adopt the suggested path, asking them for the basis of their decision to adopt it;

[0209] Receive the decision-making basis provided by the user;

[0210] The adoption behavior is classified according to the type and content of the decision-making basis;

[0211] When the adoption behavior is marked as not "affected adoption" and the decision basis points to actual business value, it is counted as a valid adoption. Based on the valid adoption and the efficiency improvement effect, it is determined whether the structured experience rule fragment passes the verification.

[0212] Specifically, identifying the user roles and department information for adopting suggestions means that the system obtains metadata such as the adopting user's position, permission level, and department within the organization through the user management module. Determining whether the adopting user is a senior or highly influential person can be based on preset rules. For example, if a user's position is department manager or above, or their historical operational data indicates that they have high decision-making weight and influence in a specific business area, they can be identified as a senior or highly influential person.

[0213] Furthermore, when adopting users are identified as senior or high-influence individuals, the system will initiate monitoring of adoption rates among non-senior or non-high-influence individuals. This aims to eliminate the "herd effect" caused by adoption by a few authoritative figures, thereby more objectively assessing the actual universality of the suggested path. If, after a senior user adopts the suggestion, the adoption rate among non-senior users increases significantly, some adoption behavior will be marked as "affected adoption," indicating that these adoptions may not be based on independent judgment but rather influenced by the behavior of senior users.

[0214] Furthermore, to gain a deeper understanding of the underlying motivations for adoption, the system provides a feedback interface to users who adopt suggested paths. This feedback interface aims to inquire about the user's decision-making basis for adopting the suggested path, such as whether it's to improve efficiency, mitigate risks, simplify operations, or other specific considerations. After receiving the user's decision-making basis, the system categorizes the adoption behavior based on its type and content, such as "efficiency-driven adoption," "risk-averse adoption," or "convenience-driven adoption."

[0215] This application's solution, by comprehensively considering the adopting user's role, their influence on other users, and the decision-making basis behind the adoption behavior, can more comprehensively and deeply evaluate the practical value and universality of experience rule fragments. Through this technical solution, validation bias caused by the adoption by a few senior personnel can be effectively avoided, ensuring that validated experience rule fragments have stronger universality and practical business value. Therefore, the accuracy and reliability of process recommendation strategies can be significantly improved, and potential risks caused by improper rule integration can be reduced, thus providing more robust and efficient decision support for intelligent office automation process management.

[0216] This application further proposes the following steps for continuously evaluating the performance of the recommended strategy for the process and adjusting the recommended strategy based on the evaluation results:

[0217] For each integrated process recommendation strategy, collect objective performance data on the adoption rate, process duration, and error rate of the process recommendation strategy during execution.

[0218] After the senior staff member completes the process operation of the recommended process strategy, a feedback interface is provided. The feedback interface generates evaluation dimensions related to the business context of the senior staff member's operation. The evaluation dimensions include efficiency improvement, risk control and ease of operation, and provides quantitative scoring options and text input boxes.

[0219] The senior personnel will evaluate the performance of the process recommendation strategy on the evaluation dimension through the quantitative scoring options and the text input box.

[0220] Keyword extraction and sentiment analysis are performed on the text feedback input by the senior personnel, and the text feedback is transformed into structured feedback data.

[0221] The structured feedback data is fused with the objective performance data to calculate the overall performance score of the process recommendation strategy;

[0222] When the overall performance score is lower than a preset threshold, the key dimensions that cause the overall performance score to drop are identified.

[0223] Based on the key dimensions, adjustment suggestions for the process recommendation strategy are generated, including adjusting the triggering conditions, recommendation priority, or related parameters of the process recommendation strategy.

[0224] Receive confirmation or correction from the senior personnel regarding the proposed adjustments.

[0225] Specifically, when evaluating each integrated process recommendation strategy, the first step is to collect objective performance data during its actual execution. This objective performance data can be understood as quantifiable metrics automatically recorded by the system. For example, adoption rate refers to the frequency with which the recommendation strategy is accepted and executed by users; process duration refers to the time required to complete the entire process after adopting the strategy; and error rate reflects the proportion of operational errors or anomalies caused by adopting the strategy. This data provides a quantitative basis for the initial evaluation of the strategy.

[0226] Furthermore, to obtain more comprehensive evaluation information, the system proactively provides a feedback interface after senior personnel complete the process operation recommended by the strategy. This feedback interface is specifically designed for senior personnel to collect their qualitative evaluations of the strategy. The feedback interface dynamically generates a series of relevant evaluation dimensions based on the specific business context of the senior personnel's operation. These evaluation dimensions typically cover aspects that senior personnel are most concerned about in their actual work, such as efficiency improvement, risk control, and ease of operation. To facilitate senior personnel's expression, the feedback interface provides quantitative scoring options, allowing them to rate each dimension, and also provides text input boxes so that senior personnel can describe their feelings, problems encountered, or suggestions for improvement in detail.

[0227] After experienced personnel evaluate the performance of the process recommendation strategy across the aforementioned evaluation dimensions using quantitative rating options and text input boxes, the system receives this evaluation information. To effectively utilize the unstructured text feedback, the system performs keyword extraction and sentiment analysis on the text feedback entered by experienced personnel. Keyword extraction aims to identify the core concepts and concerns in the feedback, while sentiment analysis is used to determine whether the experienced personnel's overall attitude towards the strategy is positive, negative, or neutral. Through these processes, the text feedback is transformed into structured feedback data, facilitating subsequent quantitative analysis and integration.

[0228] Subsequently, the structured feedback data is fused with the previously collected objective performance data to calculate a comprehensive performance score for the process recommendation strategy. This fusion aims to combine quantitative indicators with qualitative evaluations to form a more comprehensive and accurate assessment of strategy performance. For example, different weights can be assigned to the objective performance data and the structured feedback data, or a multi-dimensional weighted average method can be used for calculation.

[0229] When the calculated overall performance score falls below a preset threshold, it indicates a potential problem with the process recommendation strategy, requiring adjustment. At this point, the system further identifies the key dimensions causing the decline in the overall performance score. For example, if the efficiency improvement dimension scores high but the risk control dimension scores low, then risk control is the key dimension. Based on the identified key dimensions, the system automatically generates adjustment suggestions for the process recommendation strategy. These suggestions can be multifaceted, such as adjusting the strategy's triggering conditions to ensure it is recommended in more appropriate business contexts; adjusting the recommendation priority to achieve a more reasonable ranking when competing with other strategies; or adjusting related parameters to optimize the strategy's specific execution logic. Finally, the system receives confirmation or corrections from senior personnel regarding these adjustment suggestions to ensure that the final adjustment plan meets actual business needs and the professional judgment of senior personnel.

[0230] This application's solution effectively overcomes the limitations of relying solely on objective performance data for evaluating process recommendation strategies by incorporating qualitative feedback from experienced personnel. Through this technical solution, the application achieves a more comprehensive, in-depth, and user-friendly evaluation and adjustment of process recommendation strategies. Compared to evaluation methods that rely solely on objective data, this solution, by introducing qualitative feedback from experienced personnel, effectively captures potential hidden problems and user experience deficiencies in the strategy's practical application, thus avoiding optimization biases caused by one-sided evaluations. The professional judgment and experience of experienced personnel are systematically integrated into the evaluation process, enabling the overall performance score to more accurately reflect the strategy's true value and potential risks. Consequently, the generated adjustment suggestions are more targeted and effective, effectively addressing specific issues related to efficiency, risk, and convenience, significantly improving the adaptability, robustness, and user satisfaction of the process recommendation strategy, ultimately promoting the continuous optimization and efficient operation of intelligent office automation process management.

[0231] In some existing implementations, when merging structured feedback data provided by senior personnel with objective performance data collected by the system to calculate the overall performance score of the process recommendation strategy, inconsistencies or time misalignments may arise. For example, feedback from senior personnel may be based on their subjective feelings, while objective performance data reflects the actual system operation; there may be a time lag or misunderstanding between the two. If these data inconsistencies are not addressed, the calculated overall performance score may not accurately reflect the true performance of the process recommendation strategy, thus affecting the effectiveness of subsequent adjustment suggestions.

[0232] In response, this application further proposes a method for integrating structured feedback data with objective performance data to calculate the comprehensive performance score of the process recommendation strategy, specifically including:

[0233] The data provided by senior staff is timestamped to obtain calibrated feedback data.

[0234] Perform consistency cross-validation on the calibrated feedback data and objective performance data to identify any inconsistencies between the feedback data and objective performance data that fall outside a preset threshold range;

[0235] When an inconsistency is detected between the feedback data and the objective performance data that is outside the preset threshold range, the abnormal data marking mechanism is activated to mark the inconsistent data pairs as data to be reviewed.

[0236] Based on the severity of data inconsistency or temporal misalignment, the weights of data pairs in the fusion computation are dynamically adjusted, and the overall performance score of the process recommendation strategy is calculated.

[0237] Generate a data inconsistency report.

[0238] Specifically, timestamping the data from senior staff feedback involves aligning the evaluation data submitted by senior staff in the feedback interface with the actual time of the process operation being evaluated. This can be achieved by recording the difference between the feedback submission time and the corresponding operation completion time, or by associating the feedback data with the timestamp of the most recent related operation. The goal is to ensure that subjective feedback and objective performance data are comparable in the time dimension.

[0239] The consistency cross-validation of calibrated feedback data and objective performance data can be understood as comparing the differences between the two types of data on key indicators using pre-defined logical rules or statistical models. For example, if senior personnel report a significant improvement in efficiency, but objective performance data shows an increase in process time, there may be an inconsistency. This validation aims to identify these potential contradictions and to discover conflicts or biases between data sources.

[0240] In practical applications, when an inconsistency is found between the feedback data and the objective performance data that is outside the preset threshold range, the abnormal data marking mechanism is activated. This involves specially marking these data pairs with significant differences, for example, by setting their status to "awaiting manual review" or "abnormal data". The purpose is to remind the system or administrator to further review and process these data to prevent them from directly affecting the accuracy of the overall performance score.

[0241] Furthermore, dynamically adjusting the weights of data pairs in the fusion calculation based on the severity of data inconsistency or temporal misalignment means that when calculating the overall performance score, data pairs marked as inconsistent or severely misaligned can have their weight reduced in the total score calculation, or their influence can be temporarily excluded in extreme cases. For example, different weight decay factors can be set based on the degree of inconsistency (such as a 20% increase in objective time but a subjective evaluation of "significantly improved efficiency"). The purpose is to reduce the negative impact of abnormal data on the final evaluation results and improve the robustness of the evaluation.

[0242] In addition, generating a data inconsistency report means that the system automatically summarizes all data pairs marked as inconsistent and records in detail the type and extent of the inconsistency, the recommended processes and strategies involved, and the senior personnel involved. This report can be generated and sent to relevant managers or data analysts on a regular basis, aiming to provide a basis for data quality monitoring and assist in manual intervention and strategy optimization.

[0243] This application's solution effectively addresses the accuracy issues that may arise when integrating feedback data from senior personnel with objective performance data by introducing timestamp calibration, consistency cross-validation, outlier labeling, and dynamic weight adjustment mechanisms. Through these technical solutions, this application significantly improves the accuracy and reliability of calculating the overall performance score of the process recommendation strategy. Timestamp calibration of the feedback data ensures synchronization between subjective evaluations and objective facts in the time dimension, avoiding misjudgments caused by time sequence discrepancies. The introduction of the consistency cross-validation mechanism enables the system to proactively identify potential contradictions between senior personnel feedback and objective performance data, effectively avoiding evaluation biases that may occur in traditional methods. Furthermore, the outlier labeling mechanism and dynamic weight adjustment strategy allow the system to intelligently reduce the impact of data inconsistencies on the final evaluation results, improving the robustness of the evaluation. Therefore, the generated overall performance score more realistically and comprehensively reflects the actual effectiveness of the process recommendation strategy, providing a more accurate decision-making basis for subsequent strategy adjustments, thereby optimizing the overall efficiency and effectiveness of intelligent office automation process management.

[0244] refer to Figure 2 , Figure 2 This is a schematic diagram of the structure of an AI-based intelligent office automation process management system provided in an embodiment of the present invention, including:

[0245] The identification end is used to acquire the operation behavior data of the approver and store the preset process path information; compare the operation behavior data with the preset process path information, identify the operation sequence that differs from the preset process path information as a non-standard path adjustment, and count the frequency of the occurrence of the non-standard path adjustment.

[0246] The interactive interface is used to provide an interactive interface to senior personnel who perform the non-standard path adjustment when the frequency of the non-standard path adjustment reaches a preset threshold, so as to obtain the experience information corresponding to the non-standard path adjustment; to convert the experience information into structured experience rule fragments; to store the structured experience rule fragments, and to display the structured experience rule fragments as suggested paths in the subsequent process recommendation process.

[0247] The judgment end is used to record the adoption status of the suggested path, evaluate the efficiency improvement effect of the suggested path, and determine whether the structured experience rule fragment passes the verification based on the adoption status and the efficiency improvement effect.

[0248] The adjustment end is used to integrate the validated structured empirical rule fragments into the process recommendation logic to generate a process recommendation strategy; continuously evaluate the performance of the process recommendation strategy, and adjust the process recommendation strategy based on the evaluation results.

[0249] The AI-based intelligent office automation process management system disclosed in this application aims to address the limitations of traditional intelligent office automation systems in handling complex and ever-changing business processes and tacit experience knowledge. This system continuously monitors the operational behavior data of approvers through an identification terminal, proactively identifying non-standard path adjustments that differ from preset standard process paths and statistically analyzing their frequency. When a non-standard path adjustment reaches a preset threshold, the interaction terminal proactively provides an interface to experienced personnel to obtain the underlying experience information, thereby making tacit knowledge explicit. This experience information is then transformed into structured experience rule fragments by the interaction terminal and displayed as suggested paths. The judgment terminal records the adoption status of suggested paths and evaluates their efficiency improvement effects to verify the effectiveness of the experience rule fragments. Finally, the verified experience rule fragments are integrated into the process recommendation logic by the adjustment terminal to form a process recommendation strategy. Through continuous evaluation and adjustment, the system ensures continuous self-optimization, adapting to the evolution of business processes and significantly improving the accuracy and timeliness of path recommendations.

[0250] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. An AI-based intelligent office automation process management method, characterized in that, include: Acquire the operational behavior data of the approvers and store the preset process path information; By comparing the operation behavior data with the preset process path information, operation sequences that differ from the preset process path information are identified as non-standard path adjustments, and the frequency of occurrence of the non-standard path adjustments is statistically analyzed. When the frequency of the non-standard path adjustment reaches a preset threshold, an interactive interface is provided to senior personnel who perform the non-standard path adjustment to obtain experience information corresponding to the non-standard path adjustment. The experiential information is transformed into structured fragments of experiential rules; The structured experience rule fragments are stored, and in subsequent process recommendation, the structured experience rule fragments are displayed as suggested paths; Record the adoption status of the suggested path and evaluate the efficiency improvement effect of the suggested path. Based on the adoption status and the efficiency improvement effect, determine whether the structured empirical rule fragment passes the verification. The validated structured empirical rule fragments are integrated into the process recommendation logic to generate a process recommendation strategy; The performance of the process recommendation strategy is continuously evaluated, and the strategy is adjusted based on the evaluation results.

2. The AI-based intelligent office automation process management method according to claim 1, characterized in that, When the frequency of non-standard path adjustments reaches a preset threshold, an interactive interface is provided to experienced personnel performing the non-standard path adjustments to obtain experience information corresponding to the non-standard path adjustments, including: When the senior personnel are about to perform a key operation, monitor the key operation and, in conjunction with the business type of the form, the field values ​​already filled in, and the senior personnel's historical operation patterns, determine whether the key operation will cause a deviation from the preset process path information or involve potential compliance risks. When it is determined that the key operation deviates from the preset process path information or involves potential compliance risks, a first prompt box is provided to the senior personnel. The first prompt box asks the senior personnel about the informal rules or off-process risk mitigation mechanisms on which the key operation is based, based on the business context of the form and the predicted result of the key operation. In the first prompt box, the senior personnel are guided to understand the multiple objectives behind the key operation, including improving approval efficiency, avoiding supply chain disruption risks, or ensuring compliance with quarterly cumulative procurement requirements. After the senior personnel complete the declaration of the informal rules, the out-of-process risk mitigation mechanism, and the intention to balance multiple objectives, a risk and compliance assessment is conducted based on the declaration, the business type of the form, the filled-in field values, and the known formal and informal rules. When the risk and compliance assessment results indicate potential compliance risks, a second prompt box is provided to the senior personnel. The second prompt box points out the potential risk points and provides compliance suggestions, which the senior personnel can choose to accept and adjust their operations accordingly.

3. The AI-based intelligent office automation process management method according to claim 1, characterized in that, The process involves recording the adoption status of the suggested paths, evaluating the efficiency improvement effect of the suggested paths, and determining whether the structured empirical rule fragment passes validation based on the adoption status and the efficiency improvement effect, including: Identify the triggering conditions contained in the structured experience rule fragments, and extract the business context features of the structured experience rule fragments based on the triggering conditions; Based on the business context characteristics, form instances that match the triggering conditions are selected from historical operation behavior data; Analyze whether any anomalies occurred during the actual workflow of the form instance due to the adoption of non-standard paths; Calculate the situational risk index of the structured empirical rule fragment based on the frequency and severity of the abnormal situation; Based on the execution actions specified in the structured empirical rule fragments, the execution of the structured empirical rule fragments is simulated in test environments of different business departments; Assess the impact of the simulation execution on different segmented business scenarios, and generate a scenario universality assessment report of the structured empirical rule fragments; When the situational risk index is lower than a preset threshold and the scenario universality assessment report shows that the structured empirical rule fragment performs well in multiple relevant sub-scenarios, the structured empirical rule fragment is judged to have passed the verification.

4. The AI-based intelligent office automation process management method according to claim 1, characterized in that, The continuous evaluation of the performance of the process recommendation strategy and the adjustment of the process recommendation strategy based on the evaluation results include: For each integrated process recommendation strategy, identify the associated business departments and the downstream process nodes that may be affected; Collect process data from business departments associated with the process recommendation strategy. The process data includes approval time, task backlog, abnormal rejection rate, or compliance check pass rate of downstream departments. Maintain a hidden risk rule base, which contains potential compliance risk patterns caused by local efficiency improvements; When the process recommendation strategy improves the efficiency of its main related departments, but causes the approval time of the downstream departments to increase, the abnormal rejection rate to rise, or triggers the risk patterns in the implicit risk rule base, the comprehensive impact index of the process recommendation strategy on the overall process efficiency and compliance is calculated. Based on the comprehensive impact index, adjust the priority of the process recommendation strategy, or send an early warning notification to relevant department managers.

5. The AI-based intelligent office automation process management method according to claim 1, characterized in that, When the frequency of non-standard path adjustments reaches a preset threshold, an interactive interface is provided to experienced personnel performing the non-standard path adjustments to obtain experience information corresponding to the non-standard path adjustments, including: The interactive interface is provided after the senior personnel make the adjustment to the non-standard path; Based on the business type of the form and the historical operation context of senior personnel, preset experience templates and keyword tags are dynamically generated and displayed in the interactive interface; The experience information is received from the senior personnel through the interactive interface, including the experience template, the keyword tags, or the descriptive information they select. Provides a context-related backtracking function, which displays the senior personnel's operation sequence and related data fragments before this non-standard adjustment; Receive annotations or supplementary notes from the senior personnel on the operation sequence and the data segment as part of the experience information; A conflict pre-detection is performed on the experience information input by the senior personnel, and the conflict pre-detection compares the experience information with the existing rule base; When the conflict pre-detection detects potential conflicts or inconsistencies, the senior personnel are alerted to the potential conflicts or inconsistencies. Receive corrections or supplements from the senior personnel regarding the experience information.

6. The AI-based intelligent office automation process management method according to claim 1, characterized in that, The process of transforming the experiential information into structured fragments of experiential rules includes: Receive the experience information; The empirical information is segmented into sentences and words are identified to obtain language units; Based on a pre-defined business domain vocabulary and a set of synonyms, the language units are standardized to obtain standardized language units. Identify the business entities contained in the experience information, and determine the type and attributes of the business entities according to the business ontology definition; Analyze the syntactic structure between language units in the empirical information to identify conditions, actions, and results; Based on the conditions, actions, and results, and in conjunction with a preset rule template, the experience information is filled into the structured experience rule fragment; When the experience information cannot be directly mapped to the preset rule template, an interactive clarification interface is provided to the senior personnel, which displays a rule fragment that the system has initially understood. Receive corrections or additions to the rule fragments from the senior personnel.

7. The AI-based intelligent office automation process management method according to claim 1, characterized in that, The process involves recording the adoption status of the suggested paths, evaluating the efficiency improvement effect of the suggested paths, and determining whether the structured empirical rule fragment passes validation based on the adoption status and the efficiency improvement effect, including: Identify the user roles and department information that adopt the suggested path; Determine whether the user who adopted the recommendation is a senior or highly influential person; If so, monitor the adoption rate trend of the suggested path among non-senior or non-high-influence personnel; When the senior staff adopts the adoption and the adoption rate of the non-senior staff increases significantly, the adoption behavior is marked as "affected adoption". Provide a feedback interface to users who adopt the suggested path, asking them for the basis of their decision to adopt it; Receive the decision-making basis provided by the user; The adoption behavior is classified according to the type and content of the decision-making basis; When the adoption behavior is marked as not "affected adoption" and the decision basis points to actual business value, it is counted as a valid adoption. Based on the valid adoption and the efficiency improvement effect, it is determined whether the structured experience rule fragment passes the verification.

8. The AI-based intelligent office automation process management method according to claim 1, characterized in that, The continuous evaluation of the performance of the process recommendation strategy and the adjustment of the process recommendation strategy based on the evaluation results include: For each integrated process recommendation strategy, collect objective performance data on the adoption rate, process duration, and error rate of the process recommendation strategy during execution. After the senior personnel complete the process operation of the recommended process strategy, a feedback interface is provided. The feedback interface generates evaluation dimensions related to the business context of the senior staff member's operation. These evaluation dimensions include efficiency improvement, risk control, and ease of operation, and provide quantitative scoring options and text input boxes. The senior personnel will evaluate the performance of the process recommendation strategy on the evaluation dimension through the quantitative scoring options and the text input box. Keyword extraction and sentiment analysis are performed on the text feedback input by the senior personnel, and the text feedback is transformed into structured feedback data. The structured feedback data is fused with the objective performance data to calculate the overall performance score of the process recommendation strategy; When the overall performance score is lower than a preset threshold, the key dimensions that cause the overall performance score to drop are identified. Based on the key dimensions, adjustment suggestions for the process recommendation strategy are generated, including adjusting the triggering conditions, recommendation priority, or related parameters of the process recommendation strategy. Receive confirmation or correction from the senior personnel regarding the proposed adjustments.

9. The AI-based intelligent office automation process management method according to claim 8, characterized in that, The step of fusing the structured feedback data with the objective performance data to calculate the comprehensive performance score of the process recommendation strategy includes: The data provided by the senior personnel is timestamped to obtain calibrated feedback data. The calibrated feedback data and the objective performance data are cross-validated for consistency to identify whether there is any inconsistency between the feedback data and the objective performance data that is outside a preset threshold range; When an inconsistency is detected between the feedback data and the objective performance data that is outside a preset threshold range, an abnormal data marking mechanism is activated to mark the inconsistent data pairs as data to be reviewed. Based on the severity of data inconsistency or temporal misalignment, the weights of the data pairs in the fusion calculation are dynamically adjusted, and the overall performance score of the process recommendation strategy is calculated. Generate a data inconsistency report.

10. An AI-based intelligent office automation process management system, characterized in that, include: The identification end is used to acquire the operational behavior data of the approvers and store the preset process path information; By comparing the operation behavior data with the preset process path information, operation sequences that differ from the preset process path information are identified as non-standard path adjustments, and the frequency of occurrence of the non-standard path adjustments is statistically analyzed. The interactive interface is used to provide an interactive interface to senior personnel who perform the non-standard path adjustment when the frequency of the non-standard path adjustment reaches a preset threshold, so as to obtain the experience information corresponding to the non-standard path adjustment; to convert the experience information into structured experience rule fragments; to store the structured experience rule fragments, and to display the structured experience rule fragments as suggested paths in the subsequent process recommendation process. The judgment end is used to record the adoption status of the suggested path, evaluate the efficiency improvement effect of the suggested path, and determine whether the structured experience rule fragment passes the verification based on the adoption status and the efficiency improvement effect. The adjustment end is used to integrate the validated structured empirical rule fragments into the process recommendation logic to generate a process recommendation strategy; The performance of the process recommendation strategy is continuously evaluated, and the strategy is adjusted based on the evaluation results.

Images