Methods and systems for generating automatic responses for conversational interactions

By using machine learning models to generate automatic responses, the problem of long response times and high burdens for enterprises when handling customer inquiries has been solved, improving agent response speed and customer experience, and reducing system burden and costs.

CN123125614APending Publication Date: 2026-11-17MAERSK INC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202580029307.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2024-04-04
Filing Date
2025-04-02
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing technologies, enterprises face problems such as long query resolution times, poor customer experience, increased burden on agents to understand and resolve customer queries, and increased processing and storage requirements on customer query management systems when handling customer queries.

Method used

By using machine learning models to generate automated responses, customer intent information can be detected, relevant entities can be extracted, and conversational interactive response templates can be generated, reducing the complexity of agents understanding customer queries and query resolution time.

Benefits of technology

It reduces agent response time, improves customer experience, lowers the processing and storage requirements of the customer query management system, saves electricity costs, and reduces the employer's labor needs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123125614A_ABST
    Figure CN123125614A_ABST
Patent Text Reader

Abstract

A method and server system for generating automated responses for conversational interactions from customers are described. The method, performed by the server system, includes receiving customer intent information for the conversational interaction from a first machine learning (ML) model and performing a set of operations by a second ML model. As part of this set of operations, the method includes detecting the presence of sub-intents in the customer intent information, and, upon detecting a sub-intent, assigning a specific automatable process to the conversational interaction. The method includes extracting multiple relevant entities from the conversational interaction based on relevant text from the conversational interaction. The method also includes accessing a list of tasks associated with the specific automatable process and generating query resolution information based on processing the customer intent information against the task list. Finally, the method includes generating a conversational interaction response template for the conversational interaction based on the multiple relevant entities and the query resolution information.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This disclosure relates to the field of customer service in a conversational ecosystem, and more specifically to electronic methods and sophisticated processing systems for generating automated responses for conversational interactions with users or customers. Background Technology

[0002] Customer service is crucial for all businesses, as it can be a differentiating factor between successful and unsuccessful enterprises. To ensure good customer service, businesses employ agent teams whose task can be to proactively contact customers to resolve their issues. Furthermore, when customers contact the customer service team with their problems, the agents are responsible for resolving those issues. Today, it is essential for businesses to provide good customer service by resolving customer problems quickly and effectively to ensure repeat customers and positive word-of-mouth marketing. Generally, businesses ensure that their customers can submit their inquiries or concerns through various interaction channels, such as phone calls, emails, chat, social media, etc. This interaction between the customer and the agent can be called a conversational interaction. Upon receiving the inquiry, the agent understands the problem the customer faces and finds a suitable solution. In various examples, inquiries can be related to product usage, billing, shipping, product disposal, service, and so on. It is important to note that due to its inherent complexity, preparing an appropriate response to resolve a customer's inquiry is a labor-intensive and time-consuming process.

[0003] As a business's customer base grows, managing customer inquiries becomes considerably more complex because agents must engage in lengthy conversations with customers over extended periods. Consequently, businesses must either hire more agents or assign existing staff to longer working hours to provide a better customer experience. Because businesses struggle to recruit and train new staff, many customers experience delayed or inadequate resolution of their inquiries, resulting in a poor customer experience. Furthermore, as resolution times increase, the burden on the business's customer inquiry management system also increases in terms of processing and storage requirements.

[0004] Therefore, there is a desire to find technical solutions to improve the response time of agents when resolving customer inquiries. Summary of the Invention

[0005] There is a need for technology to overcome one or more of the limitations mentioned above, such as long query resolution times, poor customer experience, the burden on agents to understand and resolve customer queries, and the increased load on the processing and storage requirements of customer query management systems.

[0006] Various embodiments of this disclosure provide methods and systems for reducing the burden on agents by providing recommendations to agents through solution templates. Due to the reduced burden on agents, their response times are significantly improved, which in turn improves the customer experience. Furthermore, because agents spend less time on relevant portals of the customer query management system due to faster query resolution, the processing and storage requirements of the customer query management system are also reduced, which also saves on electricity costs. Furthermore, because the time spent on each query resolution is reduced, the same agent can now serve more customers, thereby also reducing the employer's labor requirements.

[0007] To achieve the above and other objectives of this disclosure, in one aspect, a computer-implemented method for generating automated responses to one or more conversational interactions from one or more customers is disclosed. This computer-implemented method is performed by a server system. The computer-implemented method includes receiving customer intent information for a conversational interaction initiated by a customer from a first machine learning (ML) model. The computer-implemented method further includes performing a set of operations by one or more second ML models. Therefore, as part of the set of operations, the method includes detecting the presence of one or more sub-intents in the customer intent information. After detecting a sub-intent, the computer-implemented method includes assigning a specific automatable process to the conversational interaction, at least in part based on comparing the customer intent information with a predefined set of automatable processes. Further, the computer-implemented method includes extracting a plurality of related entities from the conversational interaction, at least in part based on one or more related texts associated with the conversational interaction. Additionally, the computer-implemented method includes accessing a task list associated with the specific automatable process from a database. The computer-implemented method also includes generating query resolution information for the conversational interaction, at least in part based on processing the customer intent information against the task list. Finally, the computer-implemented method includes generating one or more conversational interaction response templates for the conversational interaction, based at least in part on the plurality of related entities and the query resolution information.

[0008] One advantage of some embodiments is that by receiving customer intent information for conversational interactions from a first ML model, the complex process of understanding the context and requirements of a customer query is minimized. Since the first ML model understands the customer's intent almost instantaneously, the time spent processing queries from customers is reduced.

[0009] Furthermore, detecting the presence of one or more sub-intents within the customer intent information helps determine whether a specific automatable process can be assigned to the conversational interaction. This, as can be understood, also helps reduce query resolution time. Further, accessing the task set associated with that specific automatable process facilitates the generation of query resolution information by executing the corresponding tasks. In this document, the query resolution information includes information related to the appropriate resolution of the problem faced by the customer as mentioned in the conversational interaction.

[0010] In one aspect, the step of receiving the customer intent information includes receiving the conversational interaction initiated by the customer. The step further includes extracting a set of relevant segments from the conversational interaction. Further, it includes generating first relevant text from one or more relevant texts by applying at least partially a first set of preprocessing operations to the set of relevant segments. The step further includes generating the customer intent information for the conversational interaction by the first ML model based at least partially on the first relevant text and a historical conversational interaction dataset.

[0011] One advantage of some embodiments is that extracting the relevant segment set from the conversational interaction allows the server system to process only the relevant text from the query. This reduces the processing load on the server system and improves processing speed. Furthermore, because unnecessary text is discarded, the solution recommendations generated for the agent from the relevant segment set are accurate and precise.

[0012] In one aspect, the computer-implemented method further includes fine-tuning the customer intent information by the first ML model, at least in part, based on the first relevant text and the first predefined rule set.

[0013] One advantage of some implementations is that fine-tuning customer intent information helps the server system to better understand the customer's actual intent, thereby leading to improved results.

[0014] In one aspect, the step of extracting the plurality of related entities includes extracting a first set of related entities from the plurality of related entities. To extract the first set of related entities, the computer-implemented method requires generating second related text from the one or more related texts, at least in part, based on applying a second set of preprocessing operations to the first related text. The computer-implemented method further includes processing the second related text by the one or more second ML models to generate the first set of related entities, at least in part, based on the second related text and a second predefined set of rules.

[0015] One advantage of some embodiments is that by preprocessing the first relevant text, the second relevant text required for efficient extraction of the first relevant entity set can be obtained.

[0016] In one aspect, the step of extracting the plurality of related entities includes extracting a second set of related entities from the plurality of related entities. To extract the second set of related entities, the computer-implemented method requires generating third related text from the one or more related texts, at least in part, based on applying a third set of preprocessing operations to the first related text. The computer-implemented method further includes processing the first related text via the one or more second ML models to generate an intermediate entity set, at least in part, based on a third predefined set of rules. Further, the computer-implemented method includes comparing the third related text with the intermediate entity set to generate the second set of related entities.

[0017] One advantage of some embodiments is that by preprocessing the first relevant text, the third relevant text required for efficient extraction of the second relevant entity set can be obtained.

[0018] On one hand, the steps of processing the customer intent information for the task list include executing each task from the task list to generate query resolution information for the conversational interaction.

[0019] One advantage of some embodiments is that executing the task list via the server system eliminates the need for any kind of manual intervention from the agent. This further helps to reduce query resolution time.

[0020] In one aspect, the step of detecting the presence of one or more sub-intents in the customer intent information includes detecting two or more sub-intents in the customer intent information based at least in part on the recognition of one or more connective texts in the customer intent information. In another aspect, the computer-implemented method further includes assigning non-automated processes to the conversational interaction based at least on the association of the customer intent information with at least two or more sub-intents.

[0021] One advantage of some embodiments is that by assigning non-automatable processes to conversational interactions, the conversational interaction is discarded from automation considerations. This allows the agent to take over the query resolution process for manual handling without wasting the server system's processing resources. This also acts as a security safeguard, ensuring that only the correct information is shared with the client, thus preventing irrelevant information from being pushed to the client.

[0022] According to another embodiment of this disclosure, a server system is disclosed. The server system includes a communication interface and a memory including executable instructions. The server system also includes a processor communicatively coupled to the memory. The processor is configured to execute instructions such that the server system receives, at least partially, customer intent information for a customer-initiated conversational interaction from a first machine learning (ML) model. The server system is further configured to perform a set of operations by one or more second ML models. As part of performing the set of operations, the server system is configured to detect the presence of one or more sub-intents in the customer intent information. Upon detecting a sub-intent, the server system is further configured to assign a specific automatable process to the conversational interaction, at least partially based on comparing the customer intent information with a predefined set of automatable processes. The server system is also configured to extract multiple related entities from the conversational interaction, at least partially based on one or more related texts associated with the conversational interaction. Furthermore, the server system is configured to access a task list associated with the specific automatable process from a database. Additionally, the server system is configured to generate query resolution information for the conversational interaction, at least partially based on processing the customer intent information against the task list. Finally, the server system is enabled to generate one or more conversational interaction response templates for conversational interaction, based at least in part on multiple related entities and query resolution information.

[0023] According to another aspect of this disclosure, a non-transitory computer-readable storage medium is disclosed. This non-transitory computer-readable storage medium includes computer-executable instructions that, when executed by at least one processor of a server system, cause the server system to perform a method. The method includes receiving customer intent information from a first machine learning (ML) model for a customer-initiated conversational interaction. The computer-implemented method further includes performing a set of operations by one or more second ML models. Therefore, as part of the set of operations, the method includes detecting the presence of one or more sub-intents in the customer intent information. Upon detecting a sub-intent, the computer-implemented method includes assigning a specific automatable process to the conversational interaction, at least in part based on comparing the customer intent information with a predefined set of automatable processes. Further, the computer-implemented method includes extracting a plurality of related entities from the conversational interaction, at least in part based on one or more related texts associated with the conversational interaction. Additionally, the computer-implemented method includes accessing a task list associated with the specific automatable process from a database. The computer-implemented method also includes generating query resolution information for the conversational interaction, at least in part based on processing the customer intent information against the task list. Finally, the computer-implemented method includes generating one or more conversational interaction response templates for the conversational interaction, based at least in part on the plurality of related entities and the query resolution information.

[0024] The foregoing description of the invention is merely illustrative and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, other aspects, embodiments, and features will become apparent from the accompanying drawings and the following detailed description. Attached Figure Description

[0025] To gain a more complete understanding of exemplary embodiments of this technology, reference is now made to the following description in conjunction with the accompanying drawings, wherein: Figure 1 These are example representations of conversational interactive environments according to various embodiments of the present disclosure; Figure 2 A simplified block diagram of a server system according to an embodiment of the present disclosure is shown; Figure 3 This is a flowchart depicting an embodiment of the present disclosure of a process for generating an artificial intelligence (AI) solution template for an agent in response to a customer's query email; Figure 4A and Figure 4B Together, they illustrate a process according to embodiments of this disclosure for generating an AI solution template for an agent in response to a specific query email from a customer; and Figure 5 A flowchart illustrating a method for generating one or more automated responses from one or more conversational interactions from one or more customers, according to embodiments of the present disclosure, is shown.

[0026] Unless otherwise stated, the accompanying drawings referenced in this specification should not be construed as being drawn to scale, and such drawings are merely illustrative in nature. Detailed Implementation

[0027] In the following description, numerous specific details are set forth for purposes of explanation in order to provide a thorough understanding of this disclosure. However, those skilled in the art will understand that this disclosure may be practiced without these specific details. Descriptions of well-known components and processing techniques have been omitted so as not to unnecessarily obscure the embodiments herein. The examples used herein are merely to facilitate an understanding of how the embodiments herein may be practiced, and further to enable those skilled in the art to practice the embodiments herein. Therefore, the examples should not be considered as limiting the scope of the embodiments herein.

[0028] References to "an embodiment" or "an embodiment" in this specification mean that a particular feature, structure, or characteristic described in connection with an embodiment is included in at least one embodiment of the present technology. The phrase "in an embodiment" appearing in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a single or alternative embodiment mutually exclusive with other embodiments. Furthermore, various features that may be manifested by some embodiments but not others are described. Similarly, various requirements are described, which may be required by some embodiments but not by others.

[0029] Furthermore, although the following description contains many specific details for illustrative purposes, those skilled in the art will understand that many variations and / or modifications of these details are possible within the scope of this disclosure. Similarly, although many features of this disclosure are described in relation to or in combination with each other, those skilled in the art will understand that many of these features may be provided independently of the others. Therefore, this description of the disclosure is set forth without loss of generality and without imposing any limitation on the disclosure.

[0030] Unless otherwise specifically stated, conditional language such as “can,” “may,” “may,” or “perhaps” should generally be understood in the context to indicate that certain features, elements, and / or steps are included in some embodiments, even though they are not included in other embodiments. Therefore, such conditional language is not generally intended to imply that the stated features, elements, and / or steps are necessary in any one or more embodiments in any way, or that one or more embodiments must include logic for determining, with and without user input or prompting, whether to include these features, elements, and / or steps, or whether to implement these features, elements, and / or steps in any particular embodiment.

[0031] Unless otherwise specifically stated, delimited language such as the phrase “at least one of X, Y or Z” should generally be understood in context as representing that an item, item, etc., may be X, Y or Z or any combination thereof (e.g., X, Y and / or Z). Therefore, such delimited language is generally not intended and should not imply that some embodiments require at least one X, at least one Y or at least one Z to be present.

[0032] Unless otherwise explicitly stated, articles such as “a” or “an” should generally be interpreted as including one or more of the described items. Thus, phrases such as “server system configured to…” are intended to include one or more of the described server systems / processors. One or more such enumerated devices may also be collectively configured to perform the descriptions. For example, “processors configured to perform descriptions A, B, and C” could include a first processor configured to perform description A, which works in conjunction with a second processor configured to perform descriptions B and C. The same applies to the use of definite articles used to introduce descriptions of embodiments. Furthermore, even if a specific number of the introduced descriptions of embodiments are explicitly stated, those skilled in the art will recognize that such a description should generally be interpreted as meaning at least the number described (e.g., simply stating “two descriptions” without other modifiers generally indicates at least two descriptions or two or more descriptions).

[0033] Those skilled in the art will understand that the terms used herein are generally intended to be “open-ended” terms (e.g., the terms “including” or “comprising” should be interpreted as “including but not limited to”, the term “having” should be interpreted as “having at least”, and the terms “include” or “comprise” should be interpreted as “including but not limited to”, etc.).

[0034] For clarity, the terms 'customer,' 'user,' 'client,' 'consumer,' or 'buyer' (which may be used interchangeably herein) refer to an existing or potential user of the products and / or services offered by the business or service provider. Furthermore, the terms 'customer representative' or 'agent' as used herein refer to a human agent, chatbot, self-service system, and, in a general sense, any human-machine interface or machine interface that interacts with the business's customers.

[0035] It is understood that conversational interaction between agents and clients corresponds to written digital communication such as email. It plays a crucial role in bridging the gap between clients and agents. Agents are expected to respond quickly to client inquiries with solutions that accurately match the client's requirements. However, as the number of clients and their inquiries increases, this becomes cumbersome for agents. In scenarios where agents need to perform certain tasks, the complexity of obtaining the context of the conversation from the client increases. These tasks might include accessing different internal web portals to check the status of certain processes, documents, etc., analyzing previous traces of communication with the client, analyzing past communications between the client and another agent, and so on. Furthermore, agents also need to take necessary actions and respond to the client with the most relevant information to resolve the client's problems.

[0036] Typically, various technologies are used to provide automated solutions to customer queries. In this conventional approach, upon receiving a query from a customer (i.e., a conversational interaction), a work order corresponding to the received query is created. Then, an agent within the existing enterprise resolves the work order based on a predefined set of solution templates.

[0037] However, since the predefined set of solution templates is not an exhaustive list of solutions for all possible customer queries, the responses received by customers may sometimes deviate significantly from their actual expectations. Given this, it is understandable that existing technology cannot understand the context and / or intent of conversational interactions received from customers, and therefore cannot recommend appropriate responses to the agent.

[0038] Therefore, customers suffer a poor experience because they may have to repeatedly reopen their tickets until they get the correct resolution. On the other hand, due to this poor experience, businesses end up hiring a large number of agents to manually read and resolve customer queries, incurring significant costs in recruitment, staffing, and equipment. Furthermore, the manual nature of this process leads to slow query resolution, which in turn negatively impacts the customer experience.

[0039] Therefore, various embodiments of this disclosure aim to address the aforementioned technical problems by providing a method for generating automated solution recommendations for agents engaging in conversational interactions with customers. The method of this disclosure aims to simplify the task for agents to understand the intent of conversational interactions between customers. This is achieved by identifying inherently automatable processes in the query resolution journey and automating these processes. Thus, it can be understood that the method includes determining which processes in the resolution process are automatable. Solution recommendations can then be generated for conversational interactions associated with these automatable processes. Furthermore, the agent can select one of the recommendations and, by making slight modifications, respond to the customer with the same solution. This helps optimize response time for agent-customer interactions.

[0040] Figure 1 This is an example representation of a conversational interaction environment 100 according to various embodiments of the present disclosure. Environment 100 includes a server system 102, multiple agents 104A, 104B, and 104C (collectively referred to as 'agent 104'), and multiple clients 106A, 106B, and 106C (collectively referred to as 'client 106'), each coupled to and communicating with (and / or accessing) a network 108. In a non-limiting example, server system 102 may be configured to perform one or more operations, such as, but not limited to, generating client intent information, detecting the presence of multiple sub-intents in the client intent information, generating one or more conversational interaction response templates as recommendations for agents 104 to use as responses to received conversational interactions, etc.

[0041] In one embodiment, agent 104 may be, but is not limited to, a human agent, chatbot, self-service system (such as a web or mobile digital self-service system), etc., responsible for responding to conversational interactions from client 106. In another embodiment, client 106 may be, but is not limited to, a human user, a representative of an organization / enterprise, chatbot, etc.

[0042] Network 108 may include, but is not limited to, Li-Fi networks, local area networks (LANs), wide area networks (WANs), metropolitan area networks (MANs), satellite networks, the Internet, fiber optic networks, coaxial cable networks, infrared (IR) networks, radio frequency (RF) networks, virtual networks, and / or be able to support Figure 1 Another suitable public and / or private network, or any combination thereof, for communication between two or more of the components or parts shown.

[0043] Various components in environment 100 can be connected to network 108 according to various wired and wireless communication protocols, such as Transmission Control Protocol and Internet Protocol (TCP / IP), User Datagram Protocol (UDP), 2G, 3G, 4G, 5G communication protocols, Long Term Evolution (LTE) communication protocol, Future Communication Protocol, or any combination thereof. For example, network 108 may include multiple different networks, such as private networks accessible to server system 102 and public networks (e.g., the Internet), through which server system 102, any of agent 104, and any of client 106 can communicate.

[0044] Each agent in agent 104 is associated with an electronic device for interacting with existing / potential customers (such as customer 106) located in different geographical locations. Agent 104 is depicted as being associated with various electronic devices (such as desktop computers and mobile devices) in environment 100. However, it should be understood that agent 104 may use any electronic device that can be configured to facilitate interaction between two remote individuals (e.g., individuals such as agent 104A and customer 106A). For example, agent 104B may be equipped with a mobile device, such as a smartphone, configured to facilitate interaction between agent 104B and customer 106B.

[0045] Similarly, each of clients 106 may be associated with one or more electronic devices, such as, but not limited to, mobile devices (e.g., mobile phones, smartphones, tablets, etc.), smart TVs, laptops, and desktop computers, to interact with agent 104. In the example shown, client 106A is depicted as associated with a desktop computer, client 106B is also depicted as associated with a desktop computer, and client 106C is associated with a mobile device. Similarly, agent 104A is depicted as associated with a mobile device, agent 104B is associated with a desktop computer, and agent 104C is also associated with a desktop computer.

[0046] It should be noted that one or more clients in client 106 may be associated with multiple electronic devices and may interact with agent 104 through multiple devices and / or using multiple applications (such as Google Mail®, Outlook Mail®, Apple Business Chat®, Google Business Messaging®, Facebook Messaging®, etc.), which support different text-based interaction channels (such as email, chat, etc.). Furthermore, depending on the client's needs, device-based interactions between multiple clients and multiple agents may be performed concurrently or separately.

[0047] It should be understood that client 106 and agent 104 can communicate via different applications (such as, but not limited to, mobile applications, email applications, web-based email client applications, web-based chat applications, etc.) to achieve various purposes.

[0048] In one example scenario, customer 106A can use a mobile application to initiate a conversational interaction with agent 104A. Customer 106A can initiate the interaction to inquire about invoices, bills of lading (B / L or BL), ocean bills of lading, or similar documents for products or services provided by agent 104A's employer. In another example scenario, agent 104B can initiate a conversational interaction with customer 106B related to a specific product or service as a response to a previous customer inquiry. In this scenario, the conversational interaction might involve making an offer to purchase a product or service from customer 106B, requesting feedback on products customer 106B might purchase, or resolving a customer's inquiry.

[0049] As previously described, after initiating a conversational interaction with Customer 106A, Agent 104A attempts to understand Customer 106A's question or request. Agent 104A may then have to perform a set of tasks to determine a solution to Customer 106A's question. For example, if the customer's inquiry is requesting a Bill of Lading (B / L), Agent 104A must understand that Customer 106A is requesting a B / L for a specific order placed by Customer 106A. Agent 104A must then extract customer parameters, such as customer code, shipping identifier (ID), etc., from the conversational interaction or from different portals associated with the main shipping portal, based on customer details such as email address or customer ID. Agent 104A can then access specific portals, such as the Global Customer Service System (GCSS) portal, finance-related portals, etc., to locate or generate the B / L for the customer. In one instance, the B / L may have been previously generated and stored by Customer 106A at the time of purchase, and Agent 104A can now retrieve that file from the appropriate platform using the customer parameters. Returning to the previous example, after locating the B / L document, Agent 104A can verify it and share it with Customer 106A via email. As can be understood, the process Agent 104A undertakes to fulfill the customer's request is cumbersome, time-consuming, and resource-intensive. Furthermore, because such requests are quite common, Agent 104A may have to perform the same process multiple times throughout the day, resulting in inefficient use of agent time. In addition, because many Agents 104A may have to log into such portals and spend time on them throughout the day, the portals themselves become overloaded, leading to higher processing and storage requirements, which in turn results in higher infrastructure and power costs.

[0050] To overcome this problem, a method is provided for generating solution recommendations for agent 104A in response to a conversational interaction from client 106A according to an automatable process. To this end, this disclosure describes a server system 102 that performs multiple operations to facilitate the generation of solution recommendations for agent 104.

[0051] In one embodiment, environment 100 may further include a database 110 coupled to server system 102. In the example, server system 102 coupled to database 110 is embodied in a central server (not shown) associated with the employer of agent 104; however, in other examples, server system 102 may be a separate component (acting as a hub) connected to the central server. Database 110 may be incorporated into server system 102, or it may be an individual component connected to server system 102, or it may be a database stored in a cloud storage device. In one embodiment, database 110 may store historical conversational interaction datasets, one or more machine learning (ML) models, and other necessary machine instructions required to implement various functions of server system 102, such as firmware data, operating system, etc. In an embodiment, one or more ML models may include a first ML model and one or more second ML models. Note that the historical conversational interaction dataset, the first ML model, and one or more second ML models will be explained in detail later in this disclosure. Additionally, database 110 provides a storage location for data and / or metadata obtained from various operations performed by server system 102.

[0052] In one embodiment, when a customer (such as customer 106A) initiates a conversational interaction with an agent (such as agent 104A), server system 102 is configured to also receive the conversational interaction. In some examples, the conversational interaction may be an email, chat message, etc. In an instance, the conversational interaction may include a request from customer 106A to resolve a query from customer 106A. In response to this conversational interaction, server system 102 is configured to process one or more relevant texts from the conversational interaction using a first ML model to identify one or more customer intents associated with the conversational interaction. This step results in customer intent information being further processed using one or more second ML models. This step results in the conversational interaction being categorized into at least one process category. Herein, process categories are associated with a set of tasks that should be performed to determine a solution for customer 106A's query in the received conversational interaction.

[0053] In one embodiment, server system 102 may also be configured to identify whether the process category assigned to the conversational interaction is automatable. Herein, if the process category is automatable, server system 102 may automatically implement the set of tasks that can be associated with the corresponding process category. Alternatively, if the conversational interaction requires human intervention from agent 104A and is not fully automatable, it may be discarded without further processing.

[0054] In some instances, a conversational interaction may be associated with multiple sub-intents. For example, customer 106A may request agent 104A to check if an invoice has been generated for a specific shipment, and may also request agent 104A to provide a bill of lading. In this document, multiple sub-intents correspond to both the status of requesting an invoice and the bill of lading. Therefore, server system 102 can be configured to use one or more second ML models to detect the presence of one or more sub-intents in the customer intent information. In one scenario, after detecting one of the one or more sub-intents, server system 102 can be configured to assign a specific automatable process to the conversational interaction based on a comparison with a predefined set of automatable processes. Later, server system 102 can extract multiple related entities from the conversational interaction based on one or more related texts associated with it. In some embodiments, some of the multiple entities may correspond to entities indicating important identifiers, names (such as the sender's name), etc.

[0055] In some other embodiments, server system 102 may be configured to access a list of tasks associated with a specific automatable process from database 110. Further, server system 102 may be configured to generate query resolution information for conversational interactions based at least in part on customer intent information processed against the task list. In this document, in embodiments, the task list may correspond to tasks that should be performed to determine a query resolution for a query received in a conversational interaction. Therefore, processing customer intent information against the task list may include facilitating the implementation of one or more tasks from the task list. In another embodiment, server system 102 may be configured to generate one or more conversational interaction response templates for conversational interactions based at least in part on multiple related entities and query resolution information.

[0056] In a particular embodiment, a conversational response template can be provided to agent 104A as a solution recommendation, so that in response to a conversational interaction received from client 106A, agent 104A can select the most appropriate template and send a reply to client 106A. For example, in response to receiving a query email, a template email describing a solution for the query in the received query email is provided to agent 104A as a recommendation. Reference will be made later. Figure 2 Provides a detailed explanation of the various operations required to generate a solution recommendation via server system 102.

[0057] Despite Figure 1 In this document, server system 102 is shown as being incorporated into environment 100; however, in some embodiments, server system 102 may be located outside environment 100 and communicate with the environment, for example, via network 108. In some examples, server system 102 may be implemented in a third-party external server to perform the various operations described herein.

[0058] Figure 1 The number and arrangement of systems, devices, and / or networks shown are provided as examples. Figure 1 Compared to those shown, there may be additional systems, devices, and / or networks; fewer systems, devices, and / or networks; different systems, devices, and / or networks; and / or systems, devices, and / or networks with different arrangements. Furthermore, Figure 1 The two or more systems or devices shown can be implemented within a single system or device, or Figure 1 The single system or apparatus shown may be implemented as multiple distributed systems or apparatuses. Furthermore, server system 102 should be understood as being embodied in at least one computing device communicating with network 108, the at least one computing device being specifically configured via executable instructions to perform the steps described herein, and / or embodied in at least one non-transitory computer-readable medium.

[0059] Figure 2 A simplified block diagram of a server system 200 according to an embodiment of the present disclosure is shown. Note that the server system 200 can be coupled with… Figure 1 The server system 200 is similar to server system 102. In one embodiment, server system 200 is part of an internal server operated by an organization employing users such as agent 104 (1). In some embodiments, server system 200 is embodied as cloud-based and / or software-as-a-service (SaaS) based architecture.

[0060] Server system 200 includes computer system 202 and database 204. Note that database 204 is related to... Figure 1 The database 110 is the same. The computer system 202 includes at least one processor 206 (which may be interchangeably referred to herein as 'processor 206') for executing instructions, memory 208, communication interface 210, user interface 212 and storage interface 214, which communicate with each other via bus 216.

[0061] In some embodiments, database 204 is integrated into computer system 202. For example, computer system 202 may include one or more hard drives such as database 204. User interface 212 is an interface, such as a human-machine interface (HMI) or software application, that allows a user (such as an administrator) to interact with and control server system 200 or one or more parameters associated with server system 200. It may be noted that user interface 212 may consist of several components that vary depending on the complexity and purpose of the application. Examples of components of user interface 212 may include visual elements, controls, navigation, accessibility features, etc.

[0062] Storage interface 214 is any component that provides processor 206 with access to database 204. Storage interface 214 may include, for example, an Advanced Technology Attachment (ATA) adapter, a Serial ATA (SATA) adapter, a Small Computer System Interface (SCSI) adapter, a RAID controller, a SAN adapter, a network adapter, and / or any component that provides processor 206 with access to database 204.

[0063] In a non-limiting example, database 204 is configured to store customer intent information 218, a first ML model 220, and one or more second ML models 222, etc. In another non-limiting example, database 204 is also configured to store a historical conversational interaction dataset, which can be used to train one or more ML models, such as the first ML model 220 and one or more second ML models 222. In one embodiment, the historical conversational interaction dataset may include information related to multiple conversational interactions between multiple customers and multiple agents. Examples of conversational interactions include email conversations, chat conversations, transcribed verbal communications, etc.

[0064] Processor 206 includes suitable logic, circuitry, and / or interfaces to perform operations such as receiving conversational interactions, generating one or more conversational interaction response templates, etc. Examples of processor 206 include, but are not limited to, application-specific integrated circuit (ASIC) processors, reduced instruction set computing (RISC) processors, graphics processing units (GPUs), complex instruction set computing (CISC) processors, field-programmable gate arrays (FPGAs), etc.

[0065] Memory 208 includes suitable logic, circuitry, and / or interfaces for storing a set of computer-readable instructions for performing the various operations described herein. Examples of memory 208 include random access memory (RAM), read-only memory (ROM), removable storage drives, hard disk drives (HDDs), etc. It will be apparent to those skilled in the art that the scope of this disclosure is not limited to implementing memory 208 in the server system 200 described herein. In another embodiment, memory 208 may be implemented as a database server or cloud storage device that works in conjunction with server system 200 without departing from the scope of this disclosure.

[0066] Processor 206 is operatively coupled to communication interface 210, enabling processor 206 to communicate with remote devices such as third-party servers (i.e., with remote device 224), or with electronic devices associated with agent 104 and client 106, or with any entity connected to network 108 (such as...). Figure 1 (As shown).

[0067] It should be noted that the server system 200 illustrated in the figures and described below merely illustrates devices that may benefit from embodiments of this disclosure, and therefore should not be construed as limiting the scope of this disclosure. Note that the server system 200 may include, but is not limited to, devices that can benefit from embodiments of this disclosure. Figure 2 The components described in the text are fewer or more components.

[0068] In one implementation, processor 206 includes a data receiving module 226, a data preprocessing module 228, an entity extraction module 230, an automationability determination module 232, and a solution determination module 234. It should be noted that the components described herein (such as data receiving module 226, data preprocessing module 228, entity extraction module 230, automationability determination module 232, and solution determination module 234) can be configured in various ways, including electronic circuit systems, digital arithmetic and logic blocks, and memory systems combined with software, firmware, and embedded technologies. In various non-limiting examples, data receiving module 226, data preprocessing module 228, entity extraction module 230, automationability determination module 232, and solution determination module 234 can be communicatively coupled to each other and can transfer data between modules.

[0069] In this embodiment, the data receiving module 226 includes suitable logic and / or interfaces for receiving conversational interactions initiated by a client (e.g., client 106A). The received conversational interactions can be provided to the data preprocessing module 228 for further processing.

[0070] In an embodiment, the data preprocessing module 228 includes suitable logic and / or interfaces for extracting a set of relevant segments from the conversational interaction. In a non-limiting implementation, after receiving the conversational interaction from the client 106A, the conversational interaction is preprocessed to identify the client intent associated with the conversational interaction. Preprocessing of the conversational interaction includes extracting a set of relevant segments from the conversational interaction. Herein, a set of relevant segments refers to one or more segments in the conversational interaction that may be sufficient to identify the client's intent or are important for identifying the client's intent. For example, when the conversational interaction is an email, a set of relevant segments including the email subject line and the email body may be sufficient to identify the client's intent because, as is known from a predefined email format, these segments should include the email sender's actual intent.

[0071] In a non-restrictive implementation, a simple rule-based approach or regular expression (Regex) technique can be used to extract relevant segment sets from conversational interactions. When using such an approach, a user of the server system 200, such as an administrator (not shown), can define simple rules associated with specific text patterns. Based on the rules defined for specific patterns, the text in the conversational interaction can be processed to identify matching patterns. For example, to identify the subject line in an email, a predefined subject line pattern can be used, such as "Subject:" followed by "text". The data preprocessing module 228 can then be configured to process the conversational interaction to identify the predefined pattern used for the subject line. After identifying the predefined subject line pattern, the data preprocessing module 228 can extract the subject line from the email. Similarly, the email body can be extracted from the email. In some embodiments, when the text may be complex, Natural Language Processing (NLP) techniques can be used instead of rule-based methods and Regex techniques. In some other embodiments, Regex can be used in combination with NLP techniques. It is important to note that NLP techniques are powerful for understanding the semantic meaning of text, while Regex is typically used for pattern-based matching. When using NLP techniques, a pre-trained NLP-based model or a predefined language-based library can be used to extract relevant text from the conversational interaction.

[0072] After extracting the relevant segment set, in another embodiment, the data preprocessing module 228 can be configured to generate first relevant text based at least in part on applying a first set of preprocessing operations to the relevant segment set. In an example scenario, when a query is received via email, the subject line and email body are the most relevant segments of the email. Further, the subject line and email body can be concatenated to identify the customer's intent from the email. Therefore, in one embodiment, the first set of preprocessing operations may include concatenation operations to be performed on the relevant segment set. As used herein, the term 'concatenation operation' refers to the operation of joining multiple strings of characters into a single string. In this document, characters may include alphanumeric characters, numbers, letters, etc. In a non-limiting example, the concatenation operation may be implemented in a preferred preprocessing platform using concatenation operators such as + and &.

[0073] In the embodiments, it may be noted that the methods and systems proposed in this disclosure can be used in any field or industry. However, for ease of explanation, analysis, and performance comparison, various embodiments of the proposed systems are applied to the freight and logistics industry. However, this should not be construed as limiting, and various embodiments of this disclosure are applicable to a variety of other industries such as healthcare, retail, media, tourism, criminal investigation, etc., and these will also be covered within the scope of this disclosure.

[0074] In specific scenarios, such as email, an additional set of rules can be defined. Emails that meet the following two criteria can be further delivered to identify customer intent: i. Subtype = 'Release Shipping Instructions (SI) / BL / Voyage Charter (VC)' ii. Reason for the case = 'B / L Enquiry' or 'Release B / L Web Request' In some embodiments, the first relevant text may be further preprocessed to identify a predefined set of keywords. For example, in a scenario where a customer and an agent are communicating about freight, the predefined set of keywords may include 'swb', 'seaway', 'bl', 'obl', 'mbl', 'bill', 'lading', 'waybill', 'bls', 'swbs', etc. One of the rules may also be: case sensitivity is not applied.

[0075] In some other embodiments, the data preprocessing module 228 may be further configured to filter and discard one or more conversational interactions that are not automatable during the preprocessing stage, based at least on the rules described above. For example, if at least the rules described above do not meet the criteria, the data preprocessing module 228 may exclude conversational interactions such as emails from automation considerations. Alternatively, if an email mentions a participant name or service charge (SCV) code that is different from the account code of customer 106A who made the request (i.e., sent an email with a query) to a customer service team (such as an Salesforce team), the data preprocessing module 228 may exclude the conversational interaction from automation considerations. After preprocessing and confirming that the conversational interaction meets the rules described above, the first relevant text can be provided to the entity extraction module 230 and the automatability determination module 232 for further processing by the automatability determination module 232.

[0076] In an embodiment, the automatability determination module 232 includes suitable logic and / or interfaces for generating customer intent information 218 for a conversational interaction, at least in part, based on a first relevant text and a historical conversational interaction dataset. The automatability determination module 232 can use various AI / ML models to generate the customer intent information 218. In a particular embodiment, the automatability determination module 232 can be configured to generate the customer intent information 218 using a first ML model 220. In an example, the first ML model 220 may be a large language model (LLM), such as a Transformer-based bidirectional encoder representation (BERT) model, a generative pre-trained Transformer (GPT) model, etc.

[0077] As can be understood, LLMs are typically trained using diverse and extensive datasets for language understanding. This dataset covers a wide range of topics, styles, and languages ​​to ensure the model's generalization ability. However, when an LLM is used to identify intent from text strings, it can be trained using a labeled dataset containing examples of text strings and their corresponding labels. In this document, the labeled dataset should also be diverse and cover a wide range of intents the model is expected to recognize. Therefore, the automatability determination module 232 can be configured to train a first ML model 220 to determine customer intents associated with conversational interactions based on a first relevant text and historical conversational interaction dataset. In embodiments, when considering the use of a labeled dataset to train the first ML model 220, supervised training / learning-based techniques may be preferred for training the first ML model 220. Furthermore, it may be noted that training techniques such as those for ML models are well-known in the art. Therefore, the training, validation, and testing phases for the first ML model 220 can be implemented in a manner similar to how they can be performed for any AI or ML model, and thus, for the sake of brevity, these phases will not be described further herein.

[0078] Once the first ML model 220 is trained for intent recognition, it can be used to identify customer intents associated with received conversational interactions. Furthermore, the first ML model 220 can generate corresponding customer intent information 218. As previously described, the customer intent information 218 is stored in a database 204, which can be accessed in the future for further processing or any other application.

[0079] In one embodiment, customer intent information 218 may include information related to a customer's intent. In an example scenario, the information may include a textual description of the customer's intent. It is important to note that the conversational interaction may be associated with multiple customers. Furthermore, in some embodiments, the first ML model 220 may detect multiple customer intents in the conversational interaction. Therefore, customer intent information may include information related to multiple customer intents.

[0080] In some additional embodiments, the customer intent information 218 can be further refined by training the first ML model 220 to utilize one or more intent-related rule sets, such as a first predefined rule set. Therefore, the automatability determination module 232 can be further configured to fine-tune the customer intent information 218 at least in part based on the first relevant text and the first predefined rule set. In one embodiment, the automatability determination module 232 can use the first ML model 220 to fine-tune the customer intent information 218. In a non-limiting implementation, in addition to LLM, an NLP model can also be used in collaboration with LLM as the first ML model 220 to fine-tune the customer intent information 218 for better performance.

[0081] For example, when a query is received from customer 106A as an email address, the first ML model 220 can generate a response to the following question as part of a first predefined rule set: i. Does client 106A have problems accessing a requested document in the query on the web? [Yes / No] ii. Is Client 106A requesting any updates to be shown on the bill? [Yes / No] iii. Is it a request to deliver the bill physically or online? More specifically, if the response to email delivery from the first ML model 220 is 'physically', then since the 'releasing on destination' operation is not automatable, the corresponding email may be excluded from consideration for automation. Furthermore, based on the response from the first ML model 220, web error and correction emails are also excluded.

[0082] Furthermore, conversational interactions falling within the scope of automation can be categorized into one or more automatable process categories. Therefore, after identifying the intent, i.e., the customer intent information 218, multiple actions can be performed on the intent to identify the correct scope for automation. In one embodiment, the automatability determination module 232 can be configured to perform a set of actions using one or more second ML models 222. In the example, the one or more second ML models 222 can be rule-based models, NLP-based models, Regex, any other AI or ML models, and / or combinations thereof.

[0083] In a specific, unrestricted implementation, the automatability determination module 232 can train one or more second ML models 222 based on at least the following rules: i. If the intent determines that customer 106A wants a file 'through email', 'via email', or 'on email', then exclude emails, as these cannot be considered under automated consideration.

[0084] ii. If the intention is to indicate that a specific bill is needed instead of another bill, the email will be removed due to a 'Bill TypeChange Request'.

[0085] iii. Regarding topic mining, it may be noted that the customer intent identified using the first ML model 220 is rule-based for process classification. If the topic / process category cannot be identified from the first ML model 220, another ML model, such as a KeyBert-based model, can be used. In this embodiment, a KeyBert-based model can be used to extract key phrases from emails received by customer 106A. Furthermore, the KeyBert-based model can pass these key phrases as part of the aforementioned keywords and perform multiple validations to ultimately identify the topic (or process category).

[0086] In one embodiment, conversational interactions associated with multiple customer intents may also have to be discarded from automation considerations. Therefore, the automatability determination module 232 can be configured to detect the presence of one or more sub-intents in the customer intent information 218.

[0087] In one scenario, detecting the presence of one or more sub-intents in customer intent information 218 may include detecting two or more sub-intents in customer intent information 218 based at least in part on the recognition of one or more conjunction texts in customer intent information 218. For example, the query is received via email, and customer intent information 218 for email includes one or more conjunction texts, such as 'AND'. It can then be understood that customer intent information 218 includes multiple customer intents. Therefore, the automatability determination module 232 can assign non-automatable processes to conversational interactions based at least on the association of customer intent information 218 with at least two or more sub-intents. After classifying a conversational interaction as a non-automatable process, it can be discarded without further processing.

[0088] In another scenario, detecting the presence of one or more sub-intents in customer intent information 218 may include detecting a single sub-intent. After detecting that only one sub-intent is available in customer intent information 218, the automatability determination module 232 may be configured to assign a specific automatable process to the conversational interaction, at least in part, based on comparing customer intent information 218 with a predefined set of automatable processes. Further, the predefined set of automatable processes may be based on the rules described above for process classification. In this document, the term 'automatable process' refers to a process whose steps can be automatically executed by server system 200 without any intervention from agent 104A. These processes may refer to procedures or tasks that should be performed to determine the resolution of queries received from conversational interactions from customer 106A.

[0089] In one embodiment, the entity extraction module 230 includes suitable logic and / or interfaces for extracting multiple related entities from a conversational interaction, at least in part, based on one or more related texts associated with the conversational interaction. In a non-limiting example, the one or more related texts may include first related text, second related text, third related text, and so on.

[0090] In a particular embodiment, extracting the plurality of related entities may include extracting a first set of related entities from the plurality of related entities. In a scenario where freight is being executed between the employers of customer 106A and agent 104A, the first set of related entities may include freight, number, customer number, document number, etc. It is understood that the first set of related entities may correspond to entities used as identifiers for products or services and is a necessary element that agent 104 or server system 200 can use to determine a solution to a query in a conversational interaction from customer 106A. After determining a solution for a query received in an email, these entities may need to be mentioned in a resolution email that agent 104A may send to customer 106A.

[0091] In an embodiment, to extract the first relevant entity set, the entity extraction module 230 may be configured to generate the second relevant text, at least in part, based on applying a second set of preprocessing operations to the first relevant text. It is important to note that to extract the first relevant entity set from the first relevant text, unwanted noise may need to be removed. For example, in the case of the freight industry, parameters such as freight numbers, customer IDs, and other similar parameters may appear in the subject line or email body but not in any other section of the email. Furthermore, it should be noted that when processing conversational interactions, the aforementioned parameters can be considered entities, and AI-based techniques can be considered for extracting such entities. Therefore, in a non-limiting example, the second set of preprocessing operations may include removing noise from the first relevant text, such as salutations, disclaimers, email traces, stop words, punctuation marks, any email address, URLs, numbers, etc. After removing unwanted text from the first relevant text, the second relevant text can be obtained.

[0092] In another embodiment, entity extraction module 230 may be configured to process second related text to generate a first related entity set, at least in part, based on the second related text and a second predefined rule set. In a non-limiting example, entity extraction module 230 may use one or more second ML models 222 to process the second related text. As will be understood, the one or more second ML models 222 may be any type of rule-based model that defines rules for patterns associated with the text to be extracted. Therefore, the second predefined rule set may include rules defining patterns of text to be classified as second related text. For example, freight numbers typically include alphanumeric characters, which may include numbers, letters, and sometimes special characters. Similarly, in another example, customer codes may also include alphanumeric characters. Furthermore, such entities have the characters arranged in a predefined order. Therefore, in a non-limiting scenario, the second predefined rule set may include rules specifying that if a series of alphanumeric characters is detected, it contains at least two letters followed by a predefined count of random numbers. The corresponding entities can then be classified into the first related entity set.

[0093] In another specific embodiment, extracting multiple relevant entities from a conversational interaction may include extracting a second set of relevant entities from among those entities. In a scenario where freight is being executed between the employers of customer 106A and agent 104A, the second set of relevant entities might include the sender's name, the customer's identifier, etc. It is understood that the second set of relevant entities may correspond to entities that need to be mentioned in the salutation section of a resolution email.

[0094] In an embodiment, to extract the second set of relevant entities, the entity extraction module 230 may be configured to generate the third relevant text, at least in part, based on applying a third set of preprocessing operations to the first relevant text. In a non-limiting example, the third set of preprocessing operations may include removing unwanted noise from the first relevant text. For example, in the case of freight, this step may be performed to obtain the sender's name and determine that the email signature section may be needed for further processing. Therefore, in such a scenario, the relevant text may include the email signature section.

[0095] Furthermore, in another embodiment, the entity extraction module 230 may be configured to process the first relevant text to generate an intermediate entity set, at least in part, based on a third predefined rule set. In a non-limiting example, the entity extraction module 230 may use one or more second ML models 222 (such as Named Entity Recognition (NER) models) to generate the intermediate entity set. Generally, the NER model can be used for scenarios where entities correspond to names of people, places, organizations, dates, etc., according to the third predefined rule set. Thus, it may be noted that the NER model can be trained to identify and classify such entities from text strings. In one embodiment, the intermediate entity set may include a list of names that are proper nouns, i.e., all names in an email that are names of people, places, organizations, things, etc.

[0096] In yet another embodiment, entity extraction module 230 can be configured to compare third relevant text with an intermediate entity set to generate a second relevant entity set. After comparison, one or more entities in the intermediate entity set that match entities in the third relevant text can be considered part of the second relevant entity set. For example, the third relevant text could be the signature section of an email, and the intermediate entity set could include all proper nouns in the email. Proper nouns could include names in the address section, email body, and other traces of the email. After comparing the third relevant text with the intermediate entity set, names in the intermediate entity set that match names in the signature section are the sender's names. Therefore, the matching names can be used to resolve emails that should be shared with client 106A as a response to a conversational interaction with client 106A.

[0097] After extracting multiple relevant entities from the conversational interaction, the data receiving module 226 can be configured to access a list of tasks associated with a specific automatable process from the database 204. As previously mentioned, each process may be associated with a set of tasks performed to determine a solution for a query received as a conversational interaction for customer 106A. This set of tasks may be predefined, and information associated with them may be stored in the database 204. Therefore, this information may need to be extracted from the database 204. Furthermore, the server system 200 may need to execute the set of tasks to obtain the solution. The data receiving module 226 may share the information associated with the task set with the solution determination module 234 for further processing.

[0098] In one embodiment, the solution determination module 234 includes appropriate logic and / or interfaces for generating query resolution information for conversational interaction, based at least in part on customer intent information processed against a task list. It should be noted that the task list may include tasks specific to a particular process category. For example, in the transportation and logistics industry, a task set might include checking which specific portal must be accessed to check the status of a requested document, checking if the requested document has been generated, checking if it has been internally verified, identifying the freight operator, and so on. In one embodiment, to process customer intent information against a task list, the solution determination module 234 may be further configured to execute each task from the task list to generate query resolution information for conversational interaction.

[0099] Once all tasks have been executed and the output of each executed task is positive, a relevant solution including a positive response can be generated. Alternatively, if the result of executing any task in the set is negative, a solution including a negative response can be generated. For example, if the request is to generate an invoice. In this scenario, if the invoice has already been generated, the solution for the query might be 'Yes, the invoice is generated and shared across'. On the other hand, if the invoice has not been generated, the solution determination module 234 can check the possibility of generating the invoice. If it is not possible, a message can be sent to customer 106A as a response to the query, containing: 'No, the invoice is not generated'.

[0100] In another embodiment, the resolution determination module 234 may be further configured to generate one or more conversational interaction response templates for the conversational interaction, at least in part based on multiple relevant entities and query resolution information. Hereinafter, one or more conversational interaction response templates may be generated for agent 104A as recommendations. Agent 104A may select the most appropriate template from the one or more conversational interaction response templates and share it with client 106A as a response to the conversational interaction received from client 106A.

[0101] Figure 3This is a flowchart 300 depicting, according to embodiments of the present disclosure, the process of generating an AI solution template for an agent (e.g., agent 104A) in response to a query email from a customer (e.g., customer 106A). In an example scenario, customer 106A can access a freight assistance platform on their mobile device. After accessing the platform, customer 106A places an order to transport a shipping container filled with various goods from a first location to a second location. In this scenario, it is assumed that the shipping container must cross an international border before reaching its desired destination. Therefore, it can be transported by a shipping company via cargo ship.

[0102] When placing a freight order through an online platform associated with a shipping company (i.e., a freight assistance platform), customer 106A may be presented with several options. These options are displayed to the customer via a GUI on the freight assistance platform installed on the customer's electronic device. These options may relate to setting payment methods, setting payment times, filling in freight details, providing freight instructions, editing personal details, etc. Suppose customer 106A chooses to pay after the product arrives at its destination. Once this occurs, customer 106A can use login credentials on the freight assistance platform or a personal email account (such as Google...) ® Yahoo ® Hotmail ® Send emails (etc.).

[0103] On the other hand, the transportation company may have hired a team of agents, such as Agent 104, to handle inquiries from Customer 106. Therefore, Agent 104 may also register with administrative privileges on a freight assistance platform. This could allow Agent 104 to access inquiries received from Customer 106 on an electronic device specifically associated with Agent 104 via administrator login credentials. Agent 104A might receive emails from Customer 106A and thus be responsible for resolving and closing the issues faced by Customer 106A.

[0104] As will be understood, the container shipping process is associated with the generation and management of one or more documents that can be recorded as evidence and used in the future. In one embodiment, the one or more documents may include commercial invoices, bills of lading (B / L), import licenses, arrival notices, letters of credit, etc., for imported products. In another embodiment, the one or more documents may further include B / Ls, packing lists, commercial invoices, letters of credit, certificates of origin, declarations, certificates, etc., for exported products.

[0105] It should be noted that at the end of agent 104, the freight assistance platform can be communicatively coupled with server system 200. Therefore, features facilitated by server system 200 may be available to benefit agent 104. Thus, when customer 106A sends a query email to agent 104A, server system 200 may also receive that query email. The steps implemented by server system 200 are as follows: Figure 3 Described in the text.

[0106] Upon receiving a query email, server system 200 extracts a set of relevant segments (such as 'subject' and 'email body') from the query email and concatenates them. After concatenation, server system 200 can generate first relevant text, which is the result of the concatenation. The result of the concatenation may include the subject + email body (see 302) as input to a first ML model 220 (such as LLM 304). For example, query email could be "SUBJECT OF EMAIL: REQ FOR SEAWAY BL / BL NO: 12*****89 / 22*****23 / 13*****23 / 41*****34. EMAIL FROM CUSTOMER: DEAR ABCDEF TEAM, ATTACHED PAYMENTRECEIPT FOR THE SUBJECT SHIPMENT. KINDLY POST THE “SEAWAY BL" AT EARLIEST.THANKS®ARDS, AAAAA.C, SR. EXECUTIVE – DOCUMENTATION, ABCABC LOGISTICS PVTLTD, NO.123, T2&T3, THIRD FLOOR, 0**-*****67 (Email subject: Request for Ocean Freight BL / BL NO: 12*****89 / 22*****23 / 13*****23 / 41*****34. Email from customer: Dear ABCDEF team, Attached is the payment receipt for this shipment. Please issue the "Ocean Freight BL" as soon as possible. Thank you and best regards, AAAAA.C, Senior Executive – Documentation, ABCABC Logistics Pte Ltd, 123, T2 & T3, 3rd Floor, XYZ Shopping Centre, DEFDEF Road, XXXXX, YYYYYY-123456, Mobile: +91 98******10 / Telephone: 0**-*****67)

[0107] When agent 104A receives this email, server system 200 accesses it as input. Further, first relevant text 302 is provided to LLM 304 for performing intent identification (see 306) and process classification (see 308). The intent identification 306 results in providing the customer intent associated with the query email in the form of customer intent information. Furthermore, the customer intent information is fine-tuned based on a first predefined set of rules, where LLM 304 is trained to answer the previously mentioned set of questions. After performing intent identification 306, the intent identified for the received query email could be “REQUEST TO POST SEAWAY BL FOR THE MENTIONED SHIPMENT”. Further, in a non-limiting implementation, the output of LLM 304 might resemble... Figure 4A The table shown is illustrated (see 402). It should be noted that... Figure 4A The table 402 shown may be the output of LLM 304 displayed on the UI of any electronic device of at least one of Customer 106A and Agent 104A.

[0108] refer to Figure 4A (See 400) It can be observed that the first column 404 in Table 402 shows the query email under the title 'combined text', and the second column 406 shows the customer intent under the title 'Intent'. Further, the third, fourth, and fifth columns (see 408, 410, and 412 respectively) show the answers generated by LLM 304 for the questions 'web access issue?', 'amendment query?', and 'require physical query?'. It can be observed that the answer to each of these questions is 'NO'. Therefore, after identifying the customer intent and obtaining a 'NO' answer to the question, the process moves to the step of process classification 308.

[0109] Reference Figure 3In the process classification step 308, the server system 200 may further process the first relevant text 302 to determine whether the customer intent associated with the first relevant text 302 is classifiable as an automatable process (see 310). The server system 200 may determine the automatability associated with the first relevant text 302 based on a comparison with a predefined set of automatable processes and a predefined set of rules explained earlier in this disclosure. Furthermore, in another embodiment, the server system 200 may use one or more second ML models 222 to perform process classification 308.

[0110] In some embodiments, server system 200 may train one or more second ML models 222 based on a rule set (which may also include a keyword set). The rule set may be as follows: i. If keywords from the first two lists below appear in the same sentence in the customer's intent, a process category such as 'Release SWB' can be assigned to it, provided that no words from 'excluded_list_swb' appear: 1. identification_keys_swb_1 = ['swb', 'seaway', 'seawaybill', 'seawaybl', 'waybill', 'swbbl', 'express_bl', 'swbs', 'get_swb', 'original_swb'].

[0111] 2. identification_keys_swb_2 = ['release', 'provide', 'post', 'upload', 'arrange', 'publish', 'issue', 'request', 'get_swb'].

[0112] 3. excluded_list_swb = ['send', 'edo', 'do', 'freight', 'rated', 'unrated', 'amend', 'revise', 'cdo', 'rates', 'obl', 'mbl', 'vc', 'verify_copy', 'draft', 'charge', 'not_issue', 'update', 'change', 'non-rated', 'not_issue' 'withdraw', 'print', 'destination', 'collection', 'of_issue', 'status', 'confirm', 'copy', 'counter', 'collect', 'unfreighted', 'non_rated','rectify', 'consignee', 'reupload', 're-post', 'original'].

[0113] ii. If keywords from the first two lists below appear in the same sentence in the customer intent, a process category such as 'Release OBL\MBL' can be assigned to it, provided that no words from excluded_list_obl appear: 1. identification_keys_obl_1 = ['bl', 'obl', 'mbl', 'bol', 'bill', 'bills', 'bls', 'get_ob'].

[0114] 2. identification_keys_obl_2 = ['release', 'provide', 'post', 'upload', 'arrange', 'publish', 'issue', 'request', 'final', 'get_obl'].

[0115] 3. excluded_list_obl = ['send', 'destination', 'rated', 'unrated', 'edo', 'do', 'amend', 'revise', 'print', 'cdo', 'charge', 'not_issue', 'freight', 'unfreighted', 'change', 'amendment', 'print', 'update', 'collect','split', 'pod', 'surrender', 'freight', 'address', 'instead_of', 'approve', 'switch_bl', 'courier', 'counter', 'withdraw', 'pick_up', 'collection', 'of_issue', 'status', 'confirm', 'copy', 'collect', 'slave', 'confirmation', 'overdue', 'transport_bl', 'unfreighted', 'non_rated', 'rectify', 'consignee','reupload', 're-post'].

[0116] In one embodiment, after process classification 308, if the query email is classified as non-automateable, the case can be removed (see 312). Alternatively, if the query email is classified as automatic, the process moves to the step of extracting multiple entities (see 314). The output of performing the step of process classification 308 may be similar to... Figure 4A The table shown is shown (see, 414).

[0117] Refer again Figure 4AIn a non-restrictive example, Table 414 shows two different intents assigned to certain automatable process categories. One customer intent includes 'THE CUSTOMER IS REQUESTING THE UPLOAD OFFINAL SWB FOR BOOKING #11*****12', as shown in column 416 under the heading "Combined Text," and is assigned the process category 'RELEASE SEAWAY BILL', as shown in column 418. This classification can be performed based on the rules described above. In this document, the term 'combined text' is used to refer to the customer intent. Similarly, another example of a customer intent is 'THE CUSTOMER INTENT IS TO RELEASE THE OBL ONAN URGENT BASIS', assigned the process category 'RELEASE OBL\MBL'.

[0118] Further, refer to the return Figure 3 As can be understood, multiple entities include first relevant text 302, second relevant text, and third relevant text. In this document, the second relevant text can be obtained after performing a preprocessing step (see, 316) on the first relevant text 302. Similarly, the third relevant text can be obtained after performing a preprocessing operation (such as using a NER system (see, 318)). In an embodiment, NER system 318 can preprocess the first relevant text 302 using a NER model. The output of NER system 318 is the sender's name (see, 320).

[0119] Furthermore, after identifying the process category, the entity to be extracted, and the name of the sender to be extracted, the server system 200 can execute the corresponding task set. These tasks can be executed to obtain a solution for querying the email. In a non-restrictive example of the GCSS portal (see, 322), for the process category 'RELEASE SEAWAY BILL', the task set may include 'TASK 1', 'TASK 2', 'TASK 3', 'TASK 4', and 'TASK 5', as follows. Figure 3 As shown. Each task may need to be performed and checked to see if it meets the customer's requirements in accordance with the customer's intent.

[0120] More specifically, the task set might look like this: i. As mentioned above, GCSS automation involves a total of 5 steps. Before implementing the tasks described earlier, it may also be necessary to identify the freight operator, as it is used as the carrier in the GCSS API call URL. In a non-restrictive example, the mapping is as follows: 1. If the operator:carrier in ['MSL', 'MAEU'] is 'MAEU' 2. In elif ['SGL', 'SEJJ'], the operator: carrier = 'SEJJ' 3. In elif ['SEA', 'SEAU'], the operator: carrier = 'SEAU' 4. In elif ['MCC', 'MCPU'], the operator: carrier = 'MCPU' 5. In elif ['HSU', 'SUDU'], the operator: carrier = 'SUDU' 6. In elif ['MLL', 'MAEI'], the operator: carrier = 'MAEI' ii. Task 1: Check whether the vessel has departed from port, and this can be identified based on a specific OPS task, which requires identifying whether the pick-up city is 'OUT (departure)'.

[0121] iii. Identify whether the document type and the subject matter are the same. For example, for the SWB document type, it should be 'TPDocSea Waybill Shipped', and for OBL / MBL, it should be 'TPDocNegotiable Shipped'.

[0122] iv. Task 3: Check if the Approval for Release (AFR) document task is closed.

[0123] v. Task 4: Check if the issuance of original transport documents (IOTD) is closed.

[0124] vi. Task 5: Match the account that sends emails to the transport service provider with its SCV code to see if it is marked as a Transport Document Participant (TDR Participant) in the GCSS Participant task. If any party other than the account that sent the email is marked as a TDR, the check is considered a failure.

[0125] vii. Once all tasks are completed and all checks pass, the server system 200 can then generate a response email notifying the client 106A that the requested file has been uploaded to the website.

[0126] viii. In the case of multiple shipments, the consolidated status of all shipments can be provided in a single email.

[0127] In one embodiment, the server system 200 may be further configured to generate one or more conversational interactive response templates, which in Figure 3 The process is indicated by step 324. This one or more conversational interactive response templates may be referred to as 'AI templates', 'AI solution templates', or simply 'templates' (see 326). To generate template 326, server system 200 may also consider third-party relevant text, such as the sender's name, which can be used in the salutation section of the query email. After performing this step, server system 200 provides agent 104A with one or more corresponding AI templates 326 as recommendations. Later, client 106A can select one of the recommendations and draft a response email to send to client 106A.

[0128] In one embodiment, reference Figure 4B (See 460), it can be understood that in the example scenario, server system 200 can generate three possible recommended AI templates 462, 464, and 466. In this paper, recommended AI templates 462-466 ​​are... Figure 3 The example of AI template 324 described in [the document]. The first recommended template 462 is based on... Figure 4B The method shown generates the data based on the sender's name extracted from the email signature using the NER model, the shipping number extracted from the customer query email, and the TDR participant name and SCV code extracted from the GCSS portal.

[0129] It is understood that these templates are generated based on historical responses. Furthermore, the templates mask out where information about new customers for future cases would be inserted. Further, in embodiments, two additional types of templates, such as a second recommended template 464 and a third recommended template 466, can be recommended based on the responses to one or more tasks performed in a task set performed to determine the solution for the query. For example, a second recommended template 464 can be generated if the TDR participant is the same as the account that sent the email. A third recommended template 466 can be generated if the TDR participant is anyone other than the account that sent the email.

[0130] Figure 5A flowchart illustrating a method for generating automated responses to a conversational interaction according to embodiments of the present disclosure is shown. The method 500 depicted in the flowchart can be performed, for example, by a server system 200. The sequence of operations of operation 500 may not necessarily be performed in the same order as they are presented. Furthermore, one or more operations may be grouped and performed as a single step, or an operation may have several sub-steps that can be performed in parallel or sequentially. The operations of method 500 and combinations of operations in method 500 can be implemented by, for example, hardware, firmware, processor, circuitry, and / or various means associated with the execution of software comprising one or more computer program instructions. Multiple operations are depicted in the process flow of method 500. This process flow begins with operation 502.

[0131] In 502, method 500 includes receiving customer intent information (e.g., customer intent information 218) from a first machine learning (ML) model (e.g., first ML model 220) for a conversational interaction initiated by a customer (e.g., customer 106A).

[0132] In 504, method 500 includes performing a set of operations by one or more second ML models (e.g., one or more second ML models 222). In one embodiment, the set of operations may include operations 504A-504F.

[0133] In 504A, method 500 includes detecting the presence of one or more sub-intents in customer intent information 218.

[0134] In 504B, method 500 includes, after detecting a sub-intent, assigning a specific automatable process to the conversational interaction, at least in part based on comparing customer intent information 218 with a predefined set of automatable processes.

[0135] In 504C, method 500 includes extracting multiple relevant entities from a conversational interaction based at least in part on one or more relevant texts associated with the conversational interaction.

[0136] In 504D, method 500 includes accessing a list of tasks associated with a specific automatable process from a database (e.g., database 204).

[0137] In 504E, method 500 includes generating query resolution information for conversational interaction based at least in part on customer intent information 218 processed against a task list.

[0138] In 504F, method 500 includes generating one or more conversational interaction response templates (e.g., 462, 464, and 466) for conversational interaction, based at least in part on multiple related entities and query resolution information.

[0139] refer to Figure 5 The disclosed methods or one or more operations of server system 200 can be implemented using software that includes computer-executable instructions stored on one or more computer-readable media (e.g., non-transitory computer-readable media, such as one or more optical discs, volatile memory components (e.g., DRAM or SRAM), or non-volatile memory or storage components (e.g., hard disk drives or solid-state non-volatile memory components, such as flash memory components)) and executed on a computer (e.g., any suitable computer, such as a laptop computer, netbook, webbook, tablet computing device, smartphone, or other mobile computing device). For example, such software can be executed on a single local computer or in a networked environment (e.g., via the Internet, a wide area network, a local area network, a web-based remote server, a client-server network (such as a cloud computing network), or other such networks) using one or more networked computers.

[0140] Additionally, any of the intermediate or final data created and used during the implementation of the disclosed methods or systems may also be stored on one or more computer-readable media (e.g., non-transitory computer-readable media) and is considered to be within the scope of the disclosed techniques. Furthermore, any of the data in the software-based embodiments can be uploaded, downloaded, or remotely accessed via suitable communication means. Such suitable communication means include, for example, the Internet, the World Wide Web (WWW), intranets, software applications, cable (including optical fiber), magnetic communication, electromagnetic communication (including RF, microwave, and infrared communication), electronic communication, or other such communication means.

[0141] Although the invention has been described with reference to specific exemplary embodiments, it should be noted that various modifications and changes can be made to these embodiments without departing from the broad scope of the invention. For example, the various operations, blocks, etc., described herein can be implemented and operated using hardware circuit systems (e.g., complementary metal-oxide-semiconductor (CMOS) based logic circuit systems), firmware, software, and / or hardware, (e.g., embodied in a machine-readable medium) firmware and / or software. For example, the devices and methods can be implemented using transistors, logic gates, and circuits (e.g., application-specific integrated circuit (ASIC) circuit systems and / or digital signal processor (DSP) circuit systems).

[0142] Specifically, server system 200 and its various components may be implemented using software and / or using transistors, logic gates, and circuits (e.g., integrated circuit systems, such as ASIC circuit systems). Various embodiments of the invention may include one or more computer programs stored or otherwise embodied on a computer-readable medium, wherein the computer programs are configured to cause a processor or computer to perform one or more operations. A computer-readable medium storing, embodying, or encoding a computer program or similar language may be embodied as a tangible data storage device storing one or more software programs configured to cause a processor or computer to perform one or more operations. Such operations may be any of the steps or operations described herein, for example. In some embodiments, the computer program may be stored and provided to a computer using any type of non-transitory computer-readable medium. Non-transitory computer-readable medium includes any type of tangible storage medium.

[0143] Examples of non-transitory computer-readable media include magnetic storage media (such as floppy disks, magnetic tapes, hard disk drives, etc.), optical-magnetic storage media (e.g., magneto-optical disks), optical disc read-only memory (CD-ROM), recordable optical disc (CD-R), rewritable optical disc (CD-R / W), digital versatile optical disc (DVD), Blu-ray® optical disc (BD), and semiconductor memory (such as mask ROM, programmable ROM (PROM), erasable PROM (EPROM), flash memory, random access memory (RAM), etc.). Additionally, tangible data storage devices can be embodied as one or more volatile memory devices, one or more non-volatile memory devices, and / or a combination of one or more volatile and non-volatile memory devices. In some embodiments, a computer program can be provided to a computer using any type of transient computer-readable medium. Examples of transient computer-readable media include electrical signals, optical signals, and electromagnetic waves. Transient computer-readable media can provide programs to a computer via wired communication lines (e.g., electrical wires and optical fibers) or wireless communication lines.

[0144] As discussed above, various embodiments of the present invention can be practiced with different sequences of steps and / or operations and / or with hardware elements configured differently from those disclosed. Therefore, although the invention has been described based on these exemplary embodiments, it should be noted that certain modifications, variations, and alternative constructions may be apparent and are fully within the scope of the invention.

[0145] Although various exemplary embodiments of the invention have been described herein in language specific to structural features and / or methodological actions, the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are disclosed as exemplary forms for implementing the claims.

Claims

1. A computer-implemented method, comprising: Receive customer intent information from the first machine learning (ML) model for conversational interactions initiated by the customer; as well as The following set of operations are performed by one or more second ML models: Detect the existence of one or more sub-intents in the customer intent information; Upon detecting a sub-intent, a specific automatable process is assigned to the conversational interaction, at least in part, based on comparing the customer intent information with a predefined set of automatable processes. Multiple related entities are extracted from the conversational interaction, at least in part, based on one or more related texts associated with the conversational interaction; Access the list of tasks associated with the specific automatable process from the database; At least in part, based on processing the customer intent information against the task list, query resolution information is generated for the conversational interaction; as well as Based at least in part on the plurality of related entities and the query resolution information, one or more conversational interaction response templates are generated for the conversational interaction.

2. The computer-implemented method of claim 1, wherein receiving the customer intent information includes: Receive the conversational interaction initiated by the customer; Extract the relevant segment set from the conversational interaction; The first related text in the one or more related texts is generated at least in part based on applying a first set of preprocessing operations to the set of related segments; as well as The first ML model generates the customer intent information for the conversational interaction based at least in part on the first relevant text and historical conversational interaction dataset.

3. The computer-implemented method as described in any one of claims 1 and 2, further comprising: The customer intent information is fine-tuned by the first ML model, at least in part, based on the first relevant text and the first predefined rule set.

4. The computer-implemented method of claim 2, wherein extracting the plurality of related entities comprises: The extraction of a first set of related entities from the plurality of related entities is based at least in part on the following: The second related text is generated from the one or more related texts, at least in part, by applying a second set of preprocessing operations to the first related text; as well as The second related text is processed by one or more second ML models to generate the first related entity set based at least in part on the second related text and a second predefined rule set.

5. The computer-implemented method of claim 2, wherein extracting the plurality of related entities comprises: The extraction of a second set of related entities from the plurality of related entities is based at least in part on the following: The third related text in the one or more related texts is generated at least in part based on applying a third set of preprocessing operations to the first related text; The first relevant text is processed via one or more second ML models to generate an intermediate entity set, at least in part, based on a third predefined rule set; and The third relevant text is compared with the intermediate entity set to generate the second relevant entity set.

6. The computer-implemented method of claim 1, wherein processing the customer intent information for the task list includes: Execute each task from the task list to generate query resolution information for the conversational interaction.

7. The computer-implemented method of claim 1, wherein detecting the presence of the one or more sub-intents in the customer intent information comprises: The detection of two or more sub-intents in the customer intent information is based at least in part on the recognition of one or more conjunction texts in the customer intent information.

8. The computer-implemented method of claim 1, further comprising: The non-automatable process is assigned to the conversational interaction based at least on the customer intent information being associated with at least two or more sub-intents.

9. A server system comprising: Communication interface; A memory configured to store instructions; as well as A processor communicating with the communication interface and the memory, the processor being configured to execute the instructions stored in the memory, thereby causing the server system to at least partially execute: Receive customer intent information from the first machine learning (ML) model for conversational interactions initiated by the customer; as well as The following set of operations are performed by one or more second ML models: Detect the existence of one or more sub-intents in the customer intent information; Upon detecting a sub-intent, a specific automatable process is assigned to the conversational interaction, at least in part, based on comparing the customer intent information with a predefined set of automatable processes. Multiple related entities are extracted from the conversational interaction, at least in part, based on one or more related texts associated with the conversational interaction; Access the list of tasks associated with the specific automatable process from the database; At least in part, based on processing the customer intent information against the task list, query resolution information is generated for the conversational interaction; as well as Based at least in part on the plurality of related entities and the query resolution information, one or more conversational interaction response templates are generated for the conversational interaction.

10. The server system of claim 9, wherein in order to receive the client intent information, the server system is at least partially caused to: Receive the conversational interaction initiated by the customer; Extract the relevant segment set from the conversational interaction; The first related text in the one or more related texts is generated at least in part based on applying a first set of preprocessing operations to the set of related segments; The first ML model generates the customer intent information for the conversational interaction based at least in part on the first relevant text and historical conversational interaction dataset; as well as The customer intent information is fine-tuned by the first ML model, at least in part, based on the first relevant text and the first predefined rule set.

11. The server system of claim 10, wherein in order to extract the plurality of related entities, the server system is at least partially caused to: The extraction of a first set of related entities from the plurality of related entities is based at least in part on the following: The second related text is generated from the one or more related texts, at least in part, based on applying a second set of preprocessing operations to the first related text; and The second related text is processed by one or more second ML models to generate the first related entity set based at least in part on the second related text and a second predefined rule set.

12. The server system of claim 10, wherein in order to extract the plurality of related entities, the server system is at least partially caused to: The extraction of a second set of related entities from the plurality of related entities is based at least in part on the following: The third related text in the one or more related texts is generated at least in part based on applying a third set of preprocessing operations to the first related text; The first relevant text is processed via one or more second ML models to generate an intermediate entity set, at least in part, based on a third predefined rule set; and The third relevant text is compared with the intermediate entity set to generate the second relevant entity set.

13. The server system of claim 9, wherein, in order to process the client intent information for the task list, the server system at least partially causes: Execute each task from the task list to generate query resolution information for the conversational interaction.

14. The server system of claim 9, wherein in order to detect the presence of the one or more sub-intents in the client intent information, the server system is at least partially caused to: Detecting two or more sub-intents in the customer intent information based at least in part on the recognition of one or more connective texts in the customer intent information; and After detecting at least two or more sub-intents, the non-automated process is assigned to the conversational interaction based on the customer intent information being associated with the corresponding two or more sub-intents.

15. A non-transitory computer-readable storage medium comprising computer-executable instructions, which, when executed by at least a processor of a server system, cause the server system to perform a method, the method comprising: Receive customer intent information from the first machine learning (ML) model for conversational interactions initiated by the customer; as well as The following set of operations are performed by one or more second ML models: Detect the existence of one or more sub-intents in the customer intent information; Upon detecting a sub-intent, a specific automatable process is assigned to the conversational interaction, at least in part, based on comparing the customer intent information with a predefined set of automatable processes. Multiple related entities are extracted from the conversational interaction, at least in part, based on one or more related texts associated with the conversational interaction; Access the list of tasks associated with the specific automatable process from the database; At least in part, based on processing the customer intent information against the task list, query resolution information is generated for the conversational interaction; as well as Based at least in part on the plurality of related entities and the query resolution information, one or more conversational interaction response templates are generated for the conversational interaction.