KR20210118634A - Call center service efficiency improvement system based on data and method thereof - Google Patents

Abstract

The present invention relates to a system for improving work efficiency of call center based on data, which quantitatively collects consultation content and results of call center agents who are assigned a list of prospects to evaluate the agents and change a method of assigning the list of prospects, generates a neural network model by generating tagged training data for training the neural network model on the basis of quantitative counseling information, and generates consultation support information on the basis of the neural network model, thereby increasing work efficiency of a call center, and a method thereof. According to the present invention, the system quantitatively grasps a counseling process by objectively and diversely checking a counselor's actual counseling history, and generates tagged learning information for training a deep learning model by clearly dividing a relationship between quantitatively grasped counseling process information and results of contract success (or meeting reservation success) into input and output. Through the deep learning model trained by using the learning information, the system filters malicious customers with minimal counseling resources or generates and provides counseling support information for enhancing the counseling effect for good customers to agents, thereby providing effects of increasing the success rate of prospective customer counseling, evaluating the agents to be objectively recognized, and sharing counseling skills of agents with a high success rate. The system comprises a customer selection server, an agent terminal, and an agent evaluation server.

Description

Data-based call center work efficiency improvement system and method thereof

The present invention relates to a data-based call center work efficiency improvement system and method, and in particular, quantitatively collects the consultation contents and results of a call center agent assigned a potential customer list, evaluates the agent, and then changes the potential customer list assignment method, A data-based call center work efficiency improvement system that improves call center work efficiency by generating tagged learning data for neural network model learning based on quantitative counseling information, learning the neural network model, and then generating counseling support information based on this it's about how

Traditionally, call centers provide information about services or products based on inbound calls, are used for after-sales service to relieve inconveniences in using services or products, or provide services or products to potential customers based on outbound calls. It is used for marketing purposes to induce sales by providing product information.

Among them, for marketing, contact potential customers first and provide information about products or services to induce customers' interest, and based on this, schedule a meeting to provide detailed product descriptions or induce direct contracts. Call center-based counseling accounts for a very large portion of sales of services and products.

The call center for marketing collects information about potential customers through various methods to increase consultation efficiency, distributes information about these potential customers to counselors and conducts consultations, Or, based on the success of the reservation), it judges the success of the consultation and evaluates the agent accordingly.

In fact, for the success of such consultations, it is necessary to effectively collect information on potential customers with high interest in services or products. Information of site users or users participating in events is also collected as potential customer information for related products or services. Furthermore, since potential customer information is purchased at a cost from a separate potential customer information providing service provider, considerable cost and effort are required to collect such potential customer information.

In this way, potential customer information obtained through cost and effort is distributed to agents, and the success of each agent's consultation is checked based on the success of a contract or meeting reservation to evaluate the agent's consulting work. By allowing customer information to be assigned, agents' motivation to consult is increased and product or service sales efficiency is increased.

However, this traditional method does not adequately reflect various environmental factors that may occur in actual counseling or factors external to the agent's ability, so the reliability or objectivity of the evaluation is relatively low. There may be dissatisfaction, and it is difficult to establish an active marketing policy because the efficiency is influenced by the agent's personal ability or luck, and there is a limit in improving the counseling efficiency due to frequent turnover of counselors with excellent personal ability.

In order to solve this problem, conventionally, after converting the contents of the consultation into text, the text contents are identified, the tendency of agents and customers is classified, the adequacy of the consultation contents is judged numerically, and then the consultation call with the problem is selected by the manager. Techniques have emerged that allow direct confirmation or add weights based on information that classifies counselors and customers' propensity to process counseling content based on statistics to confirm counseling quality. However, there are significant errors because it is difficult to score the counseling voice contents by parameters according to simple criteria or to simply classify them by a mechanical classification method in reality because it is difficult to cover all various situations. It is used only as simple evaluation information after consultation to check the efficiency of the call center, to point out the wrong part of a specific agent, or to allocate more potential customer information to an agent with a high evaluation score. Measures have limitations that are not easy.

Korean Patent Registration No. 10-1543366, [Method and Apparatus for Evaluating Calls or Agents Using Customer Sentiment Index] Korean Patent Registration No. 10-1824242, [Counseling quality evaluation system]

An object of the embodiments of the present invention for improving the above-described problems is to enable a quantitative understanding of the counseling process by objectively and variously confirming the actual counseling details of the counselor, and information on the quantitatively grasped counseling process and its By clearly dividing the relationship of the result, the success of the consultation (meeting reservation or contract success) into input and output, the tagged learning information for learning the deep learning model is generated, and the deep learning model trained using this learning information is used. It is to provide a data-based call center work efficiency improvement system and method to improve the efficiency of potential customer consultation through

Another object of the embodiments of the present invention is to input consultation information including contact information, keywords and voice tone and speed according to the call content, including the call time and call duration and number of calls to potential customers collected from the agent terminal as input-related information. Tagging and tagging result information including failure, reservation, or contract success according to the final consultation result as output-related information to generate learning information in which objective input and objective output are clearly tagged, and based on this learning information, multiple By learning one or more neural network models based on tagging learning data, it is possible to predict the final result according to the counseling process as early as possible, or to generate support information including counseling time, distribution, and recommended counseling content for appropriate counseling and provide it to the agent terminal. It is to provide a data-based call center work efficiency improvement system and method to increase the possibility of success.

Furthermore, another object of the embodiments of the present invention is to add information on the potential customer information collection method or source to the tagged learning information as input, thereby pre-evaluating the quality of information collected for potential customers or collecting potential customers. Data-based call center work efficiency improvement system that generates support information for classification and delivers it to the audience collection server to reduce the cost or effort for information on low-probability customers and to obtain information about potential customers economically and to provide a method therefor.

Data-based call center work efficiency improvement system according to an embodiment of the present invention for achieving the above object is a customer selection server for collecting potential customer information and allocating potential customer information to a call center agent terminal, and potential from the customer selection server Receive customer information, conduct a call consultation with the prospective customer, and contact information including call consultation attempts, call success, and call duration information, and voice tone and speed according to voice analysis information for the call content; An agent terminal that generates consultation content information including text-converted consultation information and results information including failure, reservation or contract success information according to the consultation result, contact information collected through the agent terminal, consultation information and To learn the deep learning model by collecting the result information, a plurality of deep learning models are trained by generating learning data with input and output tagged for learning the deep learning model, and the agent is evaluated through the learned deep learning model and the evaluation information is sent to the customer selection server. It includes an agent evaluation server that provides it.

As an example, the agent evaluation server further includes the collection source information about the potential customer collected from the customer selection server as an input of the training data, and calculates the expected value according to the potential customer collection source through deep learning model learning for the success of the consultation can be provided to the customer selection server.

For example, the deep learning model of the agent evaluation server includes a success rate prediction model according to the collection source of potential customers and the collected counseling information of the agent's repeated counseling process. If the predicted success rate is less than or equal to a preset criterion before the completion of counseling, counseling abandonment information may be generated and transmitted to the counselor terminal.

Furthermore, the agent evaluation server collects personal information about potential customers from the customer selection server, reflects the agent consultation schedule information of the agent terminal, and consults at least one of consultation time, call interval for consultation, call time, and keyword of consultation After generating the support information, it is provided to the agent terminal, and in the process of consulting for the same potential customer, counseling support information for follow-up counseling can be repeatedly generated and provided.

Data-based call center work efficiency improvement method according to another embodiment of the present invention, the customer screening server collecting potential customer information and allocating potential customer information to the call center agent terminal, the agent terminal is the potential customer information from the customer selection server , and performing a call consultation for the potential customer, and the agent terminal attempts to consult with the potential customer, whether the call was successful, contact information including call time information, and voice analysis of the call Creating consultation content information including voice tone and speed and text-converted consultation information according to the information, and generating result information including failure, reservation or contract success information according to the consultation result; and sending the consultation content information and result information to the agent evaluation server; After generating training data, training a plurality of deep learning models, evaluating agents through the learned deep learning models, and providing the evaluation information to a customer selection server.

As an example, the agent evaluation server collects personal information about potential customers from the customer selection server, reflects the agent consultation schedule information of the agent terminal, and at least one or more of consultation time, call interval for consultation, call time, and keyword of consultation The step of providing the counseling support information to the agent terminal after generating, and the agent evaluation server may further include the steps of repeatedly generating and providing counseling support information for follow-up counseling in the process of consulting the same potential customer in the agent terminal can

The data-based call center work efficiency improvement system and method according to an embodiment of the present invention objectively and variously confirm the actual counseling details of an agent to quantitatively identify the counseling process, and quantitatively identify information on the counseling process and the results thereof After creating the tagged learning information for deep learning model learning by clearly dividing the relationship for the success of the contract (or meeting reservation success) into input and output, through the deep learning model trained using this learning information It is possible to increase the success rate of counseling with potential customers and to evaluate agents that can be objectively recognized by filtering malicious customers with the minimum number of counseling resources or creating and providing counseling support information to enhance the counseling effect for good customers. It has the effect of being able to share the counseling skills of agents with high success rates.

Furthermore, an embodiment of the present invention tags the contact information on the call time and call holding time and frequency of the potential customer call collected from the agent terminal, the keyword according to the call content, and the counseling content information including the voice tone and speed as input-related information. and tagging result information including failure, reservation, or contract success according to the final consultation result as output-related information to generate tagged learning information with clear objective input and objective output, and multiple tagging based on this learning information By learning one or more neural network models based on learning data, it is possible to predict the final result of the counseling process as early as possible, or to generate support information including counseling timing, distribution, and recommended counseling content for appropriate counseling and provide it to the agent terminal. This increases the chances of success.

Furthermore, the embodiments of the present invention have the effect of enabling the potential customer collection server to economically secure the appropriate potential customer information by learning the information on the potential customer information collection method or source with a deep learning model and predicting the value according to its success. there is

1 is a configuration diagram showing a data-based call center work efficiency improvement system according to an embodiment of the present invention.
2 is a block diagram showing an exemplary configuration of a counselor terminal according to an embodiment of the present invention.
3 is a configuration diagram showing an exemplary configuration of a customer selection server according to an embodiment of the present invention.
4 is a configuration diagram showing an exemplary configuration of an agent evaluation server according to an embodiment of the present invention.
5 is a conceptual diagram illustrating a basic operation process of a data-based call center work efficiency improvement system according to an embodiment of the present invention.
6 is a conceptual diagram illustrating an additional operation process of a data-based call center work efficiency improvement system according to an embodiment of the present invention.

The present invention as described above will be described in detail with reference to the accompanying drawings and embodiments.

It should be noted that the technical terms used in the present invention are only used to describe specific embodiments, and are not intended to limit the present invention. In addition, the technical terms used in the present invention should be interpreted as meanings generally understood by those of ordinary skill in the art to which the present invention belongs, unless otherwise defined in particular in the present invention, and excessively comprehensive It should not be construed in the meaning of a human being or in an excessively reduced meaning. In addition, when the technical term used in the present invention is an incorrect technical term that does not accurately express the spirit of the present invention, it should be understood by being replaced with a technical term that can be correctly understood by those skilled in the art. In addition, general terms used in the present invention should be interpreted as defined in advance or according to the context before and after, and should not be interpreted in an excessively reduced meaning.

Also, as used herein, the singular expression includes the plural expression unless the context clearly dictates otherwise. In the present invention, terms such as "consisting of" or "comprising" should not be construed as necessarily including all of the various elements or several steps described in the invention, some of which elements or some steps are included. It should be construed that it may not, or may further include additional components or steps.

In addition, terms including ordinal numbers such as first, second, etc. used in the present invention may be used to describe the components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, a first component may be referred to as a second component, and similarly, the second component may also be referred to as a first component.

Hereinafter, a preferred embodiment according to the present invention will be described in detail with reference to the accompanying drawings, but the same or similar components are assigned the same reference numerals regardless of reference numerals, and redundant description thereof will be omitted.

In addition, in the description of the present invention, if it is determined that a detailed description of a related known technology may obscure the gist of the present invention, the detailed description thereof will be omitted. In addition, it should be noted that the accompanying drawings are only for easy understanding of the spirit of the present invention, and should not be construed as limiting the spirit of the present invention by the accompanying drawings.

In particular, the counselor terminal of the present invention is a smart phone having a communication function, a portable terminal, a mobile terminal, a personal digital assistant (PDA), a PMP (Portable Multimedia Player) ) terminals, personal computers, notebook computers, slate PCs, tablet PCs, ultrabooks, etc. may include various terminals, and the communication network is a well-known variety of wired and wireless communication. method can be applied. In particular, the voice call function for a call with a customer may be provided in hardware or software in the terminal, and may be configured as an external type in which a separate voice call means is separately connected. In this case, the voice call means is controlled by the agent terminal and a means for collecting call contents may be added.

In addition, the customer selection server or the agent evaluation server may be of various types of server configurations including a single computer, a parallel computer, and a distributed computer, and in particular, such a server may include a plurality of physical or logical servers through a cloud configuration.

Another configuration according to the embodiment of the present invention will be described with reference to FIGS. 1 to 6 .

1 is a configuration diagram showing a data-based call center work efficiency improvement system according to an embodiment of the present invention.

As shown, the customer selection server 200 for allocating information on the selected potential customer to the agent terminal 100 of the call center, and a voice call to the customer terminal 400 using the assigned contact information in the potential customer information Counselor evaluation server 300 that attempts to conduct counseling and collects objective information about counseling detected by the agent terminal 100 and the agent terminal 100 that detects objective information related to counseling, checks the counseling situation, and evaluates the agent (300) ) is included.

The customer selection server 200 collects potential customer information in various ways and allocates potential customer information to the call center agent terminal 100, but according to the evaluation of each agent, a different number of customer terminals (100a, 100b) 400a, 400b) Allocate prospect information with information.

The agent terminal 100 receives the potential customer information (personal information about the potential customer, customer terminal information, the reason for selecting the potential customer, etc.) from the customer selection server 100 , performs a call consultation on the potential customer, and performs a call consultation Contact information including attempts, call success, and call time information, and counseling content information including voice tone and speed and text-converted counseling information according to voice analysis information on the call content are generated, and the result is Result information including failure, reservation or contract success information is generated and transmitted to the agent evaluation server 300 . Contact information related to such consultation, consultation content information, and result information may be transmitted at once after the results are obtained. Since it is common to determine whether or not to fail after trying , it is preferable to transmit the information to the agent evaluation server 300 if the counseling is attempted or the counseling is in progress, even if the result of the counseling is not obtained.

The agent evaluation server 300 collects contact information, consultation information, and result information collected through the agent terminal 100 to generate learning data in which input and output are tagged for deep learning model learning, thereby generating a plurality of deep learning models. After learning, the agent is evaluated through the learned deep learning model, and the evaluation information is provided to the customer selection server 200 .

As described above, in the present invention, the agent evaluation server 300 collects various quantitative information about consultation using voice calls through the agent terminal 100 and collects the results (failure, reservation success, contract success) according to the consultation. By doing so, it is possible to generate learning information having an accurate input and output relationship. In deep learning models using various types of neural networks, interest in unsupervised learning is growing, which provides a lot of data and automatically classifies it and then specifies the classified object. If it is not massive, the prediction performance is low, and it is difficult to predict the correlation of complex inputs due to the nature of unsupervised learning. Therefore, for accurate analysis, it is important to train the neural network constituting the deep learning model using high-quality training data.

Therefore, in the present invention, in order to learn the deep learning model, the appropriateness of input information according to the counselor's consultation or the collected input information is generated by generating learning data that is accurately tagged with respect to input and output and learning multiple deep learning models. An evaluation model that evaluates agents based on the expected success rate can be advanced, and through this, an evaluation result that agents can accept can be provided.

Furthermore, it is possible to learn and utilize various deep learning models, such as making predictions about at least some of the inputs to understand the counselor's counseling problems or to obtain a desired result.

For example, one of the deep learning models of the agent evaluation server 300 may be a success rate prediction model according to the collection source of potential customers and the collected counseling information of the agent's repeated counseling process. In the case of using such a success rate prediction model, if the predicted success rate is less than or equal to a preset standard before the number of times of counseling for the same potential customer reaches a preset number, a counseling abandonment proposal may be transmitted to the counselor terminal 100 . Through this, it is possible to increase the overall counseling efficiency by reducing the waste of additional counseling resources for potential customers whose counseling failure is predicted in the early stage of counseling.

On the other hand, the illustrated agent evaluation server 300 receives the collection source information for the potential customer from the customer selection server 200, further includes the collection source information of the potential customer as an input of learning data, Through deep learning model training, it is possible to estimate the expected value according to the source of the audience collection. That is, by collecting various information about counseling from the agent terminal 100, the counselor's counseling process and the result for the potential customer having a specific collection source can be learned, and through this, the potential customer of the specific collection source can be treated as a normal customer. It is possible to predict the expected success rate of counseling when assigned to an agent, so that it is possible to calculate the estimated value of the potential customer information collection cost.

The agent evaluation server 300 provides the expected value of potential customer information of this specific collection source to the customer selection server 200, and the customer selection server 200 provides an economical potential based on the expected value of the potential customer information for each collection source. Collect customer information. For example, if the success rate of consulting with a prospect from a specific source is low, the cost of collecting the potential customer information should be below a certain level. Conversely, if the success rate of consulting with a potential customer from a specific source is high, the value of the potential customer information is high. Here, the source of audience information may be multiple channels that collect the relevant audience information (commercial database, information collected by events, activity analysis of members, etc.) You may.

Additionally, the agent evaluation server 300 collects personal information about potential customers from the customer selection server 200, and reflects the agent consultation schedule information of the agent terminal 100 to provide consultation time, call interval for consultation, and call After generating at least one or more counseling support information among time, counseling keywords, and suggestions (attitude, tone, tone, etc.), it is provided to the agent terminal 100, followed by the process of counseling for the same potential customer (repeated counseling) It is possible to create and provide counseling support information for counseling.

For example, for the first consultation of potential customers with specific personal information and collection sources, the agent evaluation server 300 mainly uses keywords to reserve the next consultation by making a short call within 1 minute in the morning, but in a low voice. Counseling support information suggesting to proceed with the consultation slowly can be transmitted to the counselor terminal 100, and if the first consultation is made, the success of the second consultation is determined based on the contact information and consultation information obtained through the first consultation. Counseling support information of at least one of a counseling time for counseling, a call interval for counseling, a call time, a keyword for counseling, and a suggestion is generated and provided to the counselor terminal 100 .

Since this counseling support information is generated using an analysis model that has learned the counseling success learning data of numerous agents that have been previously collected and learned, it is possible for multiple agents to share the counseling method of agents with good evaluations, so call center agents who frequently change jobs. The overall call center counseling quality (that is, the likelihood of counseling success) can be improved irrespective of the employee's turnover.

2 is a block diagram showing an exemplary configuration of the agent terminal 100 according to an embodiment of the present invention.

As shown, the server communication unit 110 for exchanging information with the customer selection server 200 and the agent evaluation server 300, the customer communication unit 150 for communicating with the customer by voice or video, and the customer communication unit A customer call detection unit 140 that detects a potential customer call attempt and call time through 150, and a voice tone and speed for the call content during a potential customer call through the customer call unit 150, and if necessary, the call content The call contents management unit 160 for recording (locally, or stored in the agent evaluation server 300), and the call contents management unit 160 to analyze the voice from the call obtained through the conversion into text information and analyze keywords or context a voice analysis unit 170 to, a customer reservation management unit 120 for registering a reservation of a customer whose meeting is reserved through consultation, and a customer contract management unit 130 for registering a contract of a customer whose contract is completed through consultation; Potential customer information is collected through the server communication unit 110 and provided to the agent, a call connection is provided to the potential customer through the customer call unit 150, and the customer call detection unit 140 and the call content management unit 160 and Customers who provide contact information and consultation information through the voice analysis unit 170 and result information through the customer reservation management unit 120 and the customer contract management unit 130 to the agent evaluation server 300 through the server communication unit 110 and a consultation management unit 180 .

Here, the customer consultation management unit 180 receives the consultation support information of at least one of the consultation time point, the call interval for consultation, the call time, and the keyword of the consultation through the agent evaluation server 300, and then further provides a suggestion unit to the agent may include

For such consultation support information, the agent terminal 100 may set a schedule for consultation in advance (customer consultation management unit 180) after receiving information about potential customers from the customer selection server 200, Consultation schedule information for potential customers and existing customers in consultation may be further provided to the agent evaluation server 300 . The agent evaluation server 300 is based on the consultation schedule information for the potential customer and the existing customer, the consultation schedule proposal, the consultation keyword proposal, or the consultation recommendation information proposal for the potential customer or the customer in the process of consultation (tone for consultation, emphasize keywords, attitudes, etc.).

3 is a configuration diagram showing an exemplary configuration of a customer selection server 200 according to an embodiment of the present invention.

As shown, the customer selection server 200 includes a customer information collection unit 210 that collects personal information or access information about the customer, and customer activity information about the customer collected through the customer information collection unit 210 . A customer information management unit 220 for classifying customers based on the analysis information on the collected customer information, and a potential customer extraction unit 240 for extracting potential customers based on the customer classification information of the customer information management unit 220; Distribution of potential customers extracted based on the agent evaluation information collected through the agent evaluation server linking unit 260 and the agent evaluation server linking unit 260 that collects evaluation information about the agent in conjunction with the agent evaluation server 300 and a counselor terminal communication unit 230 that communicates with the agent terminal 100 and provides potential customer information distributed by the agent allocator 250 to the agent terminal 100 . Of course, this configuration is an example of collecting information about potential customers directly, and instead of having such a direct potential customer selection configuration, information about related service subscribers is collected as potential customer information for new services by linking with related servers. It may also include a configuration in which information of a specific website user or user participating in an event is collected as potential customer information for a related product or service.

Here, the customer information collection unit 210 may collect age, gender, place of residence, address, hobbies, interest, and history of using similar services/products as personal information about the customer, and access target information ( URL, etc.), access frequency, and length of stay can be collected.

The customer information management unit 220 may collect search keywords, fields of interest, content interest, etc. of the customer as activity information of the customer collected through the customer information collection unit 210 . Customers are classified based on the collected customer information and analysis information on activity information. Typically, such customer classification may be classified according to an expected level of interest in a service or product to be consulted.

Furthermore, potential customer information may be purchased at a cost from a third party potential customer information providing service provider. For example, customers who have signed up for a specific insurance service can be selected as potential customers who will be interested in additional products and purchased the information, or they can purchase information about insurance service subscribers through home shopping or online direct method without a separate person in charge, or purchase information from related companies/ After being provided by the department, they are introduced to the relevant insurance service manager and use potential customer information in a way to suggest new products. As another example, if you are a golf member and owner of a premium vehicle, you may purchase information about the owner of a leased vehicle that is 3 to 4 years old and use it as a potential customer for a premium vehicle rental product.

In order to collect information about potential customers like this, a lot of effort is required or the environment must be prepared.

4 is a configuration diagram showing an exemplary configuration of the agent evaluation server 300 according to an embodiment of the present invention.

As shown, the agent evaluation server 300 basically communicates with the agent terminal to collect or provide consultation-related information and communicate with the agent terminal communication unit 310 and the customer selection server 200 to collect potential customer information and Counseling information that collects counseling schedule information, contact information according to counseling, counseling information and result information from the agent terminal 100 through the customer selection server interlocking unit 330 that provides agent evaluation information and the agent terminal communication unit 310 Potential customer information collected through the collection unit 350 and the customer selection server linkage unit 330 and the collection source information, and the contact information and consultation information collected in the consultation information collection unit are input as input, and the result of the consultation and a learning data generating unit 360 that generates tagged learning data using information as an output.

Tagged training data is an input to predict an output for multiple inputs or to obtain a target output by providing an accurate definition and criterion for the input(s) and an accurate definition and criterion for the result(s) obtained by the corresponding input When some of these are present, it means data that enables deep learning model learning, such as predicting optimal requirements for the remaining inputs. In other words, by learning a large number of tagged training data to the deep learning model, prediction of a result according to a specific input or optimization of inputs to obtain a desired result, optimization of other inputs when a specific input is fixed, etc. are predicted according to model characteristics can be

It is possible to learn various potential customer-related deep learning models through the training data generated by the training data generator 360 . For example, the illustrated counseling situation learning unit 370 is a deep learning model for estimating the expected value according to the potential customer source, a success rate prediction model for filtering out potential customers to give up, counseling time for success, time, keywords, It includes at least one of various deep learning (neural network) models, such as a proposal model for providing a suggestion for at least one of the contexts, and trains such a model using the tagged training data. In addition, the counseling situation learning unit 370 compares the actual counseling content (counseling process and the result) of the agent against the possibility of success for the counselor's counseling (the possibility of success of the success rate prediction model according to the same counseling input, that is, the level of counseling) An agent evaluation model that evaluates agents based on the basis (calculation of evaluation according to the actual counseling process and results compared to the level of counseling) can be applied.

Further, the agent evaluation unit 380 that evaluates the counseling situation for each potential customer of the agent through the agent evaluation model of the consultation situation learning unit 370 and provides it to the customer selection server 200 through the customer selection server linking unit 330 . ), a consultation proposal generation unit 320 for generating consultation support information generated through a proposal model among a plurality of deep learning models, and information for generating learning data from the agent terminal 100 and the customer selection server 200 After generating learning data through the generating unit 360 by collecting and generating learning data, a plurality of deep learning models are trained, the agent is evaluated through the learned deep learning model, and counseling support information to increase the success rate is generated. The generated counseling support information is provided to the counselor terminal 100 , and a counseling optimization management unit 340 that calculates an expected value for the collected potential customers and provides it to the customer selection server 200 is included.

As such, the agent evaluation server 300 not only evaluates the agent, but also provides various counseling proposals to the agent terminal 100 to support the agent's counseling success, while checking the quality of potential customer information in the first place, and By checking the appropriateness of cost and providing quality information on potential customer information by source to the customer selection server 200, it is also supported to enable efficient potential customer information collection.

5 is a conceptual diagram illustrating a basic operation process of a data-based call center work efficiency improvement system according to an embodiment of the present invention.

As shown in FIG. 5A , the customer selection server collects potential customer information and allocates different numbers of potential customer information to the call center agent terminal, and then provides the corresponding potential customer information to the agent terminal. This reflects the level of the agent and the agent's previous counseling evaluation.

Thereafter, as shown in FIG. 5B , the counselor terminal performs a call consultation on the potential customer received by the counselor terminal. In this process, through the agent terminal, contact information including attempts to consult with potential customers, call success, and call time information, and voice tone and speed and text-converted counseling information according to voice analysis information on the call content Generates consultation content information including, and generates result information including failure, reservation or contract success information according to the consultation result.

While the process of FIG. 5B is completed or the process of FIG. 5B is in progress, the agent terminal delivers counseling information (ie, contact information, counseling content information, and counseling result information) to the agent evaluation server as shown in FIG. 5C, and the agent evaluation server collects contact information, consultation information, and result information collected through the agent terminal, generates learning data with input and output tagged for learning the deep learning model, and then trains a plurality of deep learning models.

Thereafter, as shown in Figure 5d, the agent is evaluated through the deep learning model learned by counseling for each potential customer or by period, and the evaluation information is provided to the customer selection server, and the customer selection server is the By adjusting the numbers, you can assign a larger number of leads to agents with high ratings and fewer leads to agents with lower ratings.

6 is a conceptual diagram illustrating an additional operation process of a data-based call center work efficiency improvement system according to an embodiment of the present invention.

6a, as shown in FIG. 5a, the customer selection server collects potential customer information, and after allocating different numbers of potential customer information to the call center agent terminal, the corresponding potential customer information is provided to the agent terminal. This reflects the level of the agent and the agent's previous counseling evaluation.

When the agent terminal registers a consultation schedule for a potential customer as shown in FIG. 6b, the agent evaluation server collects information on the existing customer and the newly registered potential customer, and provides consultation proposal information for the specific potential customer to the agent terminal. can provide For example, if there is no actual counseling history yet, based on the source of the potential customer information and personal information about the potential customer, the counselor counseling schedule information of the agent terminal is reflected to reflect the counseling time, call interval for counseling, call time, and keywords of counseling (Information about tone, speed, major keywords, etc.) and proposals may be generated and then provided to the counselor terminal.

Furthermore, as shown in FIG. 6c , when an actual consultation with the customer is performed, the agent terminal provides contact information, consultation information, and result information corresponding to the consultation information to the agent evaluation server, and the agent evaluation server uses the consultation information obtained from the agent terminal. Thus, the counseling optimization model can be trained and proposal information can be generated.

As shown in Fig. 6d, the agent evaluation server provides counseling support information of at least one of the counseling time point, the call interval for counseling, the call time, and the keyword of the counseling to the agent terminal, but the conclusion of the counseling (abandonment, reservation success, contract success) etc.), repeat the process of Figs. 6c and 6d.

Through this process, a high model hit rate can be expected despite the small number of training data, and the reliability of potential customer information can also be increased.

The method according to the embodiments may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the embodiment, or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include a physical hard disk, a solid state drive (SSD), a web hard, a mass data storage means such as cloud storage, a magnetic medium such as a floppy disk and a magnetic tape ( magnetic media, optical media such as CD-ROM and DVD, magneto-optical media such as floppy disk, and ROM, RAM, Included are hardware devices specially configured to store and execute program instructions, such as flash memory and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

In the above, preferred embodiments according to the present invention have been shown and described. However, the present invention is not limited to the above-described embodiments, and without departing from the gist of the present invention appended in the claims, anyone with ordinary skill in the art to which the invention pertains will be able to implement various modifications. .

100: agent terminal 200: customer selection server
300: agent evaluation server 400: customer terminal

Claims (10)

a customer screening server for collecting potential customer information and allocating potential customer information to a call center agent terminal;
Receive potential customer information from the customer screening server, perform call consultation for the potential customer, and contact information including call consultation attempts, call success, and call time information, and voice analysis information for call content a counselor terminal for generating counseling content information including voice tone, speed, and text-converted counseling information, and generating result information including failure, reservation or contract success information according to the counseling result;
By collecting contact information, consultation information, and result information collected through the agent terminal, input and output tagged learning data are generated for deep learning model learning, and multiple deep learning models are trained, and through the learned deep learning model A data-based call center work efficiency improvement system including an agent evaluation server that evaluates agents and provides the evaluation information to a customer selection server.
The method according to claim 1, wherein the agent evaluation server,
Further including the collection source information on the potential customer collected from the customer selection server as an input of learning data, and calculating the expected value according to the potential customer collection source through deep learning model learning for the success of the consultation and providing it to the customer selection server Data-based call center work efficiency improvement system, characterized in that.
The method according to claim 2, The deep learning model of the agent evaluation server includes a success rate prediction model according to the collection source of potential customers and the counseling information collected in the repeated counseling process of the agent, and the number of times of counseling for the same potential customer is Data-based call center work efficiency improvement system, characterized in that before the preset number of times, when the predicted success rate is less than or equal to a preset criterion, the counseling abandonment information is generated and transmitted to the agent terminal.
The method according to claim 1, wherein the agent evaluation server collects personal information about the potential customer from the customer selection server, reflects the agent consultation schedule information of the agent terminal, consultation time, call interval for consultation, call time, Data-based call center work efficiency, characterized in that after generating at least one or more counseling support information among keywords and providing it to the agent terminal, repeatedly generating and providing counseling support information for follow-up counseling in the process of counseling for the same potential customer improvement system.
The method according to claim 1, wherein the agent terminal
a server communication unit for exchanging information with the customer selection server and the agent evaluation server;
a customer call unit for communicating with the customer by voice or video;
a customer call detection unit for detecting a potential customer call attempt and call time through the customer call unit;
a call content management unit for detecting a voice tone and speed for the content of the call during a potential customer call through the customer call unit and recording the call content if necessary;
a voice analysis unit that analyzes voice from the call contents obtained through the call contents management unit, converts it into text information, and analyzes keywords or context;
a customer reservation management unit for registering a reservation of a customer whose meeting is scheduled through consultation;
a customer contract management unit for registering a contract of a customer whose contract has been completed through consultation;
It collects potential customer information through the server communication unit and provides it to the agent, provides a call connection to the potential customer through the customer call unit, contact information and consultation information through the customer call detection unit, call content management unit, and voice analysis unit, and customer reservation Data-based call center work efficiency improvement system, characterized in that it comprises a customer consultation management unit that provides result information through the management unit and the customer contract management unit to the agent evaluation server through the server communication unit.
The method according to claim 5, wherein the customer consultation management unit further comprises a proposal unit that provides the counselor after receiving at least one or more counseling support information among the counseling time point, the call interval for counseling, the call time, and the keyword of the counseling through the agent evaluation server Data-based call center work efficiency improvement system, characterized in that.
The method according to claim 1, wherein the customer selection server
a customer information collection unit that collects personal information or access information about the customer;
a customer information management unit for classifying customers based on the customer activity information about the customer collected through the customer information collection unit or the analysis information on the collected customer information;
a potential customer extraction unit for extracting potential customers based on the customer classification information of the customer information management unit;
an agent evaluation server linking unit for collecting evaluation information on the agent by interworking with the agent evaluation server;
an agent allocating unit for distributing potential customers extracted based on the agent evaluation information collected through the agent evaluation server linkage unit;
Data-based call center work efficiency improvement system, characterized in that it communicates with the agent terminal and comprises an agent terminal communication unit for providing the potential customer information distributed by the agent allocation unit to the agent terminal.
The method according to claim 1, The agent evaluation server
a counselor terminal communication unit that communicates with the counselor's terminal to collect or provide counseling-related information;
a customer screening server linking unit that communicates with the customer selection server to collect potential customer information and provides agent evaluation information;
a counseling information collecting unit for collecting counseling schedule information, contact information according to counseling, counseling information, and result information from the counselor terminal through the counselor terminal communication unit;
Tagged learning data that takes as input the potential customer information collected through the customer selection server linkage unit, the collection source information, and the contact information and consultation information collected in the consultation information collection unit, and outputs the result information about the consultation a learning data generator to generate;
A deep learning model for estimating the expected value according to the source of the potential customer through the learning data generated by the learning data generator, the success rate prediction model for filtering out the potential customer to give up, the counseling time for success, time, keyword, and context Counseling situation learning unit for learning a plurality of deep learning models including at least one of the proposal models for providing at least one suggestion and the agent evaluation model that evaluates the agent based on the actual counseling contents against the possibility of success for the agent's counseling Wow;
a counselor evaluator for evaluating the counseling situation for each potential customer of the counselor through the counselor evaluation model of the counseling situation learning unit and providing it to the customer sorting server through the customer sorting server linking unit;
a counseling suggestion generator for generating counseling support information generated through a suggestion model among the plurality of deep learning models;
It collects information for generating learning data from the agent terminal and customer selection server, generates learning data through the generating unit, and trains multiple deep learning models, evaluates agents through the learned deep learning model, and determines the success rate Data base comprising a counseling optimization management unit that generates counseling support information to increase, provides the generated counseling support information to the agent terminal, calculates the expected value for the collected potential customers, and provides it to the customer selection server Call center work efficiency improvement system.
Step by which the customer selection server collects potential customer information and assigns the potential customer information to a call center agent terminal;
receiving, by the agent terminal, potential customer information from the customer selection server, and performing a call consultation on the potential customer;
Counseling that includes contact information, including information on call consultation attempts, call success, and call time, and voice tone and speed and text-converted counseling information based on voice analysis information for the call content by the agent's terminal generating content information and generating result information including failure, reservation, or contract success information according to a consultation result;
transmitting, by the agent terminal, contact information, consultation content information, and result information to the agent evaluation server;
The agent evaluation server collects contact information, consultation information, and result information collected through the agent terminal to generate learning data with input and output tagged for learning the deep learning model, and then trains a plurality of deep learning models and learns A data-based call center work efficiency improvement method comprising the step of evaluating an agent through a deep learning model and providing the evaluation information to a customer selection server.
The method according to claim 9, wherein the agent evaluation server collects personal information about the potential customer from the customer selection server, and reflects the agent consultation schedule information of the agent terminal to determine the time of the consultation, the call interval for consultation, the call time, the number of consultations A method comprising: generating at least one or more counseling support information among keywords and providing the information to a counselor terminal;
Data-based call center work efficiency improvement method, characterized in that it further comprises the step of repeatedly generating and providing, by the agent evaluation server, counseling support information for follow-up counseling in the process of consulting the same potential customer in the agent terminal.

Images