Customer touchpoint intelligent identification and intelligent follow-up method, device, equipment and medium

Abstract

The application relates to the field of artificial intelligence and financial technology, and discloses a customer touch point intelligent identification and intelligent follow-up method, device, equipment and medium, which comprises the following steps: acquiring touch point data generated by a customer in multiple interaction channels; processing the touch point data based on business rules and a machine learning model, and identifying key touch points in the customer journey; generating follow-up priorities for the key touch points based on the types of the key touch points, the portrait features of the customer and a customer intention score generated through the machine learning model; matching corresponding intelligent follow-up strategies from a preset strategy library according to the types of the key touch points and the follow-up priorities; executing the intelligent follow-up strategies to reach corresponding customers through at least one communication interface; collecting feedback data of the customers on the reaching, and optimizing and iterating the machine learning model and the intelligent follow-up strategies based on the feedback data. The application realizes intelligent identification, sorting, follow-up and optimization of customer touch points, improves conversion rate and reduces loss of sales opportunities.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and financial technology, and in particular to a method, apparatus, device and medium for intelligent identification and intelligent follow-up of customer touchpoints. Background Technology

[0002] In the service technology fields of finance and insurance, customer interaction management systems are typically used to monitor and manage customer interactions and statuses at key stages such as product consultation, browsing, quoting, and payment. In existing technologies, these systems mainly rely on predefined rules to monitor customer actions at key points in standard service processes. For example, when the system detects that a customer has not confirmed a quote within a preset time after it has been generated, or has not completed payment after submitting an order, it marks the customer as a pending "breakpoint" or task, typically providing it to customer service personnel in a list for manual follow-up.

[0003] The inventors discovered that this existing solution has significant limitations. First, its identification and screening mechanism is relatively simple and passive, heavily relying on explicit interruptions that occur within a clearly structured process and can only be detected afterward. It lacks effective automated identification and value assessment capabilities for a large number of unstructured customer potential intention behaviors that occur outside or early in the process (such as in-depth browsing of product pages, repeated clicks on marketing content, etc.), resulting in the loss of many early sales leads. Second, for the identified list of customers to be followed up, it can only perform simple listing or sorting by time, unable to automatically assess the urgency and potential value differences between different customers in the list. Therefore, it cannot provide intelligent decision support for prioritizing manual follow-ups, leading to inefficient allocation of customer service resources and the potential for high-value opportunities to be missed. These problems collectively contribute to the difficulty in improving customer follow-up efficiency and conversion rates. Summary of the Invention

[0004] This invention provides a method, apparatus, device, and medium for intelligent identification and follow-up of customer touchpoints, in order to solve the technical problems of low customer conversion efficiency and loss of high-value sales opportunities caused by the inability to effectively identify potential customer intentions and the inability to intelligently prioritize and allocate resources to a large number of customers to be followed up.

[0005] Firstly, a method for intelligent identification and follow-up of customer touchpoints is provided, including: Acquire customer touchpoint data generated across multiple interaction channels, including customer behavior data, attribute data, and historical interaction records; Based on predefined business rules and preset machine learning models, the touchpoint data is processed to identify key touchpoints in the customer journey. Based on the type of the key touchpoint, the customer profile features, and the customer intent score generated by the machine learning model, a follow-up priority is generated for the key touchpoint. Based on the type of the key touchpoint and the follow-up priority, a corresponding intelligent follow-up strategy is matched from a preset strategy library; The intelligent follow-up strategy is executed through at least one communication interface to reach the corresponding customer; Collect customer feedback data on outreach, and optimize and iterate the machine learning model and the intelligent follow-up strategy based on the feedback data.

[0006] Secondly, a customer touchpoint intelligent identification and intelligent follow-up device is provided, including: The touchpoint data acquisition module is used to acquire touchpoint data generated by customers in multiple interaction channels. The touchpoint data includes customer behavior data, attribute data, and historical interaction records. The key touchpoint module is used to process the touchpoint data based on predefined business rules and preset machine learning models to identify key touchpoints in the customer journey. The follow-up priority generation module is used to generate a follow-up priority for the key touchpoints based on the type of the key touchpoints, the customer profile features, and the customer intent score generated by the machine learning model. The intelligent strategy matching module is used to match the corresponding intelligent follow-up strategy from the preset strategy library according to the type of the key touchpoint and the follow-up priority. The strategy execution outreach module is used to execute the intelligent follow-up strategy to reach the corresponding customer through at least one communication interface; The feedback optimization and iteration module collects customer feedback data on the outreach and optimizes and iterates the machine learning model and the intelligent follow-up strategy based on the feedback data.

[0007] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described intelligent customer touchpoint identification and intelligent follow-up method.

[0008] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the aforementioned intelligent customer touchpoint identification and intelligent follow-up method.

[0009] The aforementioned solution based on intelligent customer touchpoint identification and follow-up methods, devices, equipment, and media acquires customer touchpoint data generated across multiple interaction channels, including behavior, attributes, and historical records. This data is processed based on predefined business rules and machine learning models to identify key touchpoints in the customer journey. Follow-up priorities are generated for key touchpoints based on their types, customer profiles, and customer intent scores generated by the model. Corresponding intelligent follow-up strategies are matched from a pre-defined strategy library according to the key touchpoint type and follow-up priority. The strategy is executed through at least one communication interface to reach the customer. Feedback data is collected to optimize the model and strategy. In this invention, addressing the technical problems of low customer conversion efficiency and lost high-value sales opportunities caused by the inability to effectively identify potential customer intentions and the inability to intelligently prioritize and allocate resources for a large number of customers to be followed up with, a solution integrating rule-based judgment and machine learning models for intelligent identification and dynamic ranking is proposed. This solution first integrates heterogeneous data from multiple channels and processes it using preset rules and machine learning models. This not only captures explicit interruptions in the preset process but also intelligently identifies potential intentional behaviors outside the process, thereby expanding the scope of effective customer touchpoints and solving the problems of single and missed lead sources. By generating customer intent scores based on customer profile features and machine learning models, follow-up priorities are generated for identified key touchpoints. The system can automatically rank a massive number of customers to be followed up with based on their value and intent, achieving intelligent guidance and precise allocation of customer service resources. This overcomes the blindness and inefficiency of manual ranking and significantly improves the conversion rate of high-value opportunities. Furthermore, by matching key touchpoint types and follow-up priorities with a preset strategy library and executing them through multiple communication channels, automation and a degree of personalization are achieved from strategy matching to customer outreach, reducing the delays and inconsistencies caused by manual intervention. Finally, by optimizing and iterating the machine learning model and follow-up strategy based on the feedback data, a closed-loop learning and optimization system is formed, which enables the identification accuracy, ranking rationality and reach effectiveness of the entire solution to be continuously improved, thereby systematically enhancing the intelligence level of the customer follow-up process and the sustainability of business results. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of an application environment for a customer touchpoint intelligent identification and intelligent follow-up method according to an embodiment of the present invention.

[0012] Figure 2 This is a flowchart illustrating a method for intelligent customer touchpoint identification and follow-up.

[0013] Figure 3 yes Figure 2 A flowchart illustrating a specific implementation of step S200.

[0014] Figure 4 yes Figure 2 A flowchart illustrating a specific implementation of step S300.

[0015] Figure 5 yes Figure 2 A schematic diagram of a specific implementation of step S400.

[0016] Figure 6 yes Figure 2 A schematic diagram of a specific implementation method for step S500.

[0017] Figure 7 yes Figure 2 A flowchart illustrating another specific implementation of step S500.

[0018] Figure 8 This is a schematic diagram of a customer touchpoint intelligent identification and intelligent follow-up device in one embodiment of the present invention; Figure 9 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention; Figure 10 This is a schematic diagram of a computer-readable storage medium according to an embodiment of the present invention. Detailed Implementation

[0019] 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, not all, of the embodiments of the present invention. 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.

[0020] The customer touchpoint intelligent identification and intelligent follow-up method provided in this invention can be applied to, for example... Figure 1The application environment shown includes a customer interaction terminal, multiple external interaction platforms, and an intelligent follow-up server. The intelligent follow-up server communicates with both the customer interaction terminal and the multiple external interaction platforms via a network. The customer interaction terminal refers to the terminal device used by the company's internal business personnel (such as insurance advisors, account managers, customer service agents, etc.), such as personal computers, laptops, smartphones, and tablets. Business personnel access the intelligent follow-up system through the customer interaction terminal, receive automatically generated customer follow-up task lists, priority prompts, and recommended strategies, and can manually initiate or confirm follow-up actions. The multiple external interaction platforms refer to various online and offline channels or interfaces through which the company interacts directly or indirectly with customers, including but not limited to: the company's official website and mobile applications, third-party social media platforms (such as WeChat mini-programs and WeChat Work), SMS gateways, email servers, telephone calling systems (including AI outbound calling platforms), and online customer service systems. These platforms are responsible for carrying out the original interactions with customers, generating and recording customer behavior data (such as browsing, clicking, contact information, communication, and payment), and transmitting the relevant data to the intelligent follow-up server through secure interfaces. The intelligent follow-up server is the backend service entity that deploys the intelligent customer touchpoint identification and intelligent follow-up system of this invention. Its core functions include: real-time or batch collection and fusion processing of heterogeneous customer touchpoint data from various external interaction platforms via data interfaces; processing the data using predefined business rules and built-in machine learning models to identify key touchpoints and generate customer intent scores; calculating follow-up priorities based on multi-dimensional information; matching and generating personalized follow-up strategies from a strategy library according to priority and touchpoint type; automatically executing strategies to reach customers by calling the API interfaces of the corresponding external interaction platforms; and collecting reach feedback data from various channels for continuous optimization of machine learning models and follow-up strategies. The intelligent follow-up server can be a standalone physical server, a distributed cluster of multiple servers, or a virtual service implemented based on a cloud computing platform. In this application environment, customer behavior on various external interaction platforms forms raw data, which is aggregated to the intelligent follow-up server; the server performs intelligent analysis, decision-making, and task generation, pushing tasks and strategies to customer interaction terminals for business personnel to refer to and execute, and automatically reaching customers through external interaction platforms; subsequent customer feedback is then transmitted back to the server, forming a closed loop. The present invention will now be described in detail through specific embodiments.

[0021] Please see Figure 2 As shown, Figure 2 This is a flowchart illustrating a method for intelligent customer touchpoint identification and follow-up, including the following steps: S100: Acquire customer touchpoint data across multiple interaction channels. Touchpoint data includes customer behavioral data, attribute data, and historical interaction records. Optionally, multiple interaction channels encompass various platforms where customers interact with the enterprise, such as the enterprise's official website, mobile applications, third-party social media (e.g., WeChat official accounts, mini-programs), online customer service systems, call centers, and SMS gateways. The system collects raw interaction logs from these channels in real-time or at scheduled intervals through pre-configured data interfaces. Behavioral data reflects specific customer actions, such as browsing paths on web pages or applications, page dwell time, button click records (e.g., "View Details," "Get Quote"), generated quote numbers and browsing status, initiated order actions, and payment attempt records. Attribute data refers to relatively static background information about customers, such as demographic information, family structure, and product lists obtained through registration or survey forms. Historical interaction records include summaries of past customer service communications, telephone call records, and participation and feedback from previous marketing activities. This step aims to achieve unified aggregation of fragmented, multi-source customer information to provide a complete data foundation for subsequent analysis.

[0022] S200: Based on predefined business rules and a pre-set machine learning model, the touchpoint data is processed to identify key touchpoints in the customer journey. Optionally, this step employs a dual mechanism combining rule-based judgment and model prediction to screen the aggregated touchpoint data. On one hand, the system incorporates pre-defined business rules by business experts to scan for interruptions in key business processes that conform to specific patterns, such as "not viewing or confirming the quotation within 24 hours after it is generated" in the quotation process. Events captured by these rules are marked as key touchpoints. On the other hand, the system pre-trains and deploys a machine learning model that can comprehensively analyze a series of customer behavior sequences and attributes, and output a quantitative score representing the strength of the customer's current purchase intention. The system also identifies behaviors with model scores significantly higher than the baseline threshold but that have not yet triggered any hard business rules (e.g., repeatedly comparing several product detail pages within a short period of time) as valuable key touchpoints. Through this dual mechanism, the system can not only capture explicit process interruptions but also discover behaviors of potential customers with implicit high intentions.

[0023] S300: Based on the type of key touchpoint, customer profile features, and customer intent score generated by a machine learning model, a follow-up priority is generated for each key touchpoint. Optionally, after identifying a key touchpoint, the system needs to determine its urgency and importance for follow-up. The system first determines the type of the key touchpoint (e.g., hesitation in the pricing process or interruption in the payment process). Simultaneously, the system retrieves the customer's profile features, which are comprehensive labels built based on their attribute data, historical behavior, etc., to characterize their customer group affiliation and potential value. Finally, the specific customer intent score generated by the machine learning model in step S200 for that touchpoint is combined. The system processes the touchpoint type, customer profile features, and customer intent score as input to calculate a follow-up priority for each key touchpoint. This follow-up priority directly determines the touchpoint's ranking among all pending tasks.

[0024] S400: Based on the type and follow-up priority of the key touchpoint, the system matches the corresponding intelligent follow-up strategy from a pre-defined strategy library. Optionally, the system maintains a pre-defined strategy library, which stores several standardized follow-up strategy templates for different types of key touchpoints. For example, for a touchpoint of "hesitation over pricing," the strategy library may have a pre-defined strategy template of "sending differentiated pricing comparison information"; for the "payment interruption" type, a pre-defined strategy template of "pushing a limited-time payment discount reminder" is provided. After the system generates a priority for a key touchpoint, it retrieves one or more corresponding basic strategy templates from the strategy library using its touchpoint type as an index. Subsequently, the system combines the specific context of the touchpoint (its customer profile, the specific scenario, etc.) and its priority level to select the most suitable template from the retrieved templates as the intelligent follow-up strategy for this follow-up. Higher priority touchpoints may be matched with more proactive and immediate strategies.

[0025] S500: Executes intelligent follow-up strategies to reach corresponding customers through at least one communication interface. Optionally, the system automatically calls the corresponding communication interface to execute the outreach action based on the content and requirements of the matched intelligent follow-up strategy. For example, if the strategy requires sending an SMS, the SMS gateway API is called; if the strategy requires assigning a human agent, a task order is pushed to the customer service workbench; if the strategy requires AI outbound calling, an outbound calling robot is dispatched to generate a call task. After the outreach action is executed, the system continuously monitors and collects customer feedback data, such as whether the customer opened the pushed message, clicked on the link, made a callback, or ultimately completed a payment or confirmation. This feedback data is recorded in real time and fed back to the system.

[0026] S600: Collect customer feedback data on outreach and optimize and iterate the machine learning model and intelligent follow-up strategy based on the feedback data. Optionally, based on continuously accumulated feedback data, the system will periodically or when a certain amount of data is reached, initiate an optimization and iteration process. On the one hand, new data containing feedback results will be used as training samples to retrain the machine learning model in step S200, making its intent scoring prediction more accurate. On the other hand, the outreach effect of different strategies in different scenarios (such as open rate and conversion rate) will be analyzed, thereby dynamically adjusting and optimizing the strategy selection logic or strategy content itself in the strategy library.

[0027] This invention provides a complete intelligent customer touchpoint identification and follow-up solution through steps S100 to S600. By integrating multi-channel data and fusing rules and models for identification, it significantly expands and deepens the ability to capture high-value sales opportunities and reduces lead loss. By introducing an automated priority ranking mechanism based on multi-dimensional information (type, profile, intent score), it achieves intelligent differentiation of massive follow-up tasks, enabling limited customer service resources to focus on opportunities with the highest conversion probability, greatly improving resource allocation efficiency and overall conversion rate. Through automatic matching of the strategy library and automatic multi-channel outreach, it achieves automation and a certain degree of personalization in the follow-up process, improving follow-up response speed and consistency. Finally, by establishing a closed-loop optimization mechanism based on feedback data, the system's identification accuracy, ranking rationality, and strategy effectiveness can continuously improve as the system operates, thereby systematically and sustainably enhancing the intelligence level and business output of the customer follow-up process, effectively solving the technical problems of low customer conversion efficiency and loss of high-value opportunities in existing technologies.

[0028] In some embodiments, please refer to Figure 3 , Figure 3 yes Figure 2 A flowchart illustrating a specific implementation of step S200: Step S200: Based on predefined business rules and machine learning models, the touchpoint data is processed to identify key touchpoints in the customer journey, including the following steps: S210: Identify interruption events occurring in key stages of a pre-defined business process based on business rules, and designate these interruption events as Category I critical touchpoints. Optionally, a pre-defined business process refers to a standardized sequence of steps that an enterprise, based on its service or sales logic, expects customers to complete sequentially. For example, in an insurance sales scenario, a typical business process might include key stages such as "browsing products, obtaining a quote, confirming intent, submitting an order, and completing payment." The system pre-configures business rules corresponding to these stages, which are logical judgment conditions used to detect process stagnation. For example, one rule might be defined as: "If a customer does not view or confirm a quote again within 24 hours of generating it, it is determined as a 'quotation interruption event'." Another rule might be defined as: "If a customer does not complete payment within 1 hour of placing an order, it is determined as a 'payment interruption event'." The system scans touchpoint data in real time or periodically. Once it finds that a customer's behavior meets the triggering conditions of a certain rule, it marks the event as a clear Category I critical touchpoint. These touchpoints directly reflect explicit customer churn along the established path.

[0029] S220: Input touchpoint data into a machine learning model to generate a customer intent score, and identify behavioral events with customer intent scores higher than a predetermined threshold as second-category key touchpoints. Optionally, to discover behaviors that do not violate hard business rules but still imply high purchase intent, a machine learning model is introduced for soft judgment. This model takes the touchpoint data (including the latest behavioral sequences, customer attributes, etc.) gathered in the preceding steps as input, performs non-linear calculations internally, and outputs a value between 0 and 1, called the "customer intent score." This score quantifies the model's predicted probability of "the customer completing a purchase in the near future." The system sets a predetermined decision threshold (e.g., 0.7). When the model evaluates a customer's behavior within a specific time window and the generated intent score exceeds this threshold, even if the behavior does not trigger any business rules (e.g., the customer did not place an order but browsed a product's terms page extensively for three consecutive days), the system will identify the key behavioral event that triggered this high score (such as "continuous extensive browsing") as a second-category key touchpoint. These touchpoints are used to capture customers' potential interests and early intent signals.

[0030] In some embodiments, the machine learning model is an ensemble learning model trained based on multi-dimensional features, including time features, behavior frequency features, customer attribute features, and product association features. Optionally, the machine learning model used in this embodiment is an ensemble learning model, such as LightGBM (LGBMClassifier). This model uses a large number of historical customer samples and their final conversion results (whether a purchase was made) as labels during the training phase. Each sample is represented by a carefully constructed multi-dimensional feature vector from customer touchpoint data, mainly including: Time features: such as "the time interval between the most recent key behavior and the current time (issuance interval)," "the total browsing time of all relevant pages in the past 30 days," and "the browsing time of a single session," used to capture the urgency and engagement of the behavior. Behavior frequency features: such as "the number of times the customer actively sought a quote or attempted to generate an order in the past 30 days," used to measure the frequency of the customer's proactive contact. Customer attribute features: such as "family size," "whether there are minor children," "whether there are elderly parents," and "the total number of product types purchased in the past," these static or semi-static features are used to characterize the customer's fundamentals and life stage, which are closely related to insurance needs. Product-related features, such as "the major category of the product currently being viewed or inquired about (e.g., health insurance, property insurance)" and "whether the customer has purchased similar or related products in the past," are used to associate specific product preferences. By training on these multi-dimensional features, the ensemble learning model can learn the complex correlation patterns between different feature combinations and purchase intentions. This allows for a more accurate comprehensive assessment of the intensity of intent when faced with new customer data, providing a reliable basis for distinguishing key touchpoints.

[0031] In this embodiment of the invention, through rule-based identification in S210, the system can reliably and stably capture all explicit drop-off points (first-type key touchpoints) occurring in the standard business funnel, ensuring that no basic sales opportunities are missed. Through intent scoring and threshold judgment based on a machine learning model in S220, the system possesses the ability to intelligently filter out high-intent potential leads (second-type key touchpoints) from massive amounts of seemingly normal or incomplete closed-loop behavioral data, greatly expanding the breadth and depth of sales opportunity mining and enabling early discovery of potential customers. By employing an ensemble learning model trained on multi-dimensional features, intent scoring has a solid mathematical foundation. The evaluation process comprehensively considers the time pattern, frequency, personal attributes, and product preferences of customer behavior, making the identification of second-type key touchpoints more accurate and interpretable, effectively reducing misjudgments and lead noise, and providing high-quality input for subsequent prioritization.

[0032] In some embodiments, please refer to Figure 4 , Figure 4 yes Figure 2A flowchart illustrating a specific implementation of step S300: Step S300: Based on the type of key touchpoints, customer profile features, and customer intent scores generated through a machine learning model, a follow-up priority is generated for the key touchpoints, including the following steps: S310: Determine customer value assessment results based on profile features and obtain the duration of key touchpoints. Optionally, customer profile features are an integrated, tagged representation of their attributes, historical behavior, and other information, containing signals for assessing their long-term and current potential value. In this step, the system extracts key indicators from the profile features to calculate the "customer value assessment result." For example, profile features such as "historical purchase of high-coverage or long-term insurance products," "long-term customer (long cooperation period)," and "good family asset status" can be mapped to a higher value score (such as a "high value" level or a specific score). Simultaneously, the system needs to calculate the "duration of key touchpoints," that is, the time elapsed from the moment the touchpoint is identified (e.g., an interruption or behavioral event) to the current system processing time. For example, a "payment interruption" touchpoint has occurred for 2 hours. Duration is an important indicator of opportunity urgency; generally, the shorter the duration, the greater the possibility of follow-up and recovery, and the higher the urgency.

[0033] S320: Taking customer intent score, customer value assessment result, and duration of key touchpoints as input, a quantified priority score is calculated using a preset priority ranking algorithm. Optionally, to compare and rank all pending key touchpoints on a uniform scale, the system sets up a preset priority ranking algorithm. This algorithm integrates three different dimensions of indicators—customer intent score reflecting "intent strength," customer value assessment result reflecting "customer importance," and touchpoint duration reflecting "opportunity urgency"—using a quantifiable method (e.g., weighted summation, rule-based piecewise function, or a more complex ranking model). Through this calculation, each key touchpoint is assigned a specific, comparable priority score. This score directly determines the touchpoint's position in the global list of tasks to be followed up, with higher scores ranked higher.

[0034] This invention, by introducing "customer value assessment" based on profile features and objective "touchpoint duration," supplements the "intent scoring" provided by machine learning models with two key business dimensions: customer lifetime value and opportunity window urgency. This makes priority determination more comprehensive and better aligned with the dual goals of business growth and operational efficiency. A pre-defined priority ranking algorithm integrates multi-dimensional heterogeneous indicators into a single, comparable quantitative score, transforming complex business decisions (who to follow first?) into an automated, standardized computational problem. This fundamentally solves the subjectivity, bias, and inefficiency of manual or single-dimensional ranking, ensuring that the system can automatically and continuously and accurately direct limited follow-up resources (such as customer service manpower and marketing resources) to the "most interested," "most valuable," and "most urgent" customer opportunities, thereby maximizing the return on resource investment and significantly improving overall conversion rates and customer satisfaction.

[0035] In some embodiments, please refer to Figure 5 , Figure 5 yes Figure 2 A flowchart illustrating a specific implementation of step S400: Step S400: Based on the type and follow-up priority of the key touchpoint, match the corresponding intelligent follow-up strategy from the preset strategy library, including the following steps: S410: Predefine at least one initial outreach strategy for different types of key touchpoints. Optionally, the system includes a pre-defined strategy library, primarily indexed by "key touchpoint type." Each system-defined type (e.g., "quote breakpoint," "payment breakpoint," "communication breakpoint," "lead generation breakpoint," or a more specific subcategory) is associated with one or more "initial outreach strategies." These initial strategies are standard operating procedure templates for best practices for that type of touchpoint, pre-defined by business experts based on experience. They specify the core objectives, primary channels, and content framework for follow-up. For example, for a "quote breakpoint," the core objective of its initial outreach strategy template might be "facilitating quote confirmation," the primary channel being "WeChat or SMS," and the content framework being "a short copy containing a comparison of differentiated quotes."

[0036] S420: Based on customer profile characteristics and historical interaction records, the initial outreach strategy content is personalized to generate a customized intelligent follow-up strategy. Optionally, directly using a standard template may appear rigid and reduce customer acceptance. Therefore, after selecting an initial strategy template, the system initiates a "personalization" process. The system queries the customer's profile characteristics (such as "has minor children" or "is a business owner") and their historical interaction records (such as "inquired about health insurance a week ago" or "prefers to receive WeChat messages"). Then, this information is used to dynamically fill in and fine-tune the variable parts of the template. For example, the salutation in the standard copy is changed from "Dear Customer" to "Mr. / Ms. X"; in the price comparison push, clauses related to "family health protection" or "business property risk" are highlighted; or the sending time is adjusted to the evening based on the customer's historical preferred communication time. After filling and adjustment, a general initial strategy is transformed into a customized intelligent follow-up strategy for a specific customer situation, which includes specific execution time, channels, and personalized copy or scripts.

[0037] The intelligent follow-up strategies include at least one of the following: differentiated pricing push, time-limited incentive information push, human agent task allocation, and automated outbound call task triggering. Optionally, several typical final forms of intelligent follow-up strategies are listed here: 1) Differentiated pricing push: The strategy ultimately generates a message, which may contain price adjustments based on customer profiles, comparisons of value-added services, or recommendations for protection plans tailored to specific risks, and is sent via app push or SMS. 2) Time-limited incentive information push: The strategy ultimately generates a reminder containing exclusive coupons, discount codes, or links to time-limited events, aiming to leverage a sense of urgency to encourage immediate action. 3) Human agent task allocation: The strategy ultimately generates a high-priority to-do task in the customer service workbench, along with customer profiles, intent scores, historical records, and recommended communication points, and assigns it to the most suitable customer service representative. 4) Automated outbound call task triggering: The strategy ultimately generates an outbound call task instruction, including the call timing, customer number, and AI-generated script based on historical conversations, and submits it to the AI ​​outbound call system for execution. The system selects one or a combination of strategies to execute based on touchpoint type, priority, and customer preferences.

[0038] This invention establishes a strategy library based on touchpoint type, accumulating best practice knowledge from business experts into standardized assets that can be automatically invoked by the system. This ensures that different touchpoints receive validated and targeted basic follow-up solutions, avoiding the arbitrariness and inconsistency of manual selection. Secondly, by introducing a personalized population mechanism based on customer profiles and history, standard strategies can dynamically adapt to the unique context and preferences of specific customers. This customization significantly improves the relevance and approachability of the content, thereby increasing open rates, read rates, and positive response rates, and enhancing the effectiveness of follow-up. Finally, by supporting multiple strategy formats (such as push notifications, manual calls, and outbound calls), the system can flexibly adapt to follow-up scenarios with different levels of urgency and customer preferences, forming a three-dimensional and flexible outreach system. This ensures that key information is delivered to customers in the most appropriate way, effectively promoting conversion.

[0039] In some embodiments, please refer to Figure 6 , Figure 6 yes Figure 2 A flowchart illustrating a specific implementation of step S500: Step S500: Executing an intelligent follow-up strategy to reach the corresponding customer through at least one communication interface, including: Based on the intelligent follow-up strategy and customer preferences, the system selects one or more methods from SMS, instant messaging, voice calls, or online application push notifications for outreach. Optionally, this step is responsible for translating the formulated intelligent follow-up strategy into specific customer-facing actions. The system integrates technical interfaces (APIs) for interfacing with various communication channels, such as SMS gateways, WeChat / WeChat Official Account interfaces, AI outbound calling / telephone call center interfaces, and push notification services from the enterprise's own app. During execution, the system does not mechanically use a single channel but performs intelligent channel selection and combination. First, the system parses the channel types suggested or supported by the intelligent follow-up strategy itself (e.g., "push a message"). Next, the system queries the customer's "customer preference" information, which may come from their profile characteristics (e.g., "more frequent communication with the enterprise via WeChat in historical interactions") or explicit settings in their historical interaction records (e.g., "select preference to receive SMS notifications in the personal center"). Based on the strategy requirements and customer preferences, the system determines the optimal one or more outreach methods. For example, for a high-priority payment interruption, the system might simultaneously choose to send a payment link via SMS and a pop-up notification via app push to ensure the information is delivered. For a lead requiring in-depth communication, it might prioritize voice calling, with the system automatically initiating an AI-powered outbound call task. This dynamic selection mechanism based on strategy and preferences ensures both the effectiveness of outreach and a user-friendly experience.

[0040] In some embodiments, please refer to Figure 7 , Figure 7 yes Figure 2 A flowchart illustrating a specific implementation of step S600: Step S600 involves collecting customer feedback data and optimizing and iterating the machine learning model and intelligent follow-up strategy based on the feedback data, including: S610: Record and analyze customer conversion behavior data generated after outreach. Optionally, after each outreach action (such as sending an SMS, completing an AI outbound call, or assigning a human agent task) is executed, the system will initiate a feedback monitoring cycle. During this period, the system records specific customer feedback by listening to the return status of relevant channels and monitoring the customer's subsequent behavior on the interaction platform. This feedback is structured as "customer conversion behavior data," such as: whether the message was "opened" or "clicked"; whether the coupon was "claimed" or "used"; whether the outbound call was "connected" and the "explicit intent" expressed by the customer during the call (such as "I'll think about it," "decline," or "agree to buy"); whether the task assigned to the human agent ultimately led the customer to complete key conversion events such as "successful payment" or "confirmation of insurance." The system will analyze this behavioral data, classify it as positive feedback (such as final conversion), negative feedback (such as explicit refusal), or no feedback, and associate it with previous outreach strategies, customer profiles, and original touchpoints.

[0041] S620: Customer conversion behavior data is added as new samples to the training dataset, and the machine learning model is retrained based on the updated training dataset. Optionally, to improve the prediction accuracy of the machine learning model, the system transforms the recorded and associated feedback data into new training samples. For example, for a customer who ultimately completes a purchase, all touchpoint data (features) prior to triggering follow-up behavior will be bound to the "positive sample" label (purchase = 1); for a customer who explicitly refuses or does not respond further, their corresponding data will be bound to the "negative sample" label (no purchase = 0). These new positive and negative samples are added to the model's original historical training dataset, forming a larger dataset that better reflects current market and customer behavior changes. The system will automatically trigger the model retraining process periodically (e.g., weekly or monthly) or when new samples accumulate to a certain scale. Retraining the model (e.g., an ensemble learning model) using the updated training dataset allows the model to learn the latest correlation patterns between features and purchase intentions, making it more accurate in generating customer intent scores in the future.

[0042] S630: Based on the conversion effect analysis results of various types of intelligent follow-up strategies, adjust the strategy matching rules or strategy content in the strategy library. Optionally, the system will aggregate and analyze the execution effect of various types of intelligent follow-up strategies in historical data. For example, the analysis may find that for customers with "payment breakpoints," the conversion rate (payment completion rate) of using the "limited-time incentive information push" strategy is significantly higher than that of simply using "payment reminder push"; or it may find that for customers with a certain profile, the open rate of "WeChat push" sent in the afternoon is much higher than that in the morning. Based on these conversion effect analysis results, the system can automatically or assist business experts in adjusting the strategy library. The adjustment may involve two aspects: first, adjusting the strategy matching rules, such as modifying the rules so that "quotation breakpoints" of high-value customers are preferentially matched with "human agent task allocation" rather than "automatic outbound calls"; second, optimizing the strategy content itself, such as modifying the copywriting in the "differentiated quotation push" template, or adjusting the discount and validity period in "limited-time incentives." Through this data-driven optimization, the strategy library can be continuously iterated and become more effective.

[0043] This invention, through systematic recording and analysis of conversion behavior data, transforms previously discrete and unquantified follow-up results into structured, machine learning-ready feedback information, providing an objective data foundation for system optimization. By using feedback data as new samples to continuously iterate and train the machine learning model, the model's intent recognition capabilities adapt to changes in customer behavior and the evolution of market trends, achieving self-evolution and continuous improvement of the system's core judgment capabilities, fundamentally ensuring the long-term accuracy of the source identification stage. Finally, by dynamically adjusting the strategy library based on performance data, follow-up strategies are no longer static, experience-based presets, but intelligent assets capable of selection, optimization, and fine-tuning based on actual results, thereby ensuring that the effectiveness of the outreach stage continuously increases over time. This closed-loop mechanism makes the entire system an intelligent agent capable of continuous learning and self-improvement from practice, significantly enhancing the long-term adaptability of the solution and the sustainability of business results.

[0044] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0045] In one embodiment, a customer touchpoint intelligent identification and intelligent follow-up device is provided, which corresponds one-to-one with the customer touchpoint intelligent identification and intelligent follow-up device in the above embodiment. For example... Figure 8As shown, this intelligent customer touchpoint identification and follow-up device includes a touchpoint data acquisition module 101, a key touchpoint module 102, a follow-up priority generation module 103, an intelligent strategy matching module 104, a strategy execution outreach module 105, and a feedback optimization iteration module 106. Detailed descriptions of each functional module are as follows: The touchpoint data acquisition module 101 is used to acquire touchpoint data generated by customers in multiple interaction channels. The touchpoint data includes customer behavior data, attribute data, and historical interaction records. The key touchpoint module 102 is used to process touchpoint data based on predefined business rules and preset machine learning models to identify key touchpoints in the customer journey. The follow-up priority generation module 103 is used to generate follow-up priorities for key touchpoints based on the type of key touchpoints, customer profile features, and customer intent scores generated by machine learning models. The intelligent strategy matching module 104 is used to match the corresponding intelligent follow-up strategy from the preset strategy library according to the type and follow-up priority of the key touchpoint. The strategy execution outreach module 105 is used to execute intelligent follow-up strategies to reach the corresponding customers through at least one communication interface; The feedback optimization and iteration module 106 collects customer feedback data on the outreach and optimizes and iterates the machine learning model and intelligent follow-up strategy based on the feedback data.

[0046] In one embodiment, the key contact module 102 is specifically used for: Based on business rules, identify interruption events that occur in key stages of the preset business process and treat these interruption events as the first type of key touchpoints. Touchpoint data is fed into a machine learning model to generate customer intent scores, and behavioral events with customer intent scores above a predetermined threshold are identified as second-category key touchpoints.

[0047] In one embodiment, the priority generation module 103 is specifically used for: Based on profile features, determine customer value assessment results and obtain the duration of key touchpoints; The system uses customer intent scores, customer value assessment results, and the duration of key touchpoints as inputs to calculate a quantified priority score using a pre-defined priority ranking algorithm.

[0048] In one embodiment, the intelligent policy matching module 104 is specifically used for: Predefine at least one initial reach strategy for different types of key touchpoints; Based on profile features and historical interaction records, the content of the initial outreach strategy is personalized to generate a customized intelligent follow-up strategy. The intelligent follow-up strategy includes at least one of the following: differentiated price quotation push, time-limited incentive information push, human agent task allocation, and automatic outbound call task triggering.

[0049] In some embodiments, the policy execution outreach module 105 is specifically used for: Based on intelligent follow-up strategies and customer preferences, select one or more of the following methods to reach customers: SMS, instant messaging, voice calls, or online application push notifications.

[0050] In some embodiments, the feedback optimization iteration module 106 is specifically used for: Record and analyze customer conversion behavior data generated after outreach; Customer conversion behavior data is added as new samples to the training dataset, and the machine learning model is retrained based on the updated training dataset. Based on the conversion effect analysis results of each type of intelligent follow-up strategy, adjust the strategy matching rules or strategy content in the strategy library.

[0051] This invention provides a customer touchpoint intelligent identification and intelligent follow-up device. It acquires touchpoint data generated by customers across multiple interaction channels, including behavior, attributes, and historical records. This data is processed based on predefined business rules and machine learning models to identify key touchpoints in the customer journey. Follow-up priorities are generated for key touchpoints based on their type, customer profile, and customer intent score generated by the model. A corresponding intelligent follow-up strategy is matched from a preset strategy library according to the key touchpoint type and follow-up priority. The strategy is executed through at least one communication interface to reach the customer. Feedback data is collected to optimize the model and strategy. This invention solves the technical problems of low customer conversion efficiency and lost high-value sales opportunities caused by the inability to effectively identify potential customer intent and the inability to intelligently prioritize and allocate resources to a large number of customers awaiting follow-up.

[0052] Specific limitations regarding the intelligent customer touchpoint identification and follow-up device can be found in the limitations of the intelligent customer touchpoint identification and follow-up method described above, and will not be repeated here. Each module in the aforementioned intelligent customer touchpoint identification and follow-up device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0053] In one embodiment, a computer device is provided, such as Figure 9As shown, it includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it performs the following steps: Acquire customer touchpoint data across multiple interaction channels. Touchpoint data includes customer behavior data, attribute data, and historical interaction records. Based on predefined business rules and preset machine learning models, touchpoint data is processed to identify key touchpoints in the customer journey. Based on the type of key touchpoints, customer profile features, and customer intent scores generated by machine learning models, follow-up priorities are generated for key touchpoints. Based on the type and follow-up priority of key touchpoints, the corresponding intelligent follow-up strategy is matched from the preset strategy library; Execute intelligent follow-up strategies to reach the corresponding customers through at least one communication interface; Collect customer feedback data on outreach, and optimize and iterate machine learning models and intelligent follow-up strategies based on the feedback data.

[0054] In one embodiment, such as Figure 10 As shown, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, performs the following steps: Acquire customer touchpoint data across multiple interaction channels. Touchpoint data includes customer behavior data, attribute data, and historical interaction records. Based on predefined business rules and preset machine learning models, touchpoint data is processed to identify key touchpoints in the customer journey. Based on the type of key touchpoints, customer profile features, and customer intent scores generated by machine learning models, follow-up priorities are generated for key touchpoints. Based on the type and follow-up priority of key touchpoints, the corresponding intelligent follow-up strategy is matched from the preset strategy library; Execute intelligent follow-up strategies to reach the corresponding customers through at least one communication interface; Collect customer feedback data on outreach, and optimize and iterate machine learning models and intelligent follow-up strategies based on the feedback data.

[0055] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0056] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0057] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0058] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for intelligent identification and intelligent follow-up of customer touchpoints, characterized in that, include: Acquire customer touchpoint data generated across multiple interaction channels, including customer behavior data, attribute data, and historical interaction records; Based on predefined business rules and preset machine learning models, the touchpoint data is processed to identify key touchpoints in the customer journey. Based on the type of the key touchpoint, the customer profile features, and the customer intent score generated by the machine learning model, a follow-up priority is generated for the key touchpoint. Based on the type of the key touchpoint and the follow-up priority, a corresponding intelligent follow-up strategy is matched from a preset strategy library; The intelligent follow-up strategy is executed through at least one communication interface to reach the corresponding customer; Collect customer feedback data on outreach, and optimize and iterate the machine learning model and the intelligent follow-up strategy based on the feedback data.

2. The intelligent customer touchpoint identification and intelligent follow-up method according to claim 1, characterized in that, The process, based on predefined business rules and machine learning models, analyzes the touchpoint data to identify key touchpoints in the customer journey, including: Based on the business rules, interruption events that occur in key stages of a preset business process are identified, and these interruption events are designated as first-class key touchpoints. The touchpoint data is input into the machine learning model to generate the customer intent score, and behavioral events with customer intent scores higher than a predetermined threshold are identified as second-category key touchpoints.

3. The intelligent customer touchpoint identification and intelligent follow-up method according to claim 2, characterized in that, The machine learning model is an ensemble learning model trained based on multi-dimensional features, including time features, behavior frequency features, customer attribute features, and product association features.

4. The intelligent customer touchpoint identification and intelligent follow-up method according to claim 1, characterized in that, The process of generating follow-up priorities for key touchpoints based on the type of the key touchpoint, the customer profile features, and the customer intent score generated by the machine learning model includes: Based on the profile features, determine the customer value assessment result and obtain the duration of the key touchpoints; The customer intent score, the customer value assessment result, and the duration of the key touchpoints are used as inputs, and a quantified priority score is calculated using a preset priority ranking algorithm.

5. The intelligent customer touchpoint identification and intelligent follow-up method according to claim 1, characterized in that, The step of matching a corresponding intelligent follow-up strategy from a preset strategy library based on the type of the key touchpoint and the follow-up priority includes: For each of the different types of key touchpoints, at least one initial reach strategy is predefined; Based on the profile features and the historical interaction records, the content of the initial outreach strategy is personalized to generate a customized intelligent follow-up strategy. The intelligent follow-up strategy includes at least one of the following: differentiated price quotation push, time-limited incentive information push, human agent task allocation, and automatic outbound call task triggering.

6. The intelligent customer touchpoint identification and intelligent follow-up method according to claim 1, characterized in that, The step of executing the intelligent follow-up strategy to reach the corresponding customer through at least one communication interface includes: Based on the intelligent follow-up strategy and customer preferences, one or more of the following methods can be selected for a combined outreach: SMS, instant messaging, voice calls, or online application push notifications.

7. The intelligent customer touchpoint identification and intelligent follow-up method according to claim 1, characterized in that, The process of collecting customer feedback data on outreach and optimizing and iterating the machine learning model and the intelligent follow-up strategy based on the feedback data includes: Record and analyze customer conversion behavior data generated after outreach; The customer conversion behavior data is added as a new sample to the training dataset, and the machine learning model is retrained based on the updated training dataset. Based on the conversion effect analysis results of each type of intelligent follow-up strategy, adjust the strategy matching rules or strategy content in the strategy library.

8. A customer touchpoint intelligent identification and intelligent follow-up device, characterized in that, include: The touchpoint data acquisition module is used to acquire touchpoint data generated by customers in multiple interaction channels. The touchpoint data includes customer behavior data, attribute data, and historical interaction records. The key touchpoint module is used to process the touchpoint data based on predefined business rules and preset machine learning models to identify key touchpoints in the customer journey. The follow-up priority generation module is used to generate a follow-up priority for the key touchpoints based on the type of the key touchpoints, the customer profile features, and the customer intent score generated by the machine learning model. The intelligent strategy matching module is used to match the corresponding intelligent follow-up strategy from the preset strategy library according to the type of the key touchpoint and the follow-up priority. The strategy execution outreach module is used to execute the intelligent follow-up strategy to reach the corresponding customer through at least one communication interface; The feedback optimization and iteration module collects customer feedback data on the outreach and optimizes and iterates the machine learning model and the intelligent follow-up strategy based on the feedback data.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the customer touchpoint intelligent identification and intelligent follow-up device as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the customer touchpoint intelligent identification and intelligent follow-up device as described in any one of claims 1 to 7.

Images